2015-03-01 12:37:09 can somebody tell me how i can generate the truststore for java apps on alpine? 2015-03-01 12:45:06 they call their cert store, trust store? 2015-03-01 12:51:00 looks like that 2015-03-01 12:51:32 i just want to build a minecraft docker container based on alpine linux... but openjdk7 seems to miss certs :( 2015-03-01 12:52:46 it bails out with javax.net.ssl.SSLException: java.lang.RuntimeException: Unexpected error: java.security.InvalidAlgorithmParameterException: the trustAnchors parameter must be non-empty and regarding to google it means that java is not aware of the required root certs. 2015-03-01 12:54:59 ta2p0mm5m4hu: any idea how to solve this openjdk7 issue? 2015-03-01 12:56:13 noooo. i am privileged, i did not touch even a jre for many years. 2015-03-01 12:57:00 i just think it's ironic that they call it truststore 2015-03-01 12:57:10 i dont want to touch this java stuff as well... but i wanted to build a tiny minecraft docker container... currently it looks like i have to switch back to my opensuse build for this java crap 2015-03-01 12:57:41 but it still looks like an openjdk7 package issue 2015-03-01 12:58:59 ubuntu and arch are offering a ca-certificates-java package for that 2015-03-01 13:21:27 bbl 2015-03-02 09:21:01 will tool like http://isthe.com/chongo/tech/comp/calc/ work on AL ? 2015-03-02 11:03:23 vkrishn: Does it build for you? 2015-03-02 11:04:14 not tried on AL 2015-03-02 11:04:35 Does it compile on non-alpine-systems? 2015-03-02 11:05:55 longbits.c:101:11: error: expected '=', ',', ';', 'asm' or '__attribute__' before 'int' E_FUNC int isascii(int c); 2015-03-02 11:06:31 E_FUNC is defined in decl.h … 2015-03-02 11:06:33 no idea… 2015-03-02 11:09:33 yes, I use on knoppix 2015-03-02 11:12:52 changing this E_FUNC tu extern compiles further, but still stops with errors 2015-03-02 11:44:56 compiled calc-2.12.5.0 using 2015-03-02 11:45:05 make clobber && make calc-static-only BLD_TYPE=calc-static-only 2015-03-02 11:45:15 on knoppix 2015-03-02 12:14:52 Hello 2015-03-02 12:15:07 was wondering if there was guidelines to post patches on mailing list ? 2015-03-02 12:18:56 http://wiki.alpinelinux.org/wiki/Creating_patches 2015-03-02 12:20:21 thx barthalion 2015-03-02 12:23:42 barthalion: aports is not limited to packages ? 2015-03-02 12:24:10 oh yeah indeed everything is a package anyway :D 2015-03-02 12:24:20 the world is a package 2015-03-02 12:24:52 it applies to all our projects though, for abuild/apk-tools/etc use alpine-devel mailing list 2015-03-02 12:24:57 you're welcome 2015-03-02 12:25:59 ok will do, let's send myself a test email firt ^^ 2015-03-02 12:43:20 barthalion: ok done, hope it's not too messy :) 2015-03-04 10:07:05 hi, /proc/sys/fs/protected_symlinks doesn't seem to be hardened to a vanilla alpine system. is this by design? 2015-03-04 10:07:25 s/to a/on a 2015-03-04 10:10:39 also, access to /boot/System.map from non-root users is allowed. 2015-03-04 10:11:13 please report it on our bug tracker 2015-03-04 10:11:53 both are most likely real issues, not a design 2015-03-04 10:12:38 barthalion, cool, I am very new to alpine world. I will use the bug tracker :) 2015-03-04 10:13:49 it's nothing wrong to report it here, but by putting issues you've found on bug tracker, you can be sure you won't be forgotten too fast 2015-03-04 10:14:45 true. earlier I was suspecting that I missed reading some design documents and what not :D 2015-03-04 10:18:57 I don't think we have any, to be honest 2015-03-04 10:19:10 all design happens here and on mailing lists 2015-03-04 10:19:45 OpenID integration on the bug tracker seems to be not working (I get the login page again). trying the old registration routine now :) 2015-03-04 10:32:20 barthalion, https://bugs.alpinelinux.org/issues/3962 2015-03-04 10:33:29 Alpine Linux is at the frontier of Linux hardening. It seems that, if things break due to hardening, they break first in Alpine land. Do you guys actively submit patches to upstream projects? 2015-03-04 10:37:35 why do you play with the vanilla? why not the grsec system? is hardening on vanilla even a goal? 2015-03-04 10:54:49 oh my bad, I am running a grsec kernel. Linux slowey 3.14.27-2-grsec. 2015-03-04 10:55:31 I meant to say "unmodified system" by vanilla. 2015-03-04 10:57:50 i actually readed that, instead of tA2p0mM5M4HU interpetation. 2015-03-04 11:14:44 oh? in that case pls file more bugs :) 2015-03-04 11:16:57 clandmeter: so did I :p 2015-03-04 11:17:22 we surely patch some code, no idea if we upstream our changes 2015-03-04 11:17:31 I guess it depends how much time it requires 2015-03-04 11:20:00 ideal case is the bugsubmitter fixes the bug, and upstreams the fix himself. at least that's what i try to do. 2015-03-04 11:20:17 the alpine devs are busy enough 2015-03-04 11:57:45 hopefully, I will find some time (and energy) to do some of the work involved. 2015-03-04 13:13:31 I've successfully built xpra & it's build dependencies (so gui applications inside lxc / docker can be run on a local desktop) - but have the following runtime error on both 32 & 64 bit: 2015-03-04 13:13:35 ImportError: Error relocating /usr/lib/python2.7/site-packages/xpra/x11/gtk_x11/gdk_bindings.so: XkbQueryExtension: symbol not found 2015-03-04 13:14:36 libxkbfile is installed & I've run setup-xorg-base - could this be related to lazy loading by Xorg not being supported under musl ? 2015-03-04 13:36:53 fabled: cool, now fortify covers most of the glibc interfaces + a few more (mostly out of bound reads) 2015-03-04 13:37:17 once we have tested it a bit more, I can release 0.1 2015-03-04 13:37:35 stateless, awesome 2015-03-04 13:37:46 any ideas yet how to tie all that to toolchain? 2015-03-04 13:41:51 never mind - 'setxkbmap -query' is not supported 2015-03-04 13:45:38 fabled: I did send an email to the musl ml but no one really responded, so I guess the way to go is to update the default include path 2015-03-04 13:45:47 and ship it as a separate package 2015-03-04 13:53:08 hmm 2015-03-04 13:53:18 that's part of gcc 2015-03-04 13:53:34 need to talk about that with ncopa when he returns 2015-03-04 13:53:38 hm right 2015-03-04 13:53:40 so sometime next week 2015-03-04 17:27:23 export LDFLAGS="$LDFLAGS -Wl,-z,lazy" - this takes care of lazy loading in musl for xorg applications ? 2015-03-04 17:54:07 BitL0G1c, no, musl does not support lazy 2015-03-04 17:54:16 and that will also disable one of the security features 2015-03-04 18:08:37 fabled: ok - many thx - will remove it 2015-03-05 08:49:48 arg, i built a package but i can't find the APKBUILD file. I must have forgotten to submit a patch :( 2015-03-05 08:54:26 beginner's mistake. found a sprunge, thanks to the IRC logs :) 2015-03-05 14:38:55 whom should I contact to have a ao.org domain for mirror? 2015-03-05 14:50:07 you want to set up a mirror, and just want an A record? 2015-03-05 14:50:18 like dl-7.a.o? 2015-03-05 14:57:38 yeah 2015-03-05 15:00:30 drop a note to alpine-devel with the ip and we'll get ya set up. 2015-03-05 15:01:30 cool, ty 2015-03-06 07:16:28 morning 2015-03-06 07:17:31 o/ 2015-03-06 07:24:49 looks like you have beein doing great while i have been on vacation :) 2015-03-06 07:25:25 thanks alot for keeping the wheels spinning while i was away 2015-03-06 07:27:42 cheers to rnalrd and clandmeter 2015-03-06 08:53:30 ncopa wb :) 2015-03-06 08:53:38 crow: thanks :) 2015-03-06 08:53:40 can someone update edge iso? :) 2015-03-06 08:53:56 yeah, thats a good idea 2015-03-06 08:58:13 ncopa i hope you have had good vication 2015-03-06 09:20:07 anyone in here on pi2 ? 2015-03-06 09:30:51 xen_roger, we changed few emails on alpine-devel about that just recently 2015-03-06 09:31:03 http://lists.alpinelinux.org/alpine-devel/4732.html 2015-03-06 10:12:45 xen_roger i have one, but not setup 2015-03-06 10:13:47 ncopa will it be uploaded here? http://nl.alpinelinux.org/alpine/edge/releases/x86_64/ the edge is 2015-03-06 10:13:49 iso 2015-03-06 10:14:41 http://nl.alpinelinux.org/alpine/edge/releases/x86_64/alpine-edge-150306-x86_64.iso 2015-03-06 10:21:36 fabled: I saw them, 2015-03-06 10:22:20 crow: interessted in yours ... so did the pi2 boot up no matter of setup ;) 2015-03-06 10:23:49 ncopa, wb :) 2015-03-06 10:24:27 mosquitto has a nice release (v1.4) 2015-03-06 10:24:37 xen_roger i did not put alpine on it, i put archlinuxarm just to test if it boots, and yes it does 2015-03-06 10:25:46 but requires some changes/updates, I think also a new dep(libwebsocket) 2015-03-06 10:25:53 how long did you wait for it to ship? 2015-03-06 11:19:01 i did wait 3 days, but ordered inland 2015-03-06 11:24:33 ncopa, _0/ 2015-03-06 11:24:35 back? 2015-03-06 13:09:04 argh 2015-03-06 13:09:32 fabled: does rpc.statd start for you? 2015-03-06 13:09:36 through the initscript 2015-03-06 13:09:59 it seems to hang on a wait4() here (it starts if I run it manually not through /etc/init.d/rpc.statd start) 2015-03-06 13:10:29 no special setup, just an nfs mount 2015-03-06 13:11:01 stateless, have not tested 2015-03-06 13:12:16 ok I might do a bit more testing myself, I suspect this might have to do with openrc 2015-03-06 13:56:06 I use something like http://hastebin.com/ukuqikeweb.erl for detecting usb-drives for my overlay file 2015-03-06 14:07:23 is xen virtualisation posible on alpine?? 2015-03-06 14:10:53 freedomrun, is this what you mean? http://wiki.alpinelinux.org/wiki/Tutorials_and_Howtos#Virtualization 2015-03-06 14:11:58 ah yes alacerda thank you. I`ve been a bit lost by new design. 2015-03-06 14:12:07 ;) 2015-03-06 14:12:11 :) 2015-03-06 14:12:12 np 2015-03-06 18:41:15 is there a variable for ARCH I can use in APKBUILD to give x86 or arm64 ? 2015-03-06 18:41:38 arch? 2015-03-06 18:42:32 but arm64? is that supported? and what hw do you have for arm64? i'm looking to invest into some 2015-03-06 18:49:27 nothing at the moment - just building http://people.redhat.com/sgrubb/audit & trying not to hardcode kernel include paths for $arch 2015-03-06 18:52:18 ah 2015-03-06 19:31:42 $CARCH? sth like that 2015-03-06 19:31:50 check abuild.conf, it might be defined here 2015-03-06 19:32:11 ok - thx barthalion ;-) 2015-03-06 20:14:50 oh, i meant it is called 'arch' 2015-03-06 20:14:59 or is this something different? 2015-03-06 20:15:03 hey barthalion! Got an IP addr for dl-? 2015-03-06 20:39:48 nangel: oops, sorry, didn't have time to reply; 37.187.2.63 2015-03-06 20:43:03 dl-7.a.o is yours! 2015-03-06 20:43:11 and thanks! 2015-03-06 20:45:12 cool, thank you 2015-03-06 20:45:42 I'll bring the vhost up and do the aports part tomorrow 2015-03-06 21:05:26 nangel: should I also add dl-6 to mirrors list? 2015-03-06 21:06:28 seems so 2015-03-06 21:10:56 barthalion: yes, please 2015-03-06 21:11:05 ACTION was fighting with ns2.a.o.... 2015-03-07 09:05:13 hmm, http://nl.alpinelinux.org/alpine/edge/main/x86_64/libc-dev-0.7-r0.apk has not files 2015-03-07 09:06:12 I mean has no .h files into it 2015-03-07 09:10:17 it doesn't need to 2015-03-07 09:11:04 I think something did not go ok with commit e3725c0af137717d6883 2015-03-07 09:11:07 it pulls uclibc-dev or musl-dev depending on the release 2015-03-07 09:11:25 and ships some utils and bsd headers as other subpackages 2015-03-07 09:11:58 and you say so because? 2015-03-07 09:12:06 http://nl.alpinelinux.org/alpine/v3.1/main/x86_64/libc-dev-0.6-r0.apk has 2 .h files 2015-03-07 09:12:22 edge should have 3 2015-03-07 09:12:45 files are in bsd-compat-headers subpackage 2015-03-07 09:12:56 ok 2015-03-07 09:16:28 :(, #warning usage of non-standard #include is deprecated 2015-03-07 09:16:37 would leave that to ncopa 2015-03-07 09:18:44 also upgrading alpine-sdk from v3.1 will not add bsd-compat-headers 2015-03-07 09:21:50 from AL v3.1 to edge 2015-03-07 10:03:55 it's also intended 2015-03-07 10:11:28 ok 2015-03-07 10:26:58 barthalion, had you any chance of looking at 'MariaDB's MaxScale' ? 2015-03-07 10:31:21 https://github.com/mariadb-corporation/MaxScale 2015-03-07 10:44:14 never needed to 2015-03-07 10:44:21 galera works well enough 2015-03-07 11:00:25 could not locate the source of galera code :( 2015-03-07 11:06:08 https://mariadb.com/kb/en/mariadb/what-is-mariadb-galera-cluster/ 2015-03-07 11:10:30 https://downloads.mariadb.org/mariadb-galera - 500 Internal Server Error 2015-03-07 11:10:52 archlinux does not pkg it yet 2015-03-07 11:11:13 sure, because I'm too lazy and have no environment to test it 2015-03-07 11:11:26 I maintain MariaDB there 2015-03-07 11:12:43 hmm, galera is on your local install ? 2015-03-07 11:13:15 I don't use it on Arch/Alpine 2015-03-07 11:13:41 I actually don't use it all anymore, I'm not sysadmin since half a year 2015-03-07 11:14:08 :), just thought that way, debian ? 2015-03-07 11:14:43 ubuntu 2015-03-07 11:27:06 this is something like postgres-bdr ? 2015-03-07 11:27:56 no idea 2015-03-07 22:30:14 Hi all. running into problems getting a dev environment set up for the first time to work on aports. Anyone able to assist? Wiki instructions don't seem to work 100% 2015-03-07 22:31:34 Once I've run setup-alpine, installed sdk, checkout aports, run abuild -r on one, it complains package cache is not enabled 2015-03-07 22:31:45 if I then run sudo setup-apkcache 2015-03-07 22:31:55 everything breaks, gets permission denied on apk-update 2015-03-07 22:32:39 this is with a VM running under KVM from the 3.1.2 release ISO 2015-03-07 22:34:52 Also my second attempt at doing this today, but ended up in the same place the first time 2015-03-07 22:55:08 I should have this open most of the day, so if someone has been through it, or knows where the abuild/cache/permissions intersection is going wrong, I'd love to know 2015-03-07 22:55:22 its extremely frustrating that abuild won't just work OOTB 2015-03-07 23:19:18 wyaeld: - see towards the bottom of http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#Setup_your_system_and_account - you probably have not added your user the the abuild group & logged off & on again 2015-03-07 23:20:09 sudo addgroup abuild 2015-03-07 23:20:09 sudo chgrp abuild /var/cache/distfiles 2015-03-07 23:20:09 sudo chmod g+w /var/cache/distfiles 2015-03-07 23:20:44 @bitlogic I have. have read those pages end-to-end for much of the last 24 hours 2015-03-07 23:20:57 non-root user, added to abuild, group, relogged 2015-03-07 23:21:53 apk update broke on permissions after running sudo setup-apkcache 2015-03-07 23:22:13 which I did, because abuild -r was complaining that package cache wasn't enabled 2015-03-07 23:22:51 which directory is it trying to hit? considering chowning everything I can find to the non-root user 2015-03-07 23:23:01 those instructions work if you use an LXC container as a build environment 2015-03-07 23:23:57 http://wiki.alpinelinux.org/wiki/LXC 2015-03-07 23:24:12 a container is 10 meg 2015-03-07 23:24:17 I originally started with a docker container, but it was running into different errors on things that weren't setup 2015-03-07 23:24:21 haven't tried LXC raw yet 2015-03-07 23:24:39 running a full VM from the iso should work right? 2015-03-07 23:24:46 LXC works great - I have 4 running in kvm 2015-03-07 23:25:27 alpine:/var/cache$ apk update 2015-03-07 23:25:28 ERROR: Unable to lock database: Permission denied 2015-03-07 23:25:28 ERROR: Failed to open apk database: Permission denied 2015-03-07 23:25:48 do you know where the DB it wants to use actually is? 2015-03-07 23:26:29 when i get those messages i normally reboot to clear them 2015-03-07 23:26:53 k, will try 2015-03-07 23:27:57 le sigh. "Boot failed: not a bootable disk". Looks like setup-alpine using the KVM volume toasted it somehow 2015-03-07 23:32:41 wyaeld: - if you just want a build environment setup on alpine lxc container on your host machine 2015-03-07 23:33:34 it also runs fine in KVM with virtio disks under libvirt / virsh 2015-03-07 23:33:53 that's what I'm doing 2015-03-07 23:35:02 just wary of using LXC directly. I've got a lot of experience working with docker, and decent with KVM. New to Alpine though 2015-03-07 23:35:14 you can setup an alpine lxc container on any linux 2015-03-07 23:35:52 this works on debian for example http://wiki.alpinelinux.org/wiki/LXC#Alpine_Template 2015-03-07 23:37:00 with lxc 1.0.7+ just set 'lxc.start.auto = 1' in the container config for autobooting 2015-03-07 23:38:56 & 'rc-update add lxc default' & 'rc-service lxc start' 2015-03-07 23:39:10 for those instructions, where is the actual system that the first steps running APK living? 2015-03-07 23:39:40 by default under /var/lib/lxc/ 2015-03-07 23:40:54 you could also set in /etc/lxc/lxc.conf something like 'lxc.lxcpath = /srv/lxc' if you want them somewhere else 2015-03-07 23:44:16 just use the LXC instructions for your host distribution - & then http://wiki.alpinelinux.org/wiki/LXC#Alpine_Template 2015-03-07 23:44:43 Do you know why if I launch using ISO under virtio, and then reboot, its not bootable? even after setup-alpine -> sys onto the 8GB disk image? 2015-03-07 23:52:35 ok, solved the bootable problem. For some reason setup-alpine defaults to N on one of the disk questions, which I missed 2015-03-07 23:59:57 dammit. Stuck again. Non-root user running ABUILD -r results in complaining about package cache no enabled. 2015-03-08 00:00:18 running SETUP-APKCACHE as non-root is denied 2015-03-08 00:00:38 running as root results in APK UPDATE breaking on the DB access permission 2015-03-08 00:01:07 I've now run into the same problem with Docker, KVM and LXC based installs 2015-03-08 01:15:36 making progress finally, abuild building but failing 2015-03-08 01:35:21 le sigh. Everything working, but google wants to block the SMTP because not secure enough 2015-03-08 01:48:39 patchset sent, le victory! 2015-03-08 13:41:58 hm... http://lists.alpinelinux.org/alpine-aports/0122.html anyone? 2015-03-08 14:02:50 jomat: i think the correct url of pass is: http://www.passwordstore.org/ 2015-03-08 14:03:32 mh 2015-03-09 07:36:17 ncopa, I send libwebsockets patch to ML, was wanting to use that in mosquitto 2015-03-09 07:36:46 does that mean adding to testing and then move to main ? 2015-03-09 07:37:42 yes 2015-03-09 07:37:46 i'm looking into it 2015-03-09 07:38:00 i'm also interested in mqtt websocket support :) 2015-03-09 07:43:03 also changes in mosquitto APKBUILD http://sprunge.us/iaMV 2015-03-09 07:43:56 I tried building it, seems all ok, need to test today, I tried paho lib, but think it may have some bugs 2015-03-09 07:50:04 libwebsockets also has some files in doc/ dir, that I have not pkg'ed 2015-03-09 07:51:20 also bump monit to v5.12 pls 2015-03-09 09:29:54 thats weird, I don't have alpine-devel MLs for dec18-dec21(2014) 2015-03-09 09:42:28 stateless, i'll probably take a look at your fortify stuff this week 2015-03-09 09:43:16 current plan is to add /usr/include/fortify to gcc built-in specs, and add fortify-headers (new pkg) to build-base depends 2015-03-09 09:43:32 you see any problems with that? 2015-03-09 09:43:47 maybe you can also tag new version with the latest fixes? 2015-03-09 10:51:08 Just wondering, is there any intention/motivation to move away from GCC in Alpine Linux? 2015-03-09 10:51:46 at some point, probably 2015-03-09 11:03:29 ScrumpyJack, we've talked about it earlier, and at the time llvm (which is the only real candidate atm) was not up to it 2015-03-09 11:03:33 it's improved since 2015-03-09 11:03:42 so maybe in 0,5-1 years we might consider it again 2015-03-09 11:25:26 ok 2015-03-09 11:25:48 stateless, adding the external fortify directory, e.g. /usr/include/fortify, seems simple. just needs a gcc patch 2015-03-09 11:25:59 though i wonder where to add -D_FORTIFY_SOURCE=2 default 2015-03-09 11:26:07 perhaps in specs 2015-03-09 11:26:23 i also wonder how that affects compiling kernel 2015-03-09 13:22:27 ncopa, did websocket test, works :)) 2015-03-09 13:22:40 pls push libwebsockets to main/ 2015-03-09 13:22:59 and also changes in mosquitto APKBUILD http://sprunge.us/iaMV 2015-03-09 13:24:39 vkrishn: nice 2015-03-09 13:24:43 i'll do so soonish 2015-03-09 13:24:48 thanks 2015-03-09 13:26:20 if needed I can send full patch for mosquitto changes 2015-03-09 13:38:02 stateless: do you mind rename your 'fortify' project to 'fortify-headers'? 2015-03-09 13:38:14 re taked about what to call the apk package for it 2015-03-09 13:38:30 and concluded that 'fortify-headers' is technically a better name 2015-03-09 13:38:38 but we try stick to what upstream calls it 2015-03-09 13:38:47 so we though better have upstream to rename it ;) 2015-03-09 13:51:04 ncopa, stateless: patches for fortify support in alpine: http://sprunge.us/jGPS 2015-03-09 13:51:09 seems to work 2015-03-09 13:51:26 though, fortify-headers should probably get new release, or we should cherry-pick the strncpy fix 2015-03-09 13:51:28 cool! 2015-03-09 13:51:40 http://git.2f30.org/fortify/commit/?id=d969016e9a68e396ffae0582ad0d4ea5135ed8a4 2015-03-09 13:52:07 i'd say make new release and rename it to fortify-headers ;) 2015-03-09 13:52:26 i wonder if kernel compiles still 2015-03-09 13:52:47 gentoo seems to do the same 2015-03-09 13:53:05 if someone does not like fortify they can still -U_FORTIFY_SOURCE; and i think kernel should do that 2015-03-09 13:56:22 then again, kernel should not be using libc includes so it should be ok :) 2015-03-09 13:56:37 coffe 2015-03-09 15:15:36 Is it intentional that apk doesn't exit with an error when it can't install something? 2015-03-09 15:16:00 https://www.irccloud.com/pastebin/DF5cX0vB 2015-03-09 18:46:29 hi, can someone point me in the right direction for compiling gems? I've got ruby installed on my image using the `opkg-install ruby-gems` package, but when i try to do a `gem install haml`, i get this error: http://d.pr/i/1awBD/3qL5aAPg 2015-03-09 20:45:25 I'm gonna push new io.js to testing 2015-03-09 20:45:39 I don't have an env to test the build, so I'll revert the commit if it fails 2015-03-09 20:46:29 I really need to set up personal CI with buildbot 2015-03-09 21:04:59 barthalion: cat - > APKBUILD && git push .... :) 2015-03-09 21:18:47 Yep, a FAQ for aport contributions would be useful. 2015-03-09 21:24:28 Looks like it's in the wiki. 2015-03-10 06:23:00 stateless, ping 2015-03-10 07:35:04 stateless, multiple definition of `__fortify_ppoll': needs 'static' 2015-03-10 09:15:44 ncopa, what port on msg.a.o is planned for websocket ? 2015-03-10 09:16:58 depends on when i get time 2015-03-10 09:17:03 probably after v3.2 release 2015-03-10 09:19:52 ok, but nice if possible earlier 2015-03-10 09:54:52 bah 2015-03-10 10:09:08 \o/ someone committed the s6 suite! thanks! 2015-03-10 11:58:05 uggedal: your edge image seems to be broken 2015-03-10 11:58:17 uggedal: I can't upgrade openssl thanks to stupid upstream 2015-03-10 12:08:14 ping ncopa 2015-03-10 12:08:26 hi jarruda 2015-03-10 12:09:36 hi ncopa... there is any open issue about chrony failing to start in AL 3.1.2? 2015-03-10 12:09:44 in a box with bonding 2015-03-10 12:10:01 I have it freezing on startup... 2015-03-10 12:11:20 another box running a similar network config but in AL 3.0.6 (chrony-1.29.1-r0) it starts fine 2015-03-10 12:11:45 jarruda, ncopa i have the same in a non bonding machine 2015-03-10 12:12:07 just one nick and chrony freezes 2015-03-10 12:12:13 nic* 2015-03-10 12:12:48 i tried to upgrade the chrony on this AL 3.1.2 to the edge version (chrony-1.31-r0) and it fails to 2015-03-10 12:14:42 jarruda: is the network working? 2015-03-10 12:15:11 yes... we can ping the box.. 2015-03-10 12:15:27 i have seen chrony hang if network setup is brok 2015-03-10 12:15:59 can you disable chrony at boot, make machine boot up and then try start chrony? 2015-03-10 12:16:05 does it still hang after that? 2015-03-10 12:17:40 alacerda, can you try this? 2015-03-10 12:18:00 jarruda, trting 2015-03-10 12:18:01 ;) 2015-03-10 12:18:07 trying* 2015-03-10 12:25:01 is it ok if after 1. apk add git , 2. apk del git , /etc/ssl/certs directory persists ? 2015-03-10 12:28:46 jarruda, ncopa: even if the network is working properly it doesn't start 2015-03-10 12:30:56 does it hang? 2015-03-10 12:31:03 yes 2015-03-10 12:31:18 can you try start it maually from cmd line and disable forking? 2015-03-10 12:31:24 try debug it 2015-03-10 12:31:28 or strace it 2015-03-10 12:31:30 rc-service chronyd start 2015-03-10 12:31:31 * Starting chronyd ... 2015-03-10 12:31:32 try figure why it hangs 2015-03-10 12:31:45 strace? 2015-03-10 12:32:28 wow!!! I didn't know this strace 2015-03-10 12:33:13 try run it in foreground 2015-03-10 12:34:48 chronyd -n -d 2015-03-10 12:38:01 chronyd -n -d -d 2015-03-10 12:48:11 jarruda, did you chack chronyd 1.29 ? 2015-03-10 12:48:16 check 2015-03-10 12:49:50 alacerda, I couldn't find a way to install chronyd 1.29 in the AL 3.1.2 2015-03-10 12:50:14 but if it's possible would be good to check 2015-03-10 12:54:58 hum 2015-03-10 12:55:06 i have chrony-1.31 here on my desktop 2015-03-10 12:55:09 and it works here 2015-03-10 12:55:54 weird 2015-03-10 12:55:56 =/ 2015-03-10 12:56:07 only reason chrony will "hang" is if it can't find the NTP servers 2015-03-10 12:56:29 that indicates that either it doesn't have a route to the servers (no ip, no default route, etc) 2015-03-10 12:56:30 nangel, i tcpdumped our NTP server and the requests reach it 2015-03-10 12:56:32 or DNS is down 2015-03-10 12:57:12 alacerda: and do the responses reach the device chrony is running on? 2015-03-10 12:57:12 alacerda: does tcpdump show that the ntp server responds? 2015-03-10 12:57:29 yeah, that is what i wondered, if routeback might be broken 2015-03-10 12:59:05 nangel, ncopa the route back is there since i can ping the server and it answers me 2015-03-10 12:59:32 do you have a firewall on the chrony device that blocks port 123? 2015-03-10 12:59:35 udp port 123? 2015-03-10 13:00:56 oops.. both respond 2015-03-10 13:01:11 the udp 123 port is open on the chrony device... 2015-03-10 13:07:24 fabled, if I apk add pkgA , having 5 dep, they are installed as they get downloaded or after all pkgs+deps are downloaded ? 2015-03-10 13:13:20 Any ideas on https://www.irccloud.com/pastebin/DF5cX0vB ? 2015-03-10 13:15:46 This is, at a guess, a socket returning EAGAIN. apk skips to the next package and reports an error later, but doesn't exit with an error 2015-03-10 13:20:39 will it slow down, if apk add, first downloads then installs ? 2015-03-10 13:27:16 squaremo: does it help with apk update 2015-03-10 13:30:39 ncopa: Should I try apk update separately before apk add? 2015-03-10 13:31:00 yes 2015-03-10 13:31:03 or apk add -U ... 2015-03-10 13:31:13 which is same thing effectively 2015-03-10 13:31:39 Ah OK. I am running `apk --update ...` 2015-03-10 13:32:42 Pertinent information: this is being run as part of a docker image build; i.e., as a RUN step 2015-03-10 13:33:40 based on the gliderlabs/alpine-linux image (I've reported it to gliderlabs) 2015-03-10 13:34:51 wth is gliderlabs… 2015-03-10 13:35:39 ah, ok, looks like sane docker image 2015-03-10 13:36:03 without adding rootfs from tarball 2015-03-10 13:36:05 cool 2015-03-10 13:36:17 ncopa: we should recommend it on www.a.o 2015-03-10 13:37:03 btw, i got some email while on vacation that they handed over the nasty docker repo to me 2015-03-10 13:37:10 cool #2 2015-03-10 13:37:21 i should check that out... 2015-03-10 13:38:33 Oh interesting, if I do `apk fetch ...` first, it'll exit properly if it can't fetch something. 2015-03-10 13:41:40 ncopa: so, we have whole "alpinelinux" account now? 2015-03-10 13:41:48 yes 2015-03-10 13:41:55 and the bitbucket too as i understand 2015-03-10 13:42:03 (there still remains the problem that it basically always fails ..) 2015-03-10 13:42:05 they said they sent me the bitbucket passwd 2015-03-10 13:42:09 I can take care of it, once I figure out sane CI 2015-03-10 13:42:18 (docker hub that is) 2015-03-10 13:42:19 barthalion: would be great 2015-03-10 13:43:26 can you forward me that message? 2015-03-10 13:43:36 yeah 2015-03-10 13:43:43 great, tu 2015-03-10 13:43:45 ty even 2015-03-10 13:44:03 they sent me the bitbucket passwd ecrypted with my pub key 2015-03-10 13:45:17 I think we can remove bitbucket account, or just have it as a place holder maybe 2015-03-10 13:45:24 I don't think we need it 2015-03-10 13:46:10 ok 2015-03-10 13:47:36 barthalion: do you have a docker account? 2015-03-10 13:47:53 yep 2015-03-10 14:18:21 that gliderlabs image looks really good 2015-03-10 14:48:48 on the other hand, they are making packages and not contributing back 2015-03-10 14:48:51 not good 2015-03-10 14:48:57 no cookies for them 2015-03-10 15:16:02 barthalion: the package changes made so far don't actually work, i'm planning to post patches to the mailing list once they are in a better shape 2015-03-10 16:33:14 hey 2015-03-10 16:33:21 I will rename to fortify-headers 2015-03-10 16:33:27 + make a 0.2 release with the ppoll fix 2015-03-10 16:33:28 :D 2015-03-10 16:36:13 andyshinn: great to hear; I'm mostly talking about golang 2015-03-10 16:38:10 yea, someone reported the Golang 1.4 package wasn't working yet :/ 2015-03-10 16:46:08 I was planning to take a look, but then I saw the number of patches needed for 1.3 and realized that I have no time for that 2015-03-10 16:46:44 fabled: http://dl.2f30.org/releases/fortify-headers-0.1.tar.gz 2015-03-10 16:46:55 and the repo: http://git.2f30.org/fortify-headers/ 2015-03-10 16:47:03 the release contains the ppoll fixes as well, removed the old 0.1 release entirely 2015-03-10 16:47:11 not sure if fabled is Timo but I also replied by email :) 2015-03-10 16:48:04 yup, bingo 2015-03-10 17:02:27 so have somebody created an elasticsearch package in the meantime? :) 2015-03-10 17:07:02 or maybe somebody can give me an example how to package a java application correctly on alpine? :) 2015-03-10 17:23:31 mosez: have you already looked at the packaging articles on wiki.alpinelinux.org/ 2015-03-10 17:27:42 mjones_: general packaging guides, yes. but i can not see something java related. 2015-03-10 17:52:46 stateless, i think ppoll needs to be inline too.. just like all the other funcs 2015-03-10 17:53:04 static non-inline will likely cause warning when unused 2015-03-10 17:55:55 mosez: ok, you've seen that. I don't know anything past those on the subject of java on alpine. 2015-03-10 17:56:19 mosez: I work with a lot of java webapps, but have never packaged them in apk for alpine. 2015-03-10 17:57:38 mosez: I have very strong biases towards openjdk, and I suggest that the minimum version you should use is the latest stable of OpenJDK 7 -- 8 might be good now, too. 2015-03-10 17:58:12 Hopefully you don't have a political or technical requirement to package commercial Oracle java blobs. 2015-03-10 18:01:29 mosez, i don't think we have any real java packages atm. 2015-03-10 18:03:14 Anyone have an opinion on whether java8/openjdk8 is ready for production use? I'm generally conservative with versions of java runtimes... 2015-03-10 18:03:43 fabled: ah you are right 2015-03-10 18:03:53 ...but the recent history of security bugs, plus delayed entry of several important TLS features into Java, has me changing my mind. 2015-03-10 18:05:02 fabled: ok pushed 0.2 :) 2015-03-10 18:05:26 will update dl.2f30.org later with the package 2015-03-10 18:44:41 fabled: http://dl.2f30.org/releases/fortify-headers-0.2.tar.gz 2015-03-10 18:45:08 stateless, thanks. i'll look at it more tomorrow 2015-03-10 18:45:12 :) 2015-03-10 18:45:25 i think we hope to include it sometime this week 2015-03-10 18:47:29 awesome :D 2015-03-10 18:48:11 so the pending patches for aports are: http://sprunge.us/ChNi 2015-03-10 18:48:28 if all is good, i'll push them a bit later this week 2015-03-10 18:49:13 i also finally got to fixing xattr stuff on apk 2015-03-10 18:49:21 seems like it's gonna be productive week :) 2015-03-10 18:49:31 it might be worthwhile to mention somewhere that crashes with "illegal instruction" are probably related to fortify 2015-03-10 18:49:38 yes 2015-03-10 18:49:39 or to keep it in mind 2015-03-10 18:50:29 we cannot print any more descriptive messages because no functions calls are permitted at that point 2015-03-10 18:50:33 or syscalls for that matter 2015-03-10 18:50:53 well 2015-03-10 18:51:01 fabled: the git repo url in the commit message needs to be updated, but looks good! 2015-03-10 18:51:30 i think with fortify we are not guaranteed to be hosed. so calling something would not be catastrophe 2015-03-10 18:51:40 with SSP it's worse 2015-03-10 18:51:52 any particular reason why it defaults to _FORTIFY_SOURCE=2? 2015-03-10 18:51:56 mosez: I'm looking at Arch's elasticsearch atm 2015-03-10 18:52:02 but it's still probably better to just crash explicitly with fortify too 2015-03-10 18:52:08 on multiple levels 2015-03-10 18:52:10 level 2 is handled in the same way as level 1 but may be misleading to people reading this, assuming that it will trap valid but suspicious programs 2015-03-10 18:52:21 stateless, the patch is from gentoo; they seem to default to level2 2015-03-10 18:52:21 mosez: first, it should be too hard to convert PKGBUILD to APKBUILD 2015-03-10 18:52:28 mosez: (sadly still a manual thing) 2015-03-10 18:52:30 fabled: yeah SSP is much worse indeed 2015-03-10 18:52:44 ok :) 2015-03-10 18:53:52 stateless, thanks. updated commit message. 2015-03-10 18:54:07 mosez: the one thing that would be needed is getting rid of systemd services; I bet that gentoo has openrc daemon for ES 2015-03-10 18:58:05 barthalion: Yes, it would be smart to start with arch's package I think. Since it's manual, a wiki HOWTO would seem to be useful. 2015-03-10 18:58:43 mosez: omg, of course I mean't that it shouldn't bee to hard 2015-03-10 18:59:03 mjones_: I'm too lazy to write the docs 2015-03-10 18:59:18 you just need to keep in mind that PKGBUILDs are written in bash 2015-03-10 18:59:27 while APKBUILD in POSIX shell 2015-03-10 19:00:21 fabled: in retrospect, this fortify would have trapped on heartbleed I think because of the out of bounds send() or sendto() 2015-03-10 19:00:34 glibc fortify doesn't add checkers for out of bounds reads 2015-03-10 19:01:15 stateless, not necessarily 2015-03-10 19:01:37 i don't remember details; but i do remember it involved internal buffer caching in openssl 2015-03-10 19:01:53 but yeah, it's still gonna do a lot good 2015-03-10 19:05:20 oh, gentoo's gcc doc change is wrong. -O1 (or higher), or -Os is enough to enable fortify 2015-03-10 19:05:37 no -O, or -O0 does disables it, as well as -U_FORTIFY_SOURCE 2015-03-10 19:06:20 yup 2015-03-10 19:10:10 stateless: as @fabled says, openssl was/is a special case 2015-03-10 19:10:39 stateless: static analysis didn't catch heartbleed because openssl uses/used its own perverted memory allocation routines 2015-03-10 19:11:18 @fabled: any data on performance impacts? 2015-03-10 19:13:04 no, not yet. 2015-03-10 19:13:09 @fabled: I happened to be reading this yesterday, which is likely worth a pointer from the release notes: http://mainisusuallyafunction.blogspot.com/2012/05/automatic-binary-hardening-with.html 2015-03-10 19:13:21 assuming gcc optimizes the branching correct, it should be pretty minimal 2015-03-10 19:14:16 @fabled: and this: http://www.trapkit.de/tools/checksec.html 2015-03-10 19:17:58 mjones_: ah right yeah 2015-03-10 19:19:05 I'm intrigued by x32 arch for reduced memory use, particularly in VM guests. However it has the potential to confuse some people and add a lot of extra resources/work. 2015-03-10 19:19:42 mostly the latter is a problem 2015-03-10 19:22:11 barthalion: if testing it automated, it wouldn't require much maintenance work once it was initially completed, I assume? 2015-03-10 19:22:42 just a bunch more storage for repo mirrors, and more resources for the CI, I guess. 2015-03-10 19:23:03 once it was initially completed, of course. 2015-03-10 19:23:10 we first need someone to write that automated QA 2015-03-10 19:23:40 barthalion: ah. Is there a pointer to doc on the current status of that? 2015-03-10 19:23:52 a pointer? no 2015-03-10 19:24:01 there is nothing to point to 2015-03-10 19:24:45 I'm intrested, but I haven't done a project before that I would consider sufficiently similar. 2015-03-10 19:25:08 I guess I should see what the other distros have. 2015-03-10 19:25:34 we have CI pipeline for packages, and I think that's it 2015-03-10 19:26:08 Edge must get built by a similar pipeline? 2015-03-10 19:36:56 yeah, we do that for every release 2015-03-10 19:39:17 fabled: btw our fortify does not protect against format string bugs because the checks cannot be done without potentially breaking valid programs 2015-03-10 19:39:33 compiler assistance would be required to do this properly 2015-03-10 19:41:27 and it would make it more difficult to port it to non-linux systems (the current implementation works even on Solaris 2015-03-10 19:41:49 glibc checks whether the format string contains %n and if it is in writeable memory, then it traps 2015-03-10 19:42:19 (only with fortify level 2) 2015-03-10 20:49:26 what's the best was to handle error.h includes when compiling for Alpine Linux with musl? 2015-03-10 20:56:14 Depends. For many situations, something like #ifdef _GLIBC #include error.h 2015-03-10 20:56:49 The define that might be corret is _GNU_SOURCE but don't take my word on that. 2015-03-10 20:57:00 Is this for an aport diff or fresh code or what? 2015-03-10 21:03:15 if you grep -r error.h ~/aports you will find a header you can use in abuild 2015-03-10 21:05:19 look in main/kbd/APKBUILD 2015-03-10 21:34:33 thanks gents 2015-03-10 21:45:02 arg, storage size oh 'blah' is not known 2015-03-10 21:45:07 s/oh/of 2015-03-10 21:45:24 musl specific is suppose 2015-03-10 21:46:12 Could be compiler security options? 2015-03-10 21:47:28 it fails on finding: struct timeval tv; 2015-03-10 21:48:32 storage size of tv is not known 2015-03-10 21:49:42 http://www.gnu.org/software/libc/manual/html_node/Elapsed-Time.html 2015-03-10 21:53:19 ScrumpyJack: source use #include or - looks like it should be in musl 2015-03-10 21:53:48 Yeah that's defined in /usr/include/linux/time.h, which you should have if you have headers installed 2015-03-10 21:57:11 nangel: the source uses #include 2015-03-10 21:57:50 stat /usr/include/time.h 2015-03-10 21:58:09 "no such file or directory" would be a problem 2015-03-10 22:00:02 it's there 2015-03-10 22:00:08 :/ 2015-03-10 22:00:37 im compiling davfs2 to package it up 2015-03-10 22:06:16 nangel: I'm not seeing timeval in /usr/include/time.h. Am I being silly? It's in /usr/include/linux/time.h, and in some kernel headers, though. 2015-03-10 22:07:00 uh. you are right. 2015-03-10 22:07:19 (and in bits/time.h which is included from /usr/include/time.h....depending on compiler directives) 2015-03-10 22:08:09 ScrumpyJack: temporarily add "#include " and see if it works. 2015-03-10 22:09:17 Scrubbing GNU-isms can be painful but is oh so satisfying. 2015-03-10 22:10:10 gcc was probably the most important piece of non-BSD and non-Linux open-source software to really make open-source viable. But that doesn't necessarily mean gcc is great software. 2015-03-10 22:10:46 +1 2015-03-10 22:11:07 Those curious can look up the egcs fork and how it came to replace gcc, or read about Stallman's political mandates preventing gcc from being modular. 2015-03-10 22:12:29 ScrumpyJack, if its any consolation, I'm also currently wrestling gcc to compile rtpproxy 2.0.0 2015-03-10 22:12:35 and losing 2015-03-10 22:17:02 mjones_: apk add clang llvm ? :) 2015-03-10 22:17:26 nangel: I was just thinking that. 2015-03-10 22:18:11 I wouldn't necessarily think it would solve the problem itself, but it can't hurt in helping the debug. 2015-03-10 22:19:01 ACTION dislikes monoculture and is quite relieved that clang is now viable for production use, and used in parallel CI in many organisations. 2015-03-10 22:22:42 ScrumpyJack: you should pull in sys/time.h for timeval 2015-03-10 22:22:45 not time.h 2015-03-10 22:22:59 it seems like 50% of the packages I build need me to talk upstream or patch 2015-03-10 22:23:14 stateless: thanks, i'll try that 2015-03-10 22:23:58 sys/time.h should work, but I would expect that directory to be in the compiler directive...? 2015-03-10 22:24:35 mjones_: it is a system header provided by musl, as long as build with the default include header path, it should pick it up? 2015-03-10 22:24:37 ScrumpyJack? What are you building? I decided to do ceph and its prerequisites are fairly brutal. 2015-03-10 22:24:50 lol... ScrumpyJack: take a look at main/sic/musl-fix.patch 2015-03-10 22:25:05 nangel: oops! ;) 2015-03-10 22:25:26 Scrumpyjack - install findutils & 'sudo updatedb' - & 'locate xxx.h' to help you find musl headers 2015-03-10 22:26:20 I just use find, even when locate is installed. Old habit I guess. 2015-03-10 22:26:32 BitL0G1c: I'm good with find :) 2015-03-10 22:27:46 nangel: lol, thanks for that. I'll look in aports more. Someone is bound to have seen it elsewhere 2015-03-10 22:28:19 I got a kick that ncopa actually included the full error report in the patch 2015-03-10 22:28:39 like he knew others would run into it. 2015-03-10 22:28:49 sometimes I forget a workaround I fixed :) 2015-03-10 22:29:48 Yeah, I usually do that to help myself almost as much as to help others. 2015-03-11 08:54:39 ncopa, you think we are good to push stateless' fortify-headers stuff to edge ? 2015-03-11 08:58:12 yes 2015-03-11 09:01:03 i'm testing some more ethings still 2015-03-11 09:05:12 stateless, trying to rebuild gcc with fortify: 2015-03-11 09:05:15 /usr/include/fortify/string.h: In function 'void* __fortify_memcpy(void*, const void*, size_t)': 2015-03-11 09:05:15 /usr/include/fortify/string.h:13:12: error: invalid conversion from 'void*' to 'char*' [-fpermissive] 2015-03-11 09:05:15 char *d = dest; 2015-03-11 09:05:15 ^ 2015-03-11 09:05:15 /usr/include/fortify/string.h:14:18: error: invalid conversion from 'const void*' to 'const char*' [-fpermissive] 2015-03-11 09:05:18 const char *s = src; 2015-03-11 09:05:44 why not just use 'void *' ? 2015-03-11 09:05:54 pointer arithmetics works with void* too 2015-03-11 09:06:09 in fact, you don't need those temp vars at all 2015-03-11 09:10:47 hey fabled 2015-03-11 09:11:42 hmm 2015-03-11 09:11:43 now i get 2015-03-11 09:11:44 /usr/include/sched.h:72:7: error: conflicting declaration of 'void* __fortify_memcpy(void*, const void*, size_t)' with 'C' linkage 2015-03-11 09:11:44 void *memcpy(void *__restrict, const void *__restrict, size_t); 2015-03-11 09:12:06 /usr/include/fortify/string.h:10:1: note: previous declaration with 'C++' linkage 2015-03-11 09:12:20 seems it's first included as-is, and then from inside extern "C" block 2015-03-11 09:12:25 we need the extern "C" stuff right? 2015-03-11 09:12:28 yes 2015-03-11 09:12:41 btw void * is not defined by the standard but since we are relying on GCC I guess we can use it? 2015-03-11 09:12:49 (that's why I did not use it initially) 2015-03-11 09:12:53 but this is an issue with C++ code 2015-03-11 09:13:02 we need either explicit casts or just use void * 2015-03-11 09:13:45 StarWarsFan|afk, http://sprunge.us/eXLd 2015-03-11 09:13:49 stateless, ^^ 2015-03-11 09:14:02 i'd just use void * directly 2015-03-11 09:14:20 though 2015-03-11 09:14:35 it's probably GCC extension to allow 2015-03-11 09:14:41 yeah it is 2015-03-11 09:14:45 but you have all the __attribute__ stuff there anyway... 2015-03-11 09:14:48 ok can you confirm that extern "C" works? 2015-03-11 09:14:52 yeah it is not an issue here 2015-03-11 09:15:27 though, if you want to play it safe, you can do explicit cast on the pointer types 2015-03-11 09:15:31 that should fix the warning 2015-03-11 09:15:40 gonna use void * directly for now 2015-03-11 09:18:50 yes, extern "C" { and matching } with #ifdef __cplusplus fixes the other error 2015-03-11 09:18:54 you'll need it in all the files 2015-03-11 09:19:29 yeah 2015-03-11 09:19:31 adding it now 2015-03-11 09:19:47 did you put it below the include_next? 2015-03-11 09:20:32 include/string.h:16:28: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] 2015-03-11 09:20:33 i put it inside the _FORTIFY_SOURCE #ifdef in my test, but it would not make a difference 2015-03-11 09:20:38 :( 2015-03-11 09:20:40 with g++ 2015-03-11 09:20:46 oh, ok 2015-03-11 09:20:51 I will put the cast for now 2015-03-11 09:20:52 then add explicit casts 2015-03-11 09:27:06 fabled: ok I pushed the ifdef guards + the explicit casts 2015-03-11 09:27:15 let me know if it breaks 2015-03-11 09:27:26 if you do not find any other issues I will tag a release 2015-03-11 09:27:26 :) 2015-03-11 09:29:13 stateless, ok, retrying gcc rebuild with fortify git head 2015-03-11 09:48:36 stateless, gcc still fails 2015-03-11 09:48:52 libitm seems to have c++ class with field/accessor named memset 2015-03-11 09:49:06 and the define overrides it in some headers, but not all 2015-03-11 09:55:42 stateless, seems glibc solves this by using __attribute__ ((__artificial__)) in all fortify functions 2015-03-11 09:55:49 and declaring them directly as the main symbol name 2015-03-11 09:56:04 ah 2015-03-11 09:56:15 cool 2015-03-11 09:56:18 will add that :) 2015-03-11 09:56:28 but I need to run now, I will be back in a few hours 2015-03-11 09:58:29 hmm 2015-03-11 09:58:36 stateless, though, that's maybe not the solution 2015-03-11 09:58:42 seems it affects debugging stuff only 2015-03-11 10:11:55 in c++ the rule that standard c functions may be defined as macros is not valid 2015-03-11 10:12:11 (yes this is a huge incompat in c++) 2015-03-11 10:12:26 so the solution is to disable fortify hacks ifdef __cplusplus 2015-03-11 10:12:50 (or use some c++ specific workaround in that case) 2015-03-11 10:27:26 btw in c++ the macro wouldnt work anyway when std::memset is used because there is no std::__fortify_memset 2015-03-11 10:27:58 i think you need cstring and string.h wrappers too for c++ 2015-03-11 10:39:59 there is too much bot spam here 2015-03-11 10:43:17 i suppose we could drop git commits in here 2015-03-11 10:44:02 I'd rather not 2015-03-11 11:00:46 stuff git commits to #alpine-commits or similar? 2015-03-11 11:05:36 well i can ignore the bot 2015-03-11 11:06:43 it's useful sometimes. 2015-03-11 11:10:13 hm 2015-03-11 11:10:27 nsz: thanks for the input 2015-03-11 11:10:41 should we just ignore C++ in the interim? 2015-03-11 11:11:15 not ideal I know but it requires an entire rework as the entire implementation relies on override functions with macros being legal 2015-03-11 11:13:18 i think you will have to ignore c++ 2015-03-11 11:16:17 i'm not sure what glibc does to get around c++ issues 2015-03-11 11:16:26 maybe it uses this extern inline thingie 2015-03-11 11:16:35 it doesn't override using macros 2015-03-11 11:16:49 extern inline is c99 semantics 2015-03-11 11:16:53 ah 2015-03-11 11:16:55 c++ inline works differently 2015-03-11 11:17:11 right ok, I am really not familiar with C++ at all 2015-03-11 11:17:14 but if glibc uses some magic inline attr to make gcc respect extern inline.. 2015-03-11 11:17:38 I will push a patch that ignores C++ for now 2015-03-11 11:20:55 fabled: http://git.2f30.org/fortify-headers/commit/?id=14af218559243c1e16cf871e1823f1a5d8cece9f 2015-03-11 11:21:06 should at least not break code now 2015-03-11 11:22:53 stateless, ok, i'll retry soon 2015-03-11 11:22:54 thx 2015-03-11 11:23:17 np :) 2015-03-11 11:23:25 nsz: btw, did you already move in the new building? 2015-03-11 11:23:32 I haven't seen you at lunch for some time :) 2015-03-11 11:25:51 yes 2015-03-11 11:26:07 i'm in the new building since early febr 2015-03-11 11:26:09 ah cool :) 2015-03-11 11:26:16 is it packed? :P 2015-03-11 11:26:38 no, i dont think so 2015-03-11 11:26:54 but kitchen and some auxilary spaces are a bit small 2015-03-11 11:26:57 ah right 2015-03-11 11:27:19 is it in the city centre somewhere? 2015-03-11 11:27:23 (however it's a completely open work environment without walls.. can be distracting sometimes) 2015-03-11 11:27:36 noo 2015-03-11 11:27:38 where do you work? 2015-03-11 11:27:47 it's cpc1 behind tesco 2015-03-11 11:27:48 barthalion: ARM 2015-03-11 11:27:52 aha cpc ok 2015-03-11 11:28:10 ah 2015-03-11 11:28:21 I heard they have burrito fridays or something 2015-03-11 11:28:22 :P 2015-03-11 11:28:31 or thursdays 2015-03-11 11:29:24 well i think the food portion is small whatever day it is :P 2015-03-11 11:29:31 totally 2015-03-11 11:30:17 just wondering if you met Marcin Juszkiewicz a year or two ago, but it's probably long shot 2015-03-11 11:30:29 I started 7-8 months ago 2015-03-11 11:30:30 :P 2015-03-11 11:30:50 though so :p 2015-03-11 14:41:04 please git pull git://git.alpinelinux.org/fab/aports 2015-03-11 14:41:06 thanks 2015-03-11 15:20:02 thanks 2015-03-11 16:48:51 ping kunkku 2015-03-11 18:34:55 stateless, gcc still does not compile with fortify: 2015-03-11 18:34:56 In file included from /home/tteras/aports/main/gcc/src/gcc-4.9.2/libgfortran/io/close.c:26:0: 2015-03-11 18:34:58 /home/tteras/aports/main/gcc/src/gcc-4.9.2/libgfortran/io/unix.h: In function 'sread': 2015-03-11 18:34:59 /home/tteras/aports/main/gcc/src/gcc-4.9.2/libgfortran/io/unix.h:52:17: error: 'const struct stream_vtable' has no member named '_ 2015-03-11 18:35:02 _fortify_read' 2015-03-11 18:35:04 return s->vptr->read (s, buf, nbyte); 2015-03-11 18:35:40 seems the generic #define approach will not work since the same name can be used as function pointer member in a struct 2015-03-11 19:19:39 hello 2015-03-11 19:19:54 I just got a couple of beaglebone black boards in the mail 2015-03-11 19:20:15 I would be interested in contributing in any way that I can towards porting alpine linux to the board 2015-03-11 19:20:49 zacts, please do. you got the link to mailing list archive pretty much describing where we stand. 2015-03-11 19:21:09 ok 2015-03-11 19:21:34 like i said, it should be matter of adding the proper kernel config, and u-boot script+config 2015-03-11 19:23:55 ok 2015-03-11 19:23:57 cool 2015-03-11 19:42:25 @fabled: would require a new distribution ISO, or AL would expand the rasbpi image to cover all of armhf, etc.? 2015-03-11 19:42:30 s/new/additional/ 2015-03-12 00:27:12 i think the dpkg package should depend on xz as new debs use it and dpkg-deb directly calls xz 2015-03-12 00:28:13 (eg debootstrap will fail without xz) 2015-03-12 09:34:59 algitbot:retry 2015-03-12 09:40:37 fabled: hey 2015-03-12 09:40:47 stateless, hi 2015-03-12 09:40:52 this is a violation on GCC's side as any function may also be provided as a macro 2015-03-12 09:41:00 this approach may be practically not possible though :( 2015-03-12 09:41:04 yes 2015-03-12 09:41:39 I'll have a look at the extern inline thingie 2015-03-12 11:37:21 ping kunkku 2015-03-12 13:06:22 anyone brave enough to look at a package I'm stuck with? 2015-03-12 13:06:36 i'm stuck trying to compile 2015-03-12 15:11:19 ncopa, could you pls backport sysklogd logfile rotation fix to 3.1-stable? 2015-03-12 15:11:29 it's broken there 2015-03-12 15:19:44 rnalrd: did anyone test if the patch actually works? 2015-03-12 15:21:16 well, from email by nangel I assume he has a good test case 2015-03-12 15:23:59 nice 2015-03-12 15:24:05 I'll cherry pick it 2015-03-12 15:24:14 been busy today with this netlink mdev business 2015-03-12 16:05:19 ncopa: any joy with alpine (MTA) 2015-03-12 16:07:11 ScrumpyJack: i think i pushed fixes yesterday? 2015-03-12 16:07:28 commit f293739e0db7bbddff5199f74b6150312c16a6cc 2015-03-12 16:07:28 Author: Natanael Copa 2015-03-12 16:07:28 Date: Wed Mar 11 09:20:01 2015 +0000 2015-03-12 16:07:28 main/alpine: upgrade to 2.20 2015-03-12 16:07:41 i dont use it so i dont know if it actually works... 2015-03-12 16:12:41 i must have missed it on the mailing list 2015-03-12 16:19:57 ncopa: would you mind awefully adding --with-passfile=.pine-pass to the alpine (MTA) configure script? :) 2015-03-12 18:02:36 fabled: I will probably have something for you to test tomorrow or in the weekend 2015-03-12 18:02:51 I am just converting all the headers currently 2015-03-12 19:29:49 stateless, cool 2015-03-12 19:29:51 thanks 2015-03-13 11:02:30 fabled: http://git.2f30.org/fortify-headers/commit/?id=9f8c543dc81f0c4239acae6713f5414eb7dc681d 2015-03-13 11:02:52 there may be typos in it, I went through it twice and did some rudimentary tests 2015-03-13 11:03:03 but it is very easy to make a typo that is not detected at compile time 2015-03-13 11:05:12 stateless, ok, i'll take a test spin later today 2015-03-13 11:05:32 stateless, though, would it make sense to move all the __*_orig declarations inside the corresponding fortify function ? 2015-03-13 11:06:17 perhaps 2015-03-13 11:06:28 it doesn't pollute anything as __ is reserved namespace 2015-03-13 11:06:40 I left it out so people can immediately focus on the important bits, the body of the function where checks actually happen 2015-03-13 11:06:50 instead of seeing this prologue decoration stuff 2015-03-13 11:07:46 ok 2015-03-13 11:07:57 i'll be away for a while. back later today (1-4h) 2015-03-13 11:08:27 ok :) 2015-03-13 11:08:33 let me know if there are any issues 2015-03-13 11:30:08 do you guys have a need for an LDAP client side daemon on Alpine? 2015-03-13 11:30:26 I am working on one but it is not ready yet (I can login over LDAP but that's about it) 2015-03-13 11:30:47 it is using the nscd support code in musl 1.1.7 2015-03-13 12:04:36 stateless: yes, ldap auth is something we are interested in 2015-03-13 13:22:44 ncopa: cool, my project is at http://git.2f30.org/waffle/tree/ 2015-03-13 13:22:56 not ready for public consumption at this point, but once I have something useful and I will send an alpine port for it 2015-03-13 13:25:24 stateless: nice! 2015-03-13 13:25:47 the LDAP backend needs a lot of cleanup 2015-03-13 13:25:58 add TLS support + properly escape queries etc. 2015-03-13 13:26:16 but other than that the basic structure is OK and allows new backends to be added easily 2015-03-13 13:26:41 good job with the fortify-headers btw 2015-03-13 13:26:59 thanks :) there may be small bugs that will mostly be ironed out by building/testing packages 2015-03-13 13:27:11 yeah. thats expected 2015-03-13 13:27:23 I have tested each of the interfaces separately so the overflow checks should be ok but since the conversion I may have introduced typos that go undetected at compile 2015-03-13 13:27:26 at compile time 2015-03-13 13:28:10 we will enable it by default on alpine 2015-03-13 13:28:15 at least now it won't break programs that blindly use reserved names 2015-03-13 13:28:18 nice :) 2015-03-13 13:28:22 so it will get testing 2015-03-13 13:28:30 I also subscribed to the bug tracker 2015-03-13 13:28:42 so you can assign me fortify related problems there or ping me here :) 2015-03-13 13:28:49 good 2015-03-13 13:28:51 will do 2015-03-13 13:28:56 appreciate 2015-03-13 13:29:17 i have been working on the default hotplug handler 2015-03-13 13:29:27 i have a netlink listener now 2015-03-13 13:29:43 when I have some more free time (too many projects atm) I'd like to work on something for alpine, not sure what needs to be done though 2015-03-13 13:29:58 ah 2015-03-13 13:29:59 nice 2015-03-13 13:30:07 have you seen nlmon? 2015-03-13 13:30:11 yeah, those 2, fortify and ldap auth 2015-03-13 13:30:18 http://git.r-36.net/nlmon/ 2015-03-13 13:30:26 are very nice additions 2015-03-13 13:30:34 no, but i have based my work on nldev 2015-03-13 13:30:41 sorry I meant nldev 2015-03-13 13:30:45 yup 2015-03-13 13:30:46 nlmon is just a monitor 2015-03-13 13:30:53 i have forked nldev 2015-03-13 13:30:56 cool :) 2015-03-13 13:31:05 the idea is to do what nldev does 2015-03-13 13:31:11 I am also the author of smdev which I have been using for quite some time but could get some more love/cleanup 2015-03-13 13:31:13 except, instead fork mdev for every event 2015-03-13 13:31:34 http://git.suckless.org/smdev/ 2015-03-13 13:31:37 aha 2015-03-13 13:31:41 will look at that too 2015-03-13 13:31:49 i fork the helper/handler 2015-03-13 13:31:56 and pass over the event via pipe 2015-03-13 13:32:05 handler reads it via stdin 2015-03-13 13:32:08 ah nice 2015-03-13 13:32:16 and have a timeout 2015-03-13 13:32:18 so it can exit 2015-03-13 13:32:26 yup 2015-03-13 13:32:46 the netlink listener catches that, and respawn as needed 2015-03-13 13:32:51 yeah 2015-03-13 13:32:58 that way we get minimal code in memory all times 2015-03-13 13:33:10 but still avoid fork for every event 2015-03-13 13:33:16 yup 2015-03-13 13:33:42 http://git.alpinelinux.org/cgit/ncopa/nldev/tree 2015-03-13 13:33:54 i would like to make it use mdev.conf 2015-03-13 13:34:05 but am currently stuck how to integrate mdev properly 2015-03-13 13:34:13 been discussing it on busybox ml 2015-03-13 13:34:22 for the initial population at system bootup, do you simulate the events by writing into the /sys/devices/.../uevent or do you parse /sys/devices without ever invoking the hotplug handler? 2015-03-13 13:34:50 the idea is to echo add > /sys/...../uevent 2015-03-13 13:35:03 right ok 2015-03-13 13:35:24 smdev has the -s option to parse /sys/devices (this is similar to mdev I think) 2015-03-13 13:35:36 without -s it is just invoked as a handler 2015-03-13 13:36:05 ah 2015-03-13 13:36:07 smdev 2015-03-13 13:36:14 smdev is a simple program to manage device nodes. It is 2015-03-13 13:36:14 mostly compatible with mdev but doesn't have all of its features. 2015-03-13 13:36:19 yup 2015-03-13 13:36:30 I intended to add some more stuff but haven't needed it for my system yet 2015-03-13 13:37:10 sound like i need check it up 2015-03-13 13:37:11 I also added interface renaming as an optional thing :P 2015-03-13 13:37:17 to satisfy all the hipsters 2015-03-13 13:37:22 but the implementation is trivial 2015-03-13 13:39:37 one of the other things that I have been looking at on and off is to try to use sbase on Alpine Linux, I can build quite a few ports but still there's issues that need to be addressed 2015-03-13 13:39:57 we've put a lot of effort to make most/all tools POSIX compliant + any other useful additions that are required for building stuff 2015-03-13 13:40:19 we also have some of the more heavy tools available like a POSIX compliant sed/find etc. 2015-03-13 13:40:36 they are also UTF-8 aware 2015-03-13 13:40:42 for a complete list: http://git.suckless.org/sbase/tree/README 2015-03-13 13:41:52 Alpine serves more of a test-bed for me at the moment but when these tools reach maturity they may be useful to the project as an alternative to busybox 2015-03-13 13:58:58 i am kinda happy with bb 2015-03-13 13:59:03 there is also toybox 2015-03-13 14:01:21 yeah 2015-03-13 14:01:47 the bb code is terrible though :( 2015-03-13 14:02:43 I also have a relatively long list of outstanding conformance issues with busybox but they are mostly not a problem in practice 2015-03-13 14:08:36 stateless: i'd be happy if you could look at nldev too, once fortify-source and ldap auth is done 2015-03-13 14:08:55 cool :) 2015-03-13 14:09:47 smdev might make sense 2015-03-13 14:10:06 i think some of the busybox mdev features are not needed 2015-03-13 14:10:15 we could handle MODALIAS other place 2015-03-13 14:10:22 yeah 2015-03-13 14:10:50 smdev doesn't parse an external .conf file, so that might be something we need to address (the change is straightforward though) 2015-03-13 14:11:04 and the kernel devtmpfs solves some of it 2015-03-13 14:11:18 i'd be ok with require devtmpfs in kernel config 2015-03-13 14:11:47 atm you can use smdev without devtmpfs 2015-03-13 14:11:57 i posted my nldev fork on busybox list: http://lists.busybox.net/pipermail/busybox/2015-March/082528.html 2015-03-13 14:12:13 got some insightsful feedback: http://lists.busybox.net/pipermail/busybox/2015-March/082534.html 2015-03-13 14:15:11 nice :) 2015-03-13 14:15:23 I will have a look at it in detail a bit later 2015-03-13 14:17:06 i have to admit i dont really understand Harald Becker's ideas 2015-03-13 14:17:40 they sounded really bad initially, but i'm not so sure they are as bad as i first thought 2015-03-13 14:17:57 thread started here: http://lists.busybox.net/pipermail/busybox/2015-March/082461.html 2015-03-13 14:18:01 too many words... 2015-03-13 14:24:42 he wants to make inetd style master daemon. similar to what inetd was like. 2015-03-13 14:25:01 but that operates with pipes 2015-03-13 14:34:26 in regards to fortify, I've left the FD_CLR/FD_SET as they were (override as a macro) because POSIX does not specify where these are macros or functions (they can be either or both) 2015-03-13 14:34:44 but that's ok I cannot imagine anyone using these names in any other way other than to refer to these macros/functions 2015-03-13 14:35:20 same for snprintf and sprintf for now but for a different reason 2015-03-13 14:35:45 fabled: what i dont get is, if you have a forever running daemon anyways, why not just listen on netlink and get done with it? 2015-03-13 15:18:54 seems nice http://git.2f30.org/create_ap/ 2015-03-13 15:56:36 ncopa: to test fortify locally, I applied this patch: http://sprunge.us/ChNi 2015-03-13 15:57:00 then built+installed fortify-headers, now building gcc and with the resulting gcc I can start building stuff that will use fortify, right? 2015-03-13 15:58:22 then I can just update the headers directly until things work 2015-03-13 15:59:23 probably. fabled knows 2015-03-13 15:59:38 ok will try this, should work :) 2015-03-13 16:04:13 cool I can reproduce fabled's failures, let's see if gcc builds with the latest fortify :) 2015-03-13 16:36:49 yay I built gcc :D 2015-03-13 16:46:34 binutils/gdb/perl also seem to build ok :) 2015-03-13 17:30:58 ok I've about 168 packages without issues 2015-03-13 17:31:08 I've built* 2015-03-13 17:45:48 hmm.. sounds like new punchline for AL, "stack smashing protection and pax with fortified headers" 2015-03-13 17:46:15 :P 2015-03-13 17:50:35 or just give a single name for such tech, 2015-03-13 17:50:36 joral - Jor-El - borrowed from superman 2015-03-13 17:50:36 alzone - borrowed from superman's, zod Phantom Zone 2015-03-13 17:55:43 packages built so far: http://sprunge.us/aUAV 2015-03-13 17:55:47 will let it run over the weekend :) 2015-03-13 17:59:57 was looking at morpheus, I think with newer kernel having unionfs, that project would be history 2015-03-13 18:01:54 maybe bring some nice base tools it uses to AL 2015-03-13 18:41:25 aconf doesn't work for me: http://hastebin.com/pucacoremo.vbs 2015-03-13 18:41:35 have anyone faced this befor? 2015-03-13 18:46:18 alacerda - apk add augeas-libs ? 2015-03-13 18:46:45 aldev:/etc/aconf# apk info | grep augeas 2015-03-13 18:46:46 augeas-libs 2015-03-13 18:46:46 lua5.2-augeas 2015-03-13 18:46:46 augeas 2015-03-13 18:46:46 lua5.1-augeas 2015-03-13 18:46:46 lua-augeas 2015-03-13 18:46:48 aldev:/etc/aconf# apk add augeas-libs 2015-03-13 18:46:50 OK: 435 MiB in 121 packages 2015-03-13 18:46:54 aldev:/etc/aconf# 2015-03-13 21:58:06 hey fabled 2015-03-13 21:58:53 not sure C++ has this extern inline thing, nsz mentioned it is C99 2015-03-13 21:59:12 I mean maybe GCC has an extension for this in which case it might work 2015-03-13 21:59:18 I will test it tomorrow 2015-03-13 22:06:13 inline in c99 and c++ are different but gcc has the gnu_inline attribute which gives the gnu semantics 2015-03-13 22:06:54 so if you use that attribute then there is no issue 2015-03-13 22:12:55 yeah I use that now 2015-03-13 22:12:58 good :D 2015-03-13 22:15:24 (iirc in c99 extern inline function provides an external definition in c++ and gnu it does not) 2015-03-13 22:16:48 right 2015-03-13 22:16:53 in c99 you would want a non-extern inline i think 2015-03-13 22:17:19 extern __inline __attribute__((__always_inline__,__gnu_inline__)) 2015-03-13 22:17:30 this is what's used atm (__inline for not breaking c89) 2015-03-13 22:17:38 i will read up on it some more 2015-03-13 22:19:14 hm probably non-extern inline is not enough 2015-03-13 22:19:24 so there is no pure c99 solution 2015-03-13 22:19:41 but you need asm hacks anyway 2015-03-13 22:23:17 stateless, yeah, my comment also, __gnu_inline__ is the magic thing 2015-03-13 22:23:24 you might want to add __artificial__ while at it 2015-03-13 22:23:35 adds a bit to debugging info 2015-03-13 22:26:29 yeah 2015-03-14 18:46:05 #3979 2015-03-15 06:14:09 nice, http://rlsd.dimakrasner.com/ , didn't know there were quite a few musl based distros 2015-03-15 06:14:47 good for musl, as it gets more tested 2015-03-15 09:05:44 bump testing/tengine pls (v.2.1.0) 2015-03-15 17:26:32 ncopa: I had to revert the patch for nfs-utils 1.3.2 back to 1.3.1 2015-03-15 17:26:43 for some reason rpc.statd would not start through the initscript (but it does manually) 2015-03-15 17:26:55 it is blocked on a wait4(), haven't tried to debug this yet 2015-03-16 09:18:10 is AL having support for dso ? 2015-03-16 09:18:10 I did not find any explicit entry when I ldd a file 2015-03-16 09:38:34 stateless: ok? so the nfs-utils is broken in 3.1-stable? 2015-03-16 09:52:21 was reading http://lists.busybox.net/pipermail/busybox/2015-March/082492.html 2015-03-16 09:52:30 could not quite get, "The current implementation suffers on event bursts" 2015-03-16 09:52:46 when does "event bursts" happen ? 2015-03-16 09:56:58 ncopa, is mdev.seq currently activated ? 2015-03-16 09:57:07 or I its auto 2015-03-16 09:58:44 vkrishn: i think it is enabled in mdev, but i dont think alpine uses it 2015-03-16 09:58:53 event bursts happens during hotplug 2015-03-16 09:58:57 sorry 2015-03-16 09:58:59 cold plug 2015-03-16 09:59:08 when autodetecting hardware 2015-03-16 10:00:01 but the problem is only after system is up and running and /proc/sys/kernel/hotplug is in effect 2015-03-16 10:00:26 or during boot 2015-03-16 10:01:11 during boot is what i am interested in solving 2015-03-16 10:01:24 mostly 2015-03-16 10:01:38 i do have a pretty clear idea how to solve it, finally 2015-03-16 10:01:50 and i have example code too 2015-03-16 10:02:29 saw the ML thread, yet to digest fully 2015-03-16 10:03:08 its *alot* to digest 2015-03-16 10:03:10 would be nice to see it implemented 2015-03-16 10:03:15 :) 2015-03-16 10:03:28 http://git.alpinelinux.org/cgit/ncopa/nldev/tree 2015-03-16 10:03:40 already cloned 2015-03-16 10:04:00 i think i should rename it to netlink-activator or nlactivator 2015-03-16 10:04:13 or nlsa 2015-03-16 10:04:24 netlink socket activator 2015-03-16 10:04:26 NSA 2015-03-16 10:04:26 nlacct 2015-03-16 10:04:27 :) 2015-03-16 10:04:49 acct sounds like accounting 2015-03-16 10:04:59 nlact 2015-03-16 10:05:28 nlsad 2015-03-16 10:05:34 or nlactd 2015-03-16 10:05:39 nlplug 2015-03-16 10:05:58 nlplugd 2015-03-16 10:06:20 those are good too 2015-03-16 10:06:54 nlplug or nlplugd are nice names 2015-03-16 10:07:15 or kueventd 2015-03-16 10:07:38 or kuevd 2015-03-16 10:08:33 i think that as first step, I will link in libkmod for loading the kernel modules 2015-03-16 10:08:42 in the handler/helper 2015-03-16 10:08:48 I was just toying, if there could be a method of partial tree representation of nl info, in fashion that a post parser can serially identify 2015-03-16 10:09:13 that tree is shortlived 2015-03-16 10:09:23 tree? 2015-03-16 10:09:46 hmm.... like /sys 2015-03-16 10:10:35 "Every gathering part grabs the required information, sanitizes, 2015-03-16 10:10:36 serializes and then write some kind of command to the fifo. The fifo 2015-03-16 10:10:36 management (a minimalist daemon) starts a new parser process when there 2015-03-16 10:10:36 is none running and watches it's operation (failure handling). The 2015-03-16 10:10:36 parser process reads the conf file on startup, then start to read 2015-03-16 10:10:36 commands from the fifo (written by a gathering part)" 2015-03-16 10:11:19 as laurent points out in other mail, there is only one source of the event: the kernel 2015-03-16 10:11:36 so it kinda does not make sense to collect the events from various places 2015-03-16 10:11:43 when nl outputs event, where is it written ? 2015-03-16 10:11:54 and i don't like beckers fifo idea 2015-03-16 10:12:08 yes, that what i wanted to avoid 2015-03-16 10:12:29 you mean the kernel or my hacked nldev? 2015-03-16 10:12:31 but I still need to understand 2015-03-16 10:12:44 kernel 2015-03-16 10:13:02 you need open a datagram socket (netlink) and "subscribe" to the events 2015-03-16 10:13:17 kernel will write to the fileescriptors that are subscribing 2015-03-16 10:13:45 this function shows how to subscribe to uevents from kernel via file descriptor: http://git.alpinelinux.org/cgit/ncopa/nldev/tree/nldev.c#n73 2015-03-16 10:14:10 ncopa: not sure if nfs-utils is broken in stable, I should have mentioned I run edge 2015-03-16 10:14:27 stateless: ok. thanks 2015-03-16 10:14:35 and /sys is populated by ? 2015-03-16 10:14:42 stateless: i'd like to fix it but hands are full... :-/ 2015-03-16 10:14:59 stateless: it would be awesome if you could help me debug it 2015-03-16 10:15:47 I will try to have a look at it 2015-03-16 10:16:04 vkrishn: that is not yet implemented, but the idea is that an external program/script does something equivalent to: find /sys -name uevent | while read entry; do echo "add" > $entry; done 2015-03-16 10:16:36 ok, thanks, would do some more reading 2015-03-16 10:16:41 vkrishn: i think i need add an option for the nlplugd 2015-03-16 10:17:25 ok 2015-03-16 10:17:42 because it needs to run the trigger between init_nelink_socket() and the poll loop 2015-03-16 10:18:34 i also need an option that makes nlplugd exit if handler exits with error 2015-03-16 10:19:25 i would like to call nlplugd from initramfs 2015-03-16 10:19:32 it triggers coldplug 2015-03-16 10:19:37 it runs helper 2015-03-16 10:19:42 which loads modules 2015-03-16 10:19:56 and set up /dev according mdev.conf 2015-03-16 10:20:18 and when there are block devices, it will check if it is root=... 2015-03-16 10:20:24 will it replace mdev or minimize/clean mdev 2015-03-16 10:20:27 if it is root=... then it can ecit 2015-03-16 10:20:38 the idea is that it should complement mdev 2015-03-16 10:20:51 i kinda like the mdev.conf 2015-03-16 10:21:02 but i dont like that mdev forks for every event 2015-03-16 10:21:06 that is what i want avoid 2015-03-16 10:21:22 the busybox ml has some ideas about mdev -i 2015-03-16 10:21:28 which is what i'd like 2015-03-16 10:22:01 so mdev can read events from stdin stream (or neltink socket) instead of via environment ars 2015-03-16 10:22:04 vars* 2015-03-16 10:22:45 why can't the calling program see if mdev is running, then increment a seq number 2015-03-16 10:23:00 mdev sees its >0 and refreshes 2015-03-16 10:23:14 instead of closing after event "a" 2015-03-16 10:23:21 but continues with event "b" 2015-03-16 10:25:50 but an intermediate nl daemon (small size) seems nice 2015-03-16 10:55:07 ncopa i saw few ppl complaing that it does not work ins table, but in edge it does work for them after, (not edge install and then apk update && apk upgrade) as clandmeter pushed an update for nfs-client 2015-03-16 11:18:55 ncopa: regarding nfs, i upgraded it and removed patches which seemed to be applied already. could be some are not exactly like you submiteed upstream. 2015-03-16 11:39:20 ncopa: would you be kind enough to push the alpine (MUA) patch that enables password files? it's in the mailing list :) 2015-03-16 11:57:24 with nlplugd, this would not needed, echo /sbin/mdev >/proc/sys/kernel/hotplug 2015-03-16 12:03:46 fabled: http://dl.2f30.org/releases/fortify-headers-0.3.tar.gz 2015-03-16 12:07:37 stateless, thanks. i'll try to get the toolchain commits pushed today. or tomorrow by latest. 2015-03-16 12:07:43 need to take care of few other things first 2015-03-16 12:08:51 when doing coldplug, i wonder which causes most forks: /dev/* entries or modprobe 2015-03-16 12:09:10 i think i relatively easy can reduce modprobes 2015-03-16 12:09:34 forks caused by modprobe i mean 2015-03-16 12:18:40 i'm looking at mdev code 2015-03-16 12:18:55 does not look too hard to implement the netlink message parsing in mdev 2015-03-16 12:19:05 (or read from pipe) 2015-03-16 12:19:42 if easy, can it segregate/distinguise and serialize burst event ? 2015-03-16 12:21:13 thats what it'd would do basically 2015-03-16 12:21:28 i would read resialized events from file descriptor 2015-03-16 12:21:35 which means, with every event it maybe possible to orchestrate mdev to act upon that event only 2015-03-16 12:21:52 and not redo full conf 2015-03-16 12:21:53 yes 2015-03-16 12:22:24 that would be nice 2015-03-16 12:23:04 but may require mdev code changes 2015-03-16 12:23:12 yes 2015-03-16 12:23:17 that is what i am looking at 2015-03-16 12:23:21 what mdev code changes 2015-03-16 12:25:59 fabled: awesome, thanks :) 2015-03-16 12:28:24 act to that particular event 2015-03-16 12:29:13 hmm... or it already does that 2015-03-16 12:32:48 example if there is an event, 2015-03-16 12:32:49 * Case with ACTION=change: 2015-03-16 12:32:49 * ACTION=change 2015-03-16 12:32:49 * DEVNAME=sdb 2015-03-16 12:32:49 * DEVPATH=/block/sdb 2015-03-16 12:32:49 * DEVTYPE=disk 2015-03-16 12:32:51 * DISK_MEDIA_CHANGE=1 2015-03-16 12:32:53 * MAJOR=8 2015-03-16 12:32:55 * MINOR=16 2015-03-16 12:32:57 * SUBSYSTEM=block 2015-03-16 12:38:03 Hi, trying to run docker in alpinelinux 3.2.1 as host. Can anyone share their experience, if any? Thanks! 2015-03-16 12:40:47 wait_for_seqfile(const char *seq) (ln 946, busybox-1.23.1), stops new instance of mdev, right ? 2015-03-16 12:44:59 i think it will just queue new instances of mdev 2015-03-16 12:45:11 wait til previous instance is done 2015-03-16 12:46:22 btw, i think we can disable MDEV_LOAD_FIRMWARE config 2015-03-16 12:46:27 kernel should handle that himself now 2015-03-16 12:48:02 zenny: is something broken? apk add docker should do ;) 2015-03-16 12:49:39 what I was suggesting was different, the current(old,earlier) running mdev, checks a seq num, if its >0, keep running and refresh/actUpon with newer events 2015-03-16 12:49:47 hope that's not crazy 2015-03-16 12:50:38 its crazy 2015-03-16 12:50:46 :) 2015-03-16 12:50:49 better serialize them by reading from netlink 2015-03-16 12:51:01 yes, thats next part 2015-03-16 12:51:22 nlplugd, when gets a new event, it orchestrates mdev 2015-03-16 12:51:34 zenny: there is no much add, start the service and you're good to go 2015-03-16 12:51:43 or any other application 2015-03-16 12:51:49 zenny: if you're encountering any issues, drop me or uggedal a message 2015-03-16 12:52:40 which means nlplugd should be able to feed mdev a info similar to /sys (single event) 2015-03-16 12:53:02 thus, the toying idea of shortlived tree(serialized) 2015-03-16 12:55:23 maybe I am missing something, like what happens when new instace(fork) of mdev starts, does it re-scan for changes ? 2015-03-16 12:56:34 barthalion: yep, I am confused because in ubuntu or debian there is /etc/default/docker but I couldn't find one in alipine. Nor at /etc/ 2015-03-16 12:59:44 zenny: what about /etc/conf.d/docker? 2015-03-16 13:04:21 barthalion: thanks, that's what I have been searching. 2015-03-16 13:12:28 shortlived tree(serialized), also means no getenv() 2015-03-16 13:16:14 if wait_for_seqfile queue new instances of mdev , ie wait til previous instance is done , is this not expensive, on resources 2015-03-16 13:27:13 I'm going to package redis-lua, i've a couple of doubt. 2015-03-16 13:27:35 Since it's only a .lua file, should we create two subpackages,one for 5.1, one for 5.2? 2015-03-16 13:27:57 Or one is enough, since it works for both? 2015-03-16 13:28:11 is it ok call it lua-redis? 2015-03-16 13:28:44 lua looks for lua$ver directory 2015-03-16 13:29:08 and redis-lua needs lua-socket to work 2015-03-16 13:29:26 so i suppose that two subpackages is a better choice 2015-03-16 13:33:06 fcolista: just drop lua5.1 2015-03-16 13:33:13 unles you actually need it 2015-03-16 13:33:22 no i don't 2015-03-16 13:33:49 thx 2015-03-16 13:35:09 ncopa, what about the name? 2015-03-16 13:35:16 is it fine lua-redis? 2015-03-16 13:35:25 or i should specify lua5.2-redis? 2015-03-16 13:35:36 original package name is redis-lua 2015-03-16 13:36:55 fcolista: no strong opinion from here 2015-03-16 13:37:09 but normally apk add lua- should pull in all supported lua versions 2015-03-16 13:37:40 if you want a specific lua version then lua- can be used 2015-03-16 13:38:27 this package works for both lua5.1 and lua5.2 2015-03-16 13:39:14 i'd name the aport lua-redis and have 2 subpackages lua5.1-redis and lua5.2-redis 2015-03-16 13:39:38 that was my 1st option 2015-03-16 13:39:39 thats what i normally do 2015-03-16 13:40:12 I saw your lua-mosquitto APKBUILD 2015-03-16 13:40:42 kinda hackish :) 2015-03-16 13:55:26 nice! fortify is merged 2015-03-16 13:55:43 does that mean that when all the packages are rebuilt I will have 320 package upgrades? :P 2015-03-16 13:56:49 is there some documentation and tests on this fortify? i'm very interested in details. 2015-03-16 13:57:35 rpu3uO8PEVZZ: I have a README in the repository 2015-03-16 13:57:47 http://git.2f30.org/fortify-headers/tree/README 2015-03-16 13:57:58 I've managed to build about 350 ports or so 2015-03-16 13:58:26 welcome, to alzone, joral greets you! 2015-03-16 13:58:29 there could still be runtime problems, so if you see any program crashing with Illegal Instruction it is probably due to fortify (this needs to be documented somewhere) 2015-03-16 13:58:55 and tests regarding the effectiveness of the mitigations implemented? 2015-03-16 13:59:38 well you should be able to find enough information on that on google, but bear in mind that this fortify implementation will not break valid programs 2015-03-16 14:00:19 unlike with glibc and -D_FORTIFY_SOURCE=2 (as opposed to level 1) it will potentially trap conformant programs 2015-03-16 14:00:46 performance overhead should be non-measurable too 2015-03-16 14:02:03 rpu3uO8PEVZZ: and another advantage is that this implementation will also try on out-of-bounds read interfaces not just for out-of-bounds write interfaces 2015-03-16 14:02:20 (more work needs to be done in this area, specifically should we also try on out of bounds reads in memcpy() etc?) 2015-03-16 14:02:23 s/try/trap/ 2015-03-16 14:04:38 stateless: no, we will not rebuild all packages for edge (for now) 2015-03-16 14:04:43 all updates will get it 2015-03-16 14:04:45 ok :) 2015-03-16 14:04:56 so hopefully this will be in 3.2? 2015-03-16 14:05:03 yes 2015-03-16 14:05:06 awesome :) 2015-03-16 14:05:14 yes :) 2015-03-16 14:05:34 now I can work on some cleanup + support for more checks 2015-03-16 14:06:57 I mean the general interfaces covered are more or less complete, most/all functions that I am aware of are covered 2015-03-16 14:07:22 but maybe adding checks for overflowing the source pointer in calls like memcpy() is desirable 2015-03-16 14:08:06 s/overflowing/reading past/ 2015-03-16 14:12:32 ncopa: ok will use alpine-ports, wasn't aware of that :) 2015-03-16 14:13:10 np 2015-03-16 14:13:14 oh it was you :) 2015-03-16 14:13:18 yeah :) 2015-03-16 14:13:25 well my original nickname is taken on freenode :( 2015-03-16 14:13:27 stateless, very cool. can you point at some exploit that i can test this with? like works without but not with fortify? 2015-03-16 14:14:41 rpu3uO8PEVZZ: not from the top of my head but I remember various bugs in the past that made exploitation more difficult with fortify 2015-03-16 14:15:10 thing is glibc does potentially do unsafe things when it detects an overflow, in our case the program aborts without making any function calls or system calls which is marginally better (it would matter more for SSP) 2015-03-16 14:15:24 well, i'll play with the example in the readme then. 2015-03-16 14:15:40 also format string bugs are not fortified because it cannot be done properly without compiler assistance (it can break valid usage) 2015-03-16 14:16:23 also it doesn't currently work with clang at all 2015-03-16 14:16:42 some people a bit of an investigation and it seems that there may be a bug in clang 2015-03-16 14:16:57 did a bit* 2015-03-16 14:17:48 although clang on alpine will not try to use fortify so that shouldn't affect existing users 2015-03-16 14:18:16 i'm sorry for being ignorant about this, can you tell me how this relates to -fstack-protector or grsec for example? 2015-03-16 14:18:46 fortify doesn't only work for stack lived objects 2015-03-16 14:19:26 ah, i see 2015-03-16 14:20:16 as long as the compiler can determine the size of the object involved at compile time 2015-03-16 14:20:35 you can perform size checks against it in all the functions that manipulate the object with potentially disastrous results 2015-03-16 14:22:06 hmm, i see, that's something that pax does in the kernel already iirc 2015-03-16 14:22:18 nice to have it in userland. thx! 2015-03-16 14:23:08 http://www.informit.com/articles/article.aspx?p=2036582&seqNum=6 2015-03-16 14:23:11 there is some info here 2015-03-16 14:26:20 hmm, just reading the 2004 redhat announcement, they had back then 160KB tests, maybe they can be reused for tests? 2015-03-16 14:28:57 thanks for these links! very informative 2015-03-16 14:30:26 is it possible to scan somehow automatically for packages that do not work with fortify_source=2? 2015-03-16 14:31:02 no not easily 2015-03-16 14:31:16 our implementation will handle level 2 in the same way as level 1 2015-03-16 14:31:30 but it won't provide format string checks regardless (amongst other things) 2015-03-16 14:39:50 isn't formatstrings mostly handled by: https://fedoraproject.org/wiki/Format-Security-FAQ 2015-03-16 14:41:03 so there's no difference between level 1&2? 2015-03-16 14:44:00 ncopa, with shortlived tree of events, there would not be need of 2015-03-16 14:44:00 } else { <--block line 1071 in mdev.c(busybox) 2015-03-16 14:44:00 it will just read the temp tree, maybe -p (partials, /sys/partials) 2015-03-16 14:44:00 if (argv[1] && strcmp(argv[1], "-p") == 0) { 2015-03-16 14:44:00 which is populated by nlplugd, also likely no need for nldev-handler.c 2015-03-16 14:44:01 but still crazy... 2015-03-16 14:46:08 how efficiently partials are handled, that is a different issue 2015-03-16 14:49:51 which means, moving nlplugd to busybox 2015-03-16 15:00:08 rpu3uO8PEVZZ: yeah those are just warnings, if you want the program to terminate on such problems then it is not going to happen with the current fortify implementation 2015-03-16 15:00:40 the glibc does it is it assumes that if you have a %n in your format string and the location of the format string is in writeable memory, then the program will crash at runtime regardless 2015-03-16 15:01:23 so char fmt[] = "foo%n\n"; printf(fmt, &n); will crash but printf("foo%n\n", &n); won't 2015-03-16 15:01:53 either way it is valid from a conformance point of view (potentially unsafe if fmt is user controllable obviously) 2015-03-16 15:03:01 that's why level 2 fortify exists, so you can only enable it when you willingly want to break valid programs, it doesn't make sense to be the default option in a distro 2015-03-16 15:06:22 i see 2015-03-16 15:07:15 in our case, level 2 is the same as level 1, you do not even have the option to break valid programs :D 2015-03-16 15:07:33 and if valid usage is broken as a result of fortify, it is a bug in fortify 2015-03-16 15:09:36 hmm, i found only exploits circumventing the fortify format string protection 2015-03-16 15:09:46 yeah 2015-03-16 15:09:55 http://phrack.org/issues/67/9.html 2015-03-16 15:10:00 yeah I've read the paper 2015-03-16 15:10:14 for the others more than you :) 2015-03-16 15:11:38 I do not think there's much you can do to redirect execution after you've hit the check but you can potentially redirect execution before you hit the check function 2015-03-16 15:11:51 depends on the situation 2015-03-16 15:12:26 the only thing that happens once the overflow has been detected is to __builtin_trap() 2015-03-16 15:23:12 it builds fine on my dev machine :( but it needs the perl-file-readbackwards package... 2015-03-16 15:27:15 sorry, the perl-file-readbackwards is needed by the x2goserver package 2015-03-16 15:32:29 Jean-Scotch: it also failed on 32bit, x86 2015-03-16 15:33:27 shouldn't the build logs include more information? 2015-03-16 15:34:43 stateless: they do, but the irc bot does not give the proper url 2015-03-16 15:34:47 http://bld1.alpinelinux.org/buildlogs/build-edge-x86/testing/nx-libs/nx-libs-3.5.0.30-r0.log 2015-03-16 15:34:49 ah 2015-03-16 15:34:53 Strange as I did a update && upgrade and a fresh install of the git tree juste before building it on my dev machine. 2015-03-16 15:34:55 thanks 2015-03-16 15:35:39 the per package logs are in http://bld1.alpinelinux.org/buildlogs/build-edge-x86/ 2015-03-16 15:35:45 at least for x86 2015-03-16 15:35:46 building packages on a xeon e5-2650 with 16 threads and 32G of ram is quite nice :D 2015-03-16 15:37:09 cpuinfo: http://sprunge.us/AXYW 2015-03-16 15:37:29 24 cores is kinda nice too :) 2015-03-16 15:37:40 ah crap :P 2015-03-16 15:37:54 SSD too I take it? :P 2015-03-16 15:38:03 unfortunally no 2015-03-16 15:38:23 reminds me i got ssd disks for my wandbord arm build boxes 2015-03-16 15:38:33 i need to set up those :-/ 2015-03-16 15:38:37 what box is that? 2015-03-16 15:38:41 mine is a "regular" workstation at work 2015-03-16 15:42:23 i got 2 wandboards: http://www.wandboard.org/ 2015-03-16 15:42:36 the quad variant 2015-03-16 15:43:35 no I meant the 24 core box :P 2015-03-16 15:44:43 its the build server(s) 2015-03-16 15:44:47 i have 2 of them 2015-03-16 15:44:53 hp server something i think 2015-03-16 15:45:12 nice :) 2015-03-16 15:54:51 ScrumpyJack, luufs is a COW fs ? 2015-03-16 16:03:08 I just reran the build on my dev machine and it ran fine :-/ 2015-03-16 16:18:02 according to the dev at x2go, the probable cause of the failure is the amount of core user by buildozer (I only use 4 on my dev machine). He says that's probably a race condition into the build process 2015-03-16 16:21:26 and that's a bug in nx-libs but not one ealily fixed... 2015-03-16 16:28:17 may I resubmit with "make -j 4" as it's working with 4 core on my box ? 2015-03-16 16:47:24 Jean-Scotch: resubmit with -j1, please 2015-03-16 16:48:22 ha! ok 2015-03-16 16:49:39 there is still a possibility that it'll fail with 4 threads 2015-03-16 16:57:04 ok running a completely fortified system with X seems to work, now building firefox :) 2015-03-16 17:07:10 nice 2015-03-16 17:07:21 vkrishn: nope 2015-03-16 17:11:00 it's a union filesystem, so it overlays the underlying fs. 2015-03-16 17:15:39 I am unable to find some docs on how to use it in real setup ? 2015-03-16 17:16:04 what does "overlays" mean ? 2015-03-16 17:18:51 does luufs-doc not work? 2015-03-16 17:19:34 luufs /bottom /upper /mountpoint 2015-03-16 17:20:33 /mountpoint will contain both the /bottom and /upper directories (unified) 2015-03-16 17:20:53 but /bottom will be read only and /upper will be read-write 2015-03-16 17:22:36 if you use in any of your opensource project, pls let me know 2015-03-16 17:22:46 how you made use of it 2015-03-16 17:23:54 since for now, thats the only unionfs we have in AL 2015-03-16 17:24:10 for exemple, /a with all the important things like busybox can be mounted on /bin in read only and /b can be mounted in /bin too, so you can add things to /bin, but you can't modify /bin/busybox 2015-03-16 17:24:21 ok main/unionfs-fuse is there 2015-03-16 17:24:51 yes, aufs and unionfs are two other such tools 2015-03-16 17:26:08 that means one cannot run `apk update` or `apk upgrade` ? 2015-03-16 17:27:16 it means you can't modify what you mount from your /bottom fs . i the /bin exemple, i guess it would break lots of things, including apk 2015-03-16 17:28:14 that's the reason I was looking for some docs or projects that uses it effectively 2015-03-16 17:31:16 you could use it to serve webpages that you dont want modified. maybe some config files. put an include statement in your main_config file and put the configs you don't want to be changed even by root in /a/main_config then mount it somewhere where you can have the includes too 2015-03-16 17:32:41 ok, would think over it 2015-03-16 17:32:57 i guess you could use it any time you want "stuff not to touch" to live along side of "stuff you can touch" :) 2015-03-16 17:33:12 :) 2015-03-16 17:37:55 I was kinda pleased when I read unionfs, till I read its features 2015-03-16 17:38:20 coz i was expecting cow 2015-03-16 17:39:36 seems http://rlsd.dimakrasner.com/ uses it 2015-03-16 17:39:47 he wrote it 2015-03-16 17:40:09 he also wrote logfence which i like 2015-03-16 17:40:24 (it's in my list of packages to make) 2015-03-16 17:40:38 yes, I saw that seems nice, theoritically, need to use and see 2015-03-16 17:41:39 also downloaded the iso, neat, runs under <48mb 2015-03-16 17:41:40 agreed. got for fencing syslog 2015-03-16 17:41:45 s/got/good 2015-03-16 17:44:05 i like some of the security decisions that he's made with rlsd 2015-03-16 18:50:09 how long does it take to get a new python module into testing? 2015-03-16 18:51:21 mosez: have you sent a patch? 2015-03-16 18:51:23 mosez, I assume you submitted the aports/patch 2015-03-16 18:51:28 depends how fast you get someone to merge it 2015-03-16 18:52:02 (or write the APKBUILD, actually) 2015-03-16 18:52:09 less than 6 hours, I guess 2015-03-16 18:52:34 ScrumpyJack: what's the point of luufs when there is overlayfs? 2015-03-16 18:52:34 not yet sent. but i would do it :) 2015-03-16 18:52:52 barthalion: more choice 2015-03-16 18:52:56 just wanted to ask if somebody can tell me how long it takes to get it in in a normal case :) 2015-03-16 18:53:27 so i will setup en dev environment and create packages for py-numpy and elasticsearch 8) 2015-03-16 18:53:30 mosez: the greatest chances are in 8-16 CET 2015-03-16 18:53:41 ScrumpyJack: so actually no point :p 2015-03-16 18:53:46 excellent... i'm located in germany :D 2015-03-16 18:54:02 and if you have APKBUILD already, I can take a look 2015-03-16 18:54:19 i will do it now... have just read the guide :) 2015-03-16 18:54:37 i'm more familar with suse packaging until now :P 2015-03-16 18:55:09 APKBUILDs are way more simple 2015-03-16 18:55:21 so it shouldn't be a challenge, if you know spec files 2015-03-16 18:55:23 does it make sense to create a package for http://overviewer.org/ as well? it's a python application related to minecraft. 2015-03-16 18:55:44 yeah, looks quite easy 2015-03-16 18:55:58 and i can see the gentoo roots within the init files as well 2015-03-16 18:56:00 sure, if you feel it deserves packaging 2015-03-16 18:56:08 yeah, we use openrc 2015-03-16 18:56:18 barthalion: how are you getting on with the docker build? 2015-03-16 18:56:27 I've not been following that 2015-03-16 18:56:44 i'm a suse engineer... but i really love the container size of alpine for docker :) 2015-03-16 18:59:18 12mb compared with 600mb is a small difference 2015-03-16 19:12:41 12mb? they are doing something wrong :) 2015-03-16 19:13:02 my alpine base images are around 3mb 2015-03-16 19:17:55 kaniini: ask andyshinn :D 2015-03-16 19:18:11 https://github.com/gliderlabs/docker-alpine this is the base image i'm using 2015-03-16 19:18:28 ah, sorry... 12mb is the size of my memcached container :D 2015-03-16 19:19:31 6mb is my base image including bash 8) 2015-03-16 19:19:55 and ncurses 2015-03-16 19:20:03 https://github.com/dockhippie/alpine 2015-03-16 19:20:19 did you know you can throw away suse 2015-03-16 19:20:24 and run docker on alpine itself 2015-03-16 19:20:37 imagine the potential 2015-03-16 19:20:53 kaniini: i know... but eat your own dogfood, you know? ;) 2015-03-16 19:21:05 i would not eat that dog food 2015-03-16 19:21:17 (or any dog food for that matter, i am not a dog) 2015-03-16 19:21:21 i'm fine with that 2015-03-16 19:21:54 opensuse is the first distribution where i think ruby gem packaging is more or less done right 2015-03-16 19:22:08 out of the perspective of a ruby developer 2015-03-16 19:22:57 mosez: I am getting the following when I try to pull your repos: "FATA[0005] Error pulling image (latest) from webhippie/alpine, ApplyLayer exit status 1 chmod /bin/bbsuid: permission denied" 2015-03-16 19:23:17 zenny: wtf? never seen that before 2015-03-16 19:24:16 zenny: http://bugs.alpinelinux.org/issues/3934 maybe related to this? 2015-03-16 19:24:36 directly the first hit for google search :D 2015-03-16 19:25:17 zenny: as you can see in the dockerfile i'm not doing something special 2015-03-16 19:26:31 as i see the armhf message, i have to try alpine on my rpi 2015-03-16 19:28:22 does "gem install bundler" work on a normal install of alpine edge? 2015-03-16 19:28:23 mosez: changing the parameters in sysctl.conf did let pull the image, but got "FATA[0022] Error response from daemon: Cannot start container 7e42160de19ad9647960aaff8bbd6a0504fd3093564bde67d7f2b8e0001cb5c3: setup mount namespace mount system mounting mqueue into /var/lib/docker/vfs/dir/7e42160de19ad9647960aaff8bbd6a0504fd3093564bde67d7f2b8e0001cb5c3/dev/mqueue no such device" error. 2015-03-16 19:28:55 i'm getting "Unable to download data from https://rubygems.org/ - SSL_connect returned=1 errno=0 state=SSLv3 read server certificate B: certificate verify failed (https://api.rubygems.org/latest_specs.4.8.gz)" within my docker container 2015-03-16 19:28:55 mosez: the container could not be started. 2015-03-16 19:29:26 zenny: maybe i'm the wrong person for getting it running on alpine. i never tried to run docker images on an alpine host 2015-03-16 19:30:22 kaniini: do you know what's going wrong? 2015-03-16 19:32:31 ncopa, is it possible to apply this: http://lists.alpinelinux.org/alpine-aports/0277.html ? 2015-03-16 19:32:34 mosez: thanks. Btw, did you checked https://github.com/rancherio/os a 20MB host for docker, impressive like alpine docker image 2015-03-16 19:33:27 zenny: i know that, but i'm just watching it to see how it works out :) 2015-03-16 19:35:47 thanks ncopa 2015-03-16 19:38:48 hum... it really looks like certs are not overall handled nice on alpine, is that possible? :) 2015-03-16 19:38:54 mosez: like you stated, it does not seem to be the problem with your github image repo, but with alpine host for docker. I can pull your image with rancheros host without any errors. 2015-03-16 19:39:07 got problems in the past with certs in combination with java as well 2015-03-16 19:39:28 zenny: good to know :) 2015-03-16 19:40:40 i think i will come back to that stuff tomorrow morning 2015-03-16 19:51:53 I appended sysctl.conf with two lines as stated here: http://bugs.alpinelinux.org/issues/3934 and could pull the image, but the container cannot start with errror "FATA[0022] Error response from daemon: Cannot start container 7e42160de19ad9647960aaff8bbd6a0504fd3093564bde67d7f2b8e0001cb5c3: setup mount namespace mount system mounting mqueue into /var/lib/docker/vfs/dir/7e42160de19ad9647960aaff8bbd6a0504fd3093564bde67d7f2b8e0001 2015-03-16 20:04:45 Posted a bug report at http://bugs.alpinelinux.org/issues/4010 2015-03-16 20:06:42 is there a way to blacklist a package upgrade? 2015-03-16 20:07:30 http://wiki.alpinelinux.org/wiki/Alpine_Linux_package_management#Holding_a_specific_package_back 2015-03-16 20:07:34 cheers 2015-03-16 20:26:47 ScrumpyJack: I'm nowhere atm, too busy at work and too tired at home to care 2015-03-16 20:39:53 ncopa: I forgot to increment the pkgrel in nx-libs :( 2015-03-16 20:40:27 is it possible it is the reason of the new failure? 2015-03-16 22:27:13 i can not get gem running within my alpine container :( 2015-03-16 22:50:06 can i overwrite the values of /etc/abuild.conf with env variables? 2015-03-17 07:13:22 morning 2015-03-17 07:21:50 morning 2015-03-17 08:20:50 algitbot: pls log alpine-linux 2015-03-17 08:29:02 ncopa, would any of the patches http://ftp.de.debian.org/debian/pool/main/x/xchat/xchat_2.8.8-7.1.debian.tar.gz , be useful 2015-03-17 08:29:36 dunno 2015-03-17 08:29:44 i doubt 2015-03-17 08:29:57 i use hexchat instead of xchat nowdays 2015-03-17 08:30:05 i think xchat is dead upstream 2015-03-17 08:31:06 vkrishn: do you have a specific issue you need fix in xchat? 2015-03-17 08:31:09 ok would download and try to move to it 2015-03-17 08:31:55 no, just the weird dcc discussion on alpine-linux, that made me look into the code 2015-03-17 08:32:15 hum, we shuold probably try upgrade xorg-server 2015-03-17 08:32:25 and found the patches, and also saw its been not updated since long time 2015-03-17 08:32:41 its dead upstream afaik 2015-03-17 08:32:53 ok downloading hexchat 2015-03-17 08:33:06 there is a migration doc on their website 2015-03-17 08:33:22 so you can move over your current config and logs and all 2015-03-17 08:33:32 ok 2015-03-17 10:19:24 morning :) 2015-03-17 11:01:14 here comes xorg 1.17 2015-03-17 11:07:39 wee. 2015-03-17 11:08:04 ncopa: few drivers are included in xorg now, right? 2015-03-17 11:15:48 okay nvm, I just saw line with cirrus, geode and v4l. 2015-03-17 11:44:07 look at this: http://hastebin.com/averirujar.vbs 2015-03-17 11:45:13 I did: ln -s /usr/lib/libfreetype.so.6.11.3 /usr/lib/libfreetype.so.6 2015-03-17 11:45:19 apk fix 2015-03-17 11:45:19 (1/3) Reinstalling mkfontscale (1.0.7-r2) 2015-03-17 11:45:19 (2/3) Reinstalling mkfontdir (1.0.5-r1) 2015-03-17 11:45:20 (3/3) Reinstalling fontconfig (2.11.1-r1) 2015-03-17 11:45:20 Executing busybox-1.22.1-r14.trigger 2015-03-17 11:45:20 Executing mkfontscale-1.0.7-r2.trigger 2015-03-17 11:45:22 Executing mkfontdir-1.0.5-r1.trigger 2015-03-17 11:45:24 Executing fontconfig-2.11.1-r1.trigger 2015-03-17 11:45:26 OK: 409 MiB in 220 packages 2015-03-17 11:53:36 I've been struggling a bit with an aport. anyone about? 2015-03-17 11:54:32 i've patched several times, and i'm stuck on the next error 2015-03-17 11:54:41 (which compiling) 2015-03-17 11:56:24 s/which/when 2015-03-17 12:02:10 ncopa: so I take it this is a fortified xorg? :D 2015-03-17 12:02:20 given that gcc was updated too 2015-03-17 12:02:25 scadu: i only saw xf86-video-modesetting driver shipped with xorg 2015-03-17 12:02:46 stateless: yes, xorg should been built with fortify 2015-03-17 12:02:51 :) 2015-03-17 12:03:33 those should be built with fortify: http://sprunge.us/LNJQ 2015-03-17 12:05:47 yeah cool 2015-03-17 13:04:40 will AL v3.2 have lua v5.1 ? 2015-03-17 13:13:08 ah thought i was joined here as well :) 2015-03-17 13:55:24 Mar 17 12:27:16 localhost kern.alert kernel: [ 3664.923100] grsec: From 192.168.64.22: denied resource overstep by requesting 4096 for RLIMIT_CORE against limit 0 for /usr/bin/marco[marco:3198] uid/euid:1000/1000 gid/egid:1000/1000, parent /usr/bin/mate-session[mate-session: 2015-03-17 13:56:03 do I need to do anything in order to grsec leave my marco start properly ? 2015-03-17 14:42:31  HexChat: 2.10.0 ** OS: Linux 3.9.6-64 x86_64 ** Distro: Debian 7.1 ** CPU: 2 x Intel(R) Core(TM)2 Duo CPU E7500 @ 2.93GHz (GenuineIntel) @ 1.60GHz ** RAM: Physical: 5.8GiB, 72.3% free ** Disk: Total: 110.8GiB, 28.5% free ** VGA: Intel Corporation 4 Series Chipset Integrated Graphics Controller ** Sound: HDA-Intel - HDA Intel ** Ethernet: Atheros Communications Inc. AR8121/AR8113/AR8114 Gigabit or Fast Ethernet ** 2015-03-17 14:42:31 Uptime: 1w 0d 21h 10m 45s ** 2015-03-17 14:43:24 oops. sorry 2015-03-17 17:28:29 i have created my own docker container for building packages :D https://registry.hub.docker.com/u/webhippie/apkbuild/ 2015-03-17 17:57:29 mosez: check out https://github.com/andyshinn/docker-alpine-abuild for additional ideas 2015-03-17 18:19:16 mosez in your case the bash is package someone would like to build? "docker run -ti --rm -v $(pwd):/source webhippie/apkbuild:latest bash" or how can you actually build an package with your docker, maybe i missed the whole docker thing. 2015-03-17 18:27:18 crow: no, bash is the command to run in the container, it is just an interactive bash shell with a pre-configured environment with stuff to run abuild 2015-03-17 18:29:44 so how would i compile an APKGBUILD in this case? 2015-03-17 18:34:39 in that case, he is assuming your PWD is the current folder with your APKGBUILD file and patches, etc, and it gets passed in to the container with -v $(pwd):/source and now /source inside the container is where you would run abuild 2015-03-17 18:37:37 I love how everyone mistakes APKBUILD with APKGBUILD sooner or later 2015-03-17 18:38:18 andyshinn: i have the aports clone at pwd and i change into the testing folder for building the package. i will take a look at your container :) 2015-03-17 18:39:04 hum... which url should i define for source if the tarball comes from github? 2015-03-17 18:39:31 e.g. https://github.com/overviewer/Minecraft-Overviewer/tarball/v0.12.0 2015-03-17 18:40:05 saveas-https://github.com/overviewer/Minecraft-Overviewer/tarball/v0.12.0/overviewer-0.12.0.tgz? 2015-03-17 18:40:06 https://github.com/overviewer/Minecraft-Overviewer/archive/v0.12.0.tar.gz 2015-03-17 18:40:14 this is what you want 2015-03-17 18:40:19 barthalion: thx 2015-03-17 19:14:31 barthalion well yes sorry, i use both system :) 2015-03-17 20:23:59 is xorg supposed to work with linux-grsec ? 2015-03-17 20:24:08 can't seem to run X 2015-03-17 20:31:35 xorg runs fine on alpine 2015-03-17 20:33:23 rpu3uO8PEVZZ: it doesn't here 2015-03-17 20:33:25 :D 2015-03-17 20:34:09 what's the problem? 2015-03-17 21:08:46 So, I see python-2.7.9 in stable; is there a python3 migration plan, or any discussion on the matter? 2015-03-17 21:09:31 It seems some of the distributions are being relatively aggressive about moving to 3, which is surprising but welcome. 2015-03-17 21:37:24 mjones_: http://lists.alpinelinux.org/alpine-devel/4663.html 2015-03-17 21:41:42 scadu: thanks for the pointer! 2015-03-17 21:42:13 scadu: argh, the worst part is I now remember reading this post. 2015-03-17 21:44:42 mjones_: my memory isn't better :> 2015-03-17 21:46:17 mjones_: tldr we have python3 in testing, making it "stable" is matter of resources 2015-03-17 21:46:28 might happen in 3.3 unless ncopa clones himself 2015-03-18 07:16:46 I think this https://registry.hub.docker.com/_/alpine/ resolves the issue for us, congrats andyshinn 2015-03-18 07:58:08 barthalion: nice! 2015-03-18 07:58:51 Ha! 2015-03-18 07:59:02 mysql client alpine docker image is 16MB 2015-03-18 07:59:24 compared to ubuntu 14.04 docker image is 232MB 2015-03-18 08:39:30 ncopa: do you remember why we wrote update-ca-certificates in lua? 2015-03-18 08:43:25 looks like that was done by fabled 2015-03-18 08:43:38 there is a ref to #2846 in git commit 2015-03-18 08:44:19 performance too 2015-03-18 08:44:34 http://bugs.alpinelinux.org/issues/2846#note-4 2015-03-18 08:44:47 on his arm box it went from 3-20 seconds to 0.25 secs 2015-03-18 08:45:04 barthalion: ^^^ 2015-03-18 08:45:19 ok, got it 2015-03-18 08:45:41 I was thinking about rewriting it to shell, but I correctly thought there is a reason behind using lua 2015-03-18 08:46:27 if you want get rid of lua dependency, then I'd recommend do it in C 2015-03-18 08:48:41 should probably not be more than 200 lines C code 2015-03-18 08:49:17 yeah 2015-03-18 08:49:23 I'll take a look probably 2015-03-18 08:50:23 please ask fabled to do peer review of it 2015-03-18 08:59:15 barthalion, yes, it was performance issue only 2015-03-18 08:59:35 and lua sounded better than perl or python 2015-03-18 08:59:48 C would've worked, but it was faster to do it in lua 2015-03-18 09:00:29 sure thing 2015-03-18 09:43:14 clandmeter: are you working on bugs.a.o now? 2015-03-18 09:43:21 it does not pick up git commits 2015-03-18 09:43:26 nope 2015-03-18 09:43:28 didnt touch it 2015-03-18 09:43:45 seems like it is not syncing git repo 2015-03-18 09:46:40 ok, something happend with mqtt-exec 2015-03-18 09:46:41 hum 2015-03-18 12:16:34 coredumb: you found the problem apparently? 2015-03-18 12:19:43 rpu3uO8PEVZZ: yeah 2015-03-18 12:19:48 stable vs edge 2015-03-18 12:20:00 good job! 2015-03-18 12:20:05 you'll see it's better than i3 ^^ 2015-03-18 12:20:25 can it do sticky windows? 2015-03-18 12:20:53 rpu3uO8PEVZZ: sticky like ? 2015-03-18 12:23:36 a window that is persistent over all virtual desktops without it being a dock 2015-03-18 12:24:13 there's no dock notion in spectrwm afaik 2015-03-18 12:24:36 no sure it's possible to stick windows like that 2015-03-18 12:24:37 well "without it being a dock" 2015-03-18 12:25:02 i must admit i use far from all the features ^^ 2015-03-18 12:30:23 of course 2015-03-18 12:30:54 https://opensource.conformal.com/cgi-bin/man-cgi?spectrwm doesn't look like there's the stickiness notion 2015-03-18 12:32:57 in that case (and from my POV) there's no benefit for me to move from i3wm 2015-03-18 12:34:23 ^^ 2015-03-18 14:09:12 bump fossil pls (v1.32) 2015-03-18 14:11:44 and task v2.4.2 2015-03-18 14:38:34 vkrishn: fossil got noew version numbering scheme? 2015-03-18 14:38:45 yes 2015-03-18 14:38:49 hum 2015-03-18 14:38:59 old version is 20150119112900 2015-03-18 14:39:02 new is 1.32 2015-03-18 14:39:12 there is one release between them 2015-03-18 14:39:38 old number is a trillion or so bigger than new '1' 2015-03-18 14:39:47 the downloaded bin was around 2.3mb 2015-03-18 14:39:57 :) 2015-03-18 14:40:13 now its back to 1.19mb 2015-03-18 14:40:49 vkrishn: the problem is that apk upgrade will think the previous is newer 2015-03-18 14:40:51 I think on change logs they had 1.x scheme 2015-03-18 14:41:01 so you will not get it with apk upgrade 2015-03-18 14:41:48 i mean, how should you know which is newest: 20150119112900 or 1.32? 2015-03-18 14:42:00 oh well 2015-03-18 14:42:11 good that it is in testing still 2015-03-18 14:42:29 maybe convert that to date from, https://www.fossil-scm.org/fossil/timeline?c=version-1.32;y=ci 2015-03-18 14:43:22 but would need APKBUILD changes 2015-03-18 14:43:32 maybe saveas: 2015-03-18 14:56:07 http://lists.alpinelinux.org/alpine-aports/0287.html 2015-03-18 14:56:10 ncopa, please ^^ 2015-03-18 15:02:22 http://bld2.alpinelinux.org/buildlogs/build-edge-x86_64/main/pciutils/pciutils-3.3.0-r0.log 2015-03-18 15:02:34 pciutils needs linux-headers 2015-03-18 15:09:52 i wonder if this might be a fortify error: http://bld1.alpinelinux.org/buildlogs/build-edge-x86/main/pciutils/pciutils-3.3.0-r1.log 2015-03-18 15:10:01 seems only to affect x86 2015-03-18 15:15:26 ncopa, no 2015-03-18 15:15:33 it's missing _syscall definition 2015-03-18 15:15:39 or _syscall5 2015-03-18 15:15:49 that sounds glibc internal thingy 2015-03-18 15:17:21 it might be just missing #include 2015-03-18 15:17:26 since pread() is defined there 2015-03-18 15:18:57 weird that it worked on x86_64 2015-03-18 15:19:06 i think there's platform specific #ifdefs 2015-03-18 15:19:18 ok 2015-03-18 15:19:20 i'm off for today 2015-03-18 15:19:21 ttyl 2015-03-18 15:24:48 looks like glibc stuff indeed 2015-03-18 16:03:31 ncopa: thanks for patching alpine (the MUA) :) 2015-03-18 18:57:37 does anyone know this error: 2015-03-18 18:57:41 /usr/lib/gcc/x86_64-alpine-linux-musl/4.9.2/../../../../x86_64-alpine-linux-musl/bin/ld: warning: libltdl.so.7, needed by /usr/lib/libcanberra-gtk.so.0, not found (try using -rpath or -rpath-link) 2015-03-18 18:57:45 ? 2015-03-18 18:57:59 searching for any clue but with no success 2015-03-18 18:59:20 alacerda: ok 2015-03-18 18:59:44 That's a linker error obvs, with a GUI library. 2015-03-18 19:00:09 You're compiling with gcc. Did you have to fix a header error with any .h files or anything? 2015-03-18 19:00:47 alacerda: do this: "apk add libltdl" 2015-03-18 19:00:48 i didn't touch any file... i am just doing an: abuild -r 2015-03-18 19:01:03 i've added libltdl to makedepends 2015-03-18 19:01:18 but the error continues 2015-03-18 19:01:28 Is it installed? Does libltdl.so.* exist on system? 2015-03-18 19:02:00 "find /usr -name libltdl\* -print" 2015-03-18 19:02:09 mjones_, it is my dev machine nothing is installed: it install when i do: abuild -r 2015-03-18 19:02:47 Then just add it manually with apk and see if that fixes it. 2015-03-18 19:02:56 alacerda: do this: "apk add libltdl" 2015-03-18 19:03:12 dont(you need libltdl-dev in makedepends? 2015-03-18 19:03:28 the is no libltdl-dev 2015-03-18 19:03:42 apk search libltdl 2015-03-18 19:03:48 doesn't show any -dev 2015-03-18 19:05:15 alacerda: are you running 3.1.2? 2015-03-18 19:05:34 yes: VERSION_ID=3.1.2 2015-03-18 19:06:24 I think it is a sub package of something else because I do not see a dir named libltdl nor under main nor under testing... 2015-03-18 19:06:34 Jean-Scotch: there's no -dev and it's trying to link against the runtim dynamic lib 2015-03-18 19:06:39 I guess you need the -dev of the main package 2015-03-18 19:07:10 # apk info --who-owns /usr/lib/libltdl.so.7.3.1 2015-03-18 19:07:19 /usr/lib/libltdl.so.7.3.1 is owned by libltdl-2.4.3-r0 2015-03-18 19:07:50 libltdl-2.4.3-r0 depends on: 2015-03-18 19:07:50 so:libc.musl-x86_64.so.1 2015-03-18 19:08:28 mmm. "cd /usr/lib && ln -s libltdl.so.7.3.1 libltdl.so.7" 2015-03-18 19:08:52 Is this a clean build? 2015-03-18 19:08:57 mjones_, i dont think it is feasible in a APKBUILD 2015-03-18 19:09:17 alacerda: do it manually, seee if it fixes it 2015-03-18 19:09:37 ln: libltdl.so.7: File exists 2015-03-18 19:09:38 :) 2015-03-18 19:10:06 then it's a -rpath problem I guess? 2015-03-18 19:10:22 It doesn't say the sombols are wrong. 2015-03-18 19:10:29 s/sombols/symbols/ 2015-03-18 19:10:41 libtool-dev maybe? 2015-03-18 19:10:41 as apk indo libltdl talks about libtool 2015-03-18 19:10:41 s/indo/info/ 2015-03-18 19:10:42 yes it is a subpackage of libtoll as seen in it's APKBUILD 2015-03-18 19:10:42 *libtool 2015-03-18 19:11:05 Jean-Scotch, let me try ... 2015-03-18 19:13:14 Jean-Scotch, there is no libtool-dev --> http://nl.alpinelinux.org/alpine/v3.1/main/x86_64/ 2015-03-18 19:17:46 alacerda: building from aports git master? or 3.1-stable branch? 2015-03-18 19:18:48 tdtrask, how can I check this? 2015-03-18 19:19:13 run 'git branch' from aports directory 2015-03-18 19:19:25 * master 2015-03-18 19:20:11 ACTION is not surprised you have problems building aports git master on 3.1 box 2015-03-18 19:20:34 you should be running edge 2015-03-18 19:20:40 hum... 2015-03-18 19:20:50 yeah, I guess everyone develops on edge 2015-03-18 19:20:56 except people taking care of backports 2015-03-18 19:21:00 but me :( 2015-03-18 19:21:07 build 3.1 packages on 3.1, ... 2015-03-18 19:21:25 ncopa, told me sth about this before but i didnt ge tthe point.. now i see! 2015-03-18 19:21:32 thanks all 2015-03-18 19:21:34 because you're installing 3.1 dependencies when building edge packages 2015-03-18 19:21:39 they don't sync up 2015-03-18 19:21:40 I am going to install an edge box 2015-03-19 00:23:23 Someone here was working on pdns in testing...? 2015-03-19 01:11:18 The git.alpinelinux.org favicon is the old favicon, not the new one. 2015-03-19 04:13:45 hahaha, I see a few of the ceph dependencies are already in testing courtesy of ncopa! 2015-03-19 04:13:58 Serves me right for not setting up an aports first. 2015-03-19 05:37:09 Best time of the day for beating on code. 2015-03-19 05:37:25 ACTION needs a standing desk. 2015-03-19 06:09:22 openssl-dev is an undocumented dependency to abuild-keygen, and thus to alpine-sdk 2015-03-19 06:34:20 ncopa, testing/wrk would require -march=i586 to build on x86 2015-03-19 06:50:11 mjones_: I'd love a standing desk too, looks incredibly comfortable 2015-03-19 07:09:03 barthalion: Ikea has a new electric-lift model at a very aggressive price; apparently they're in high demand. 2015-03-19 07:11:06 I'm too lazy to travel to the other end of the city to see if they are selling it here as well 2015-03-19 07:11:08 :p 2015-03-19 07:13:43 Do we have thoughts on libressl-portable coexisting with openssl? Libre builds easily, but we have no aport. 2015-03-19 07:14:01 Obviously they share binary name. 2015-03-19 07:16:09 I think fabled or ncopa talked about libressl on alpine-devel some time ago 2015-03-19 07:16:13 (the mailing list) 2015-03-19 07:16:21 and for now we stick to openssl 2015-03-19 07:24:24 hi, i dont mind adding libressl to testing repo 2015-03-19 07:24:57 but will will not jump on every new openssl replacement that pops up 2015-03-19 07:25:28 we might switch to libressl at some point, but currently we wait a bit to see how things develops 2015-03-19 07:26:57 there should come out a new openssl release today? 2015-03-19 07:28:58 ncopa: understood. But how would you like to handle the names? e.g. libressl binary name is still 'openssl'. 2015-03-19 07:29:35 ncopa: yes, but nobody knows when 2015-03-19 07:30:34 ncopa: I assume we want a way to have both installed at the same time. 2015-03-19 07:31:22 ncopa: I don't know if there's any plan to have two different packages satisfy the same dependency, like (openssl|libressl) or (gcc|clang) 2015-03-19 07:32:41 well, clang doesn't satisfy the same dependency after all 2015-03-19 07:32:56 and packaging libressl can be tricky as it might have different abi than openssl 2015-03-19 07:33:13 so apk will cringe 2015-03-19 07:34:06 I read a bunch about the symbol compat for the ABI. 2015-03-19 07:34:21 In fact I read so much I can't remember if they plan for sure to keep the same one. 2015-03-19 07:34:32 Actually, OpenSSL's changes with some regularity, right? 2015-03-19 07:34:50 well, they used to not 2015-03-19 07:34:53 except last release 2015-03-19 07:35:03 ACTION wonders if it shouldn't ship with all names changed to 'libressl', at least for now. 2015-03-19 07:35:06 when they even didn't bump soname, leaving a lot of apps broken 2015-03-19 07:35:07 hi all. Very OT question. I've a stupid device which has a bugged firmware,who's sending tagged packets to vlan 4095. This vlan can't be used (and linux correctly does not allow that). I'm wondering to use a P2P with that device and apply this patch http://comments.gmane.org/gmane.linux.network/244582 (just to recover this device and configure it in a right vlan) 2015-03-19 07:35:16 Do you think it could work? 2015-03-19 07:36:13 I didn't know 4095 can't be used. It's within 4096... 2015-03-19 07:36:33 me too 2015-03-19 07:36:59 but it's kinda reserved for internal purpose 2015-03-19 07:37:17 dot1q itself, it seems 2015-03-19 07:39:36 fcolista: I'd try it. You can't get into the device some easier way...? 2015-03-19 07:43:11 mjones_, no 2015-03-19 07:43:15 no serial 2015-03-19 07:43:18 only net 2015-03-19 07:43:34 there's jtag, but it might need some hw hack 2015-03-19 07:43:46 fcolista: must be a D-Link device :P 2015-03-19 07:44:08 It's Corinex... 2015-03-19 07:44:10 Yeah, you don't want to mess with jtag if you can avoid it, unless you have an electronics bench. 2015-03-19 07:44:11 Powerline 2015-03-19 07:44:18 right... 2015-03-19 07:44:26 fcolista: I'd try the patch. All you have to lose is your time... 2015-03-19 07:44:38 it's fine with that 2015-03-19 07:44:48 Put it on a private cable though 2015-03-19 07:44:55 It's going to be P2P 2015-03-19 07:45:03 since 4095 cannot be added on a switch 2015-03-19 07:45:03 As you said. 2015-03-19 07:45:09 yep :) 2015-03-19 07:46:12 I learned something new today. It's always easier to remember that silly thing not to do. 2015-03-19 08:27:23 non-free repo has amd and intel microcode, but I see no loaders anywhere in the system... 2015-03-19 12:11:15 https://github.com/scadu/aports rebranded polarssl to mbedtls and packages rebuilt. 2015-03-19 12:13:39 merged, thanks 2015-03-19 13:13:26 any idea what is planB features in mbedtls ? 2015-03-19 13:15:17 hmm.. , https://tls.mbed.org/core-features , "Suite B" 2015-03-19 13:24:00 scadu barthalion here http://git.alpinelinux.org/cgit/aports/commit/?id=3a49adfaacb3217928700ba8737ab74be9a1ab6c is the url= wrong. as there is no www in the https://tls.mbed.org/ 2015-03-19 14:30:54 barthalion: i think i remember you talking about removing the lua dependency on ca-certificates the other day, is there a bug i can follow for that? 2015-03-19 14:36:16 andyshinn: lua is there for performance. see #2846 2015-03-19 14:36:52 you will likely need implement it in C to get rid of lua 2015-03-19 14:40:58 hmm, do you know if there were any attempts at optimizing the original shell script? i can't actually find the original 2015-03-19 15:05:49 brilliant 2015-03-19 15:06:00 also openssl-1.0.1 has now gone through re-indent 2015-03-19 15:06:05 and none of the patches apply 2015-03-19 15:06:12 even though it's security only changes 2015-03-19 15:06:14 sigh 2015-03-19 15:11:11 urgh 2015-03-19 15:11:25 so... it's gonna take a while to rebase all openssl-1.0.1 patches 2015-03-19 15:11:38 since the idiots did indent 2015-03-19 15:11:47 so that no patch will apply 2015-03-19 15:11:49 we are not alone ... 2015-03-19 15:12:03 and i need to run now for a while 2015-03-19 15:12:19 maybe patch has "ignore white space" option 2015-03-19 15:14:41 andyshinn: original is probably somewhere in git; I'm going to rewrite it in C 2015-03-19 15:14:48 crow: will fix it in a moment 2015-03-19 15:48:36 fabled: patch --ignore-whitespace .... 2015-03-19 15:49:28 we can impossible be the first who wants to patch 1.0.1... 2015-03-19 15:51:20 mmm 2015-03-19 15:51:22 no worky 2015-03-19 15:51:29 i think it's more than whitespace 2015-03-19 15:51:35 ok i see what you mean... 2015-03-19 15:51:40 sigh... 2015-03-19 15:57:43 looks like 0001-crypto-hmac-support-EVP_MD_CTX_FLAG_ONESHOT-and-set-.patch need to be manually merged :-/ 2015-03-19 15:58:08 that one can be copied from edge 2015-03-19 15:58:13 ok 2015-03-19 15:58:56 but i think the padlock ones are problematic 2015-03-19 15:59:11 it is now we should start regret that we didnt get our patches in to upstream :-/ 2015-03-19 15:59:25 i have a bug open on them 2015-03-19 16:01:27 i'm looking if we can just copy the edge patches 2015-03-19 16:01:33 i did rebase them all for 1.0.2 2015-03-19 16:05:44 ok, all edge patches except the parallel build one applies. not sure if they build or are correct 2015-03-19 16:14:08 ok 2015-03-19 16:14:16 i have a commit - now to just test if things work 2015-03-19 16:16:29 ok. http://sprunge.us/EIBT thats the commit. have not been tested yet. 2015-03-19 16:16:35 but will push after some testing 2015-03-19 16:17:50 ttyl 2015-03-19 16:18:03 i need to run. will maybe push later today if get a moment to test. 2015-03-19 17:49:35 ncopa, your mqtt-exec saved my life 2015-03-19 19:13:13 I see some pkg like snort,collectd use libmodbus, does it not require packging http://libmodbus.org/ ? 2015-03-19 19:36:15 ok 2015-03-19 19:36:24 seems that openssl-1.0.1m does not have HIGH fixes 2015-03-19 19:36:42 (there was one LOW->HIGH CVE classification change, but it was fixed in 1.0.1k already) 2015-03-19 19:36:56 so 1.0.1m has only moderate or low fixes 2015-03-19 19:37:07 i'll do some more testing and push the updated to our stable branches tomorrow 2015-03-19 19:37:43 (now i need sleep -- it's more catastrophic to push untested stuff to stable branches when you are tired... than to delay few hours the moderate fixes) 2015-03-19 19:37:53 ttyl 2015-03-19 19:47:25 vkrishn: I find no reference to modbus in either collectd nor snort packages in _main_ 2015-03-19 19:47:44 vkrishn: there is no reference to building without it, either. 2015-03-19 19:48:10 I think snort has its own implementation, even collecd, but not sure, 2015-03-19 19:50:35 I would guess so. 2015-03-19 19:51:39 Not reinventing the wheel is a good rule of thumb, but sometimes eliminating redundant code can go too far. Like all of Microsoft's products, that have dependencies on all of the rest of Microsoft's products. 2015-03-19 19:53:05 Historically they have zero notion of reimplementations, third-party implementations, or separating the spec from the embodiment. 2015-03-19 19:53:48 RFCs make Microsoft look risible by comparison. 2015-03-19 19:56:00 I wonder what if they make XP free, since it no longer supported 2015-03-19 19:57:02 or is it just an economically evil thought 2015-03-19 19:59:36 making closed-source open is hard, because of third-party stuff. Plus patent indemnification, etc. 2015-03-19 19:59:50 It was hard with Java/JDK. 2015-03-19 20:00:12 ReactOS is the best bet for a free XP, and it's not a good bet. 2015-03-19 20:00:41 Plus it would compete with the new stuff. 2015-03-19 20:01:00 free in sense for personal use 2015-03-19 20:01:31 that way it would be less problematic 2015-03-19 20:02:06 Free for personal use it worthless in a legal sense, the same way that source code that isn't FOSS is mostly useless. 2015-03-19 20:02:06 they already have office free for edu, iirc 2015-03-19 20:02:26 Remember qmail before it was public domained by DJB? A royal pain. 2015-03-19 20:02:58 But they want people to use office, and they want those free-office people to generate proprietary documents and to get angry at the ones who can't or won't open the documents. 2015-03-19 20:03:05 They don't want anybody using XP for any reason. 2015-03-19 20:03:15 :) 2015-03-19 20:03:25 Hell, I dislike most Microsoft things, and _I_ sure as hell don't want anyone using XP. 2015-03-19 20:03:39 Get a modern, reasonably secure-ish operating system. 2015-03-19 20:03:49 With TCP windows larger than 64k, goddamnit. 2015-03-19 20:04:22 And SNI. And a browser that isn't IE6 or almost as bad. And PNG support, and SVG support. 2015-03-19 20:04:36 And Elliptic-Curve crypto support in the Schannel. 2015-03-19 20:06:08 And TRIM/UNMAP for block storage. I barely touch Windows, I don't use it personally, and I can still think of dozens of reasons no one should use XP. 2015-03-19 20:06:50 SMB2 is a lot less bad than SMB1, and you need Vista for SMB2. 2015-03-19 20:08:06 so almost of no use with current protocols 2015-03-19 20:12:27 Anyway, if someone wants something that can use those terrible closed source blob drivers from XP, but be securable and possibly usable one day, check out ReactOS. 2015-03-19 20:12:49 In the meantime, don't buy anything that doesn't support open-source drivers. 2015-03-19 20:16:15 no, just thinking a how easy would have been for vendors to create game console for teenagers using slightly old hardware, without having to think of licenses 2015-03-19 20:19:48 I see. One of the attractions of consoles is security, a certain kind of idempotence or statelessness, and needing zero maintenance. Those things don't apply to old Windows. 2015-03-19 20:20:07 Old PC games are cheap, but now there's Steam and competitors... 2015-03-19 20:25:23 vkrishn: they already have everything free for education 2015-03-19 20:25:40 vkrishn: try to find a linux workstation in anything below university in Poland, good luck 2015-03-19 20:25:49 vkrishn: and it's even hard in universities, actually 2015-03-19 20:29:14 barthalion, yes, I occassionally download newer live versions to try and recommend to family members 2015-03-19 20:29:53 recent was edubuntu 2015-03-19 20:30:59 My impression is that the less-wealthy nations feel like something is being withheld from them when it's suggested they use open source operating systems and software. 2015-03-19 20:31:22 So while wealthy Germany uses quite a lot of Linux, I bet there's not so much in Thailand. 2015-03-19 20:31:35 I switched more than a decade back 2015-03-19 20:32:28 though always had to use them in offices 2015-03-19 20:34:00 steamos looks nice, now that many game frameworks are coming to linux, should be a hit 2015-03-19 20:35:32 In the recent past I had to admin a fair number of Windows servers from Linux desktops. rdesktop and its fork xfreerdp work extremely well for this. 2015-03-19 20:37:49 Education tends to be significantly behind the curve in the States, at all levels. 2015-03-19 20:38:23 Education picked up on Java when industry wanted Java, and has -- rightly or wrongly -- mostly stuck with it. Better than what they used previously. 2015-03-19 20:39:35 Behind-the-curve institutions will end up as customers for the forward-leaning SaaS, PaaS operations, in many cases. 2015-03-19 20:41:00 sometimes have you own staff to pay and handle such infra costs more 2015-03-19 20:41:09 having your* 2015-03-19 20:41:35 Sure, and plenty of cases where it makes sense to do in-house. 2015-03-19 20:42:32 But when your organisation is spending most of its cycles figuring out how to get java and Flash updated so to stem the spate of drive-by malware compromises, that's not an investment in software or infrastructure. 2015-03-19 20:42:46 That's spending 80% of your resources keeping the lights on, as one Gardner study reported. 2015-03-19 20:44:56 ;) 2015-03-19 20:46:21 mjones_: kind of true about Poland, probably 2015-03-19 20:46:39 for quite a long our tax office considered using then-openoffice as unlegal 2015-03-19 20:46:47 because you had no receipt that you bought it 2015-03-19 20:46:57 same about linux distributions 2015-03-19 20:47:34 s/unlegal/illegal/ 2015-03-19 20:47:42 illegal 2015-03-19 20:47:51 doh, ping's so high :c 2015-03-19 20:48:16 brain so slow 2015-03-19 20:51:11 barthalion: I love you, too ;* 2015-03-19 20:51:18 ;) 2015-03-19 20:51:24 its not difficult to create a fore-front org for os software, charging the os at $1/or procesing fee, making things look as needed 2015-03-19 20:51:56 even a goverment agency (Non Profitable) can do 2015-03-19 20:52:15 Bureaucracy is problem enough already. 2015-03-19 20:52:52 Something that is useful though, is being able to pay an OSS foundation for 'support' 2015-03-19 20:53:33 yes 2015-03-19 20:53:36 As in, you don't really need support services, but it's easy to contribute money when your organisation or the taxman see the invoice marked as 'support services'. 2015-03-19 20:54:14 And just maybe you'll have an issue later and they can help you out also, but that's not the real reason for the payment. 2015-03-19 20:55:07 it mainly showing think on paper 2015-03-19 20:56:12 things* 2015-03-19 20:56:21 sleepy :) 2015-03-19 20:59:39 and there are few things like , guarantee, accountability, damage responsibility 2015-03-19 21:01:44 Those don't exist in software, and rarely enough in hardware or tech services. 2015-03-19 21:02:20 Anybody that claims they're paying for indemnification is likely a shortsighted fool, and probably an insecure bungler. 2015-03-19 21:03:09 Unless you're paying an insurer, paying for indemnification is likely a fallacy. 2015-03-19 21:03:35 I did get source-code escrow one time, though. It would have worked, too, if it wasn't for those meddling kids and their dog! 2015-03-19 21:10:06 lol 2015-03-19 21:11:39 took me a while to remember where I heard that :p 2015-03-19 23:58:33 ./configure --without-fuse --without-tcmalloc --without-libatomic-ops 2015-03-19 23:59:13 My first pass at ceph dropped those features, but now I have to go back and figure out the implications... 2015-03-19 23:59:34 I mean, who doesn't want atomic ops? 2015-03-20 00:40:23 libatomic_ops is in main. tcmalloc is pretty much an aesthetic choice, at least for now. 2015-03-20 00:55:36 "TCMalloc is faster than the glibc 2.3 malloc (available as a separate library called ptmalloc2) and other mallocs that I have tested. ptmalloc2 takes approximately 300 nanoseconds to execute a malloc/free pair on a 2.8 GHz P4 (for small objects). The TCMalloc implementation takes approximately 50 nanoseconds for the same operation pair." 2015-03-20 00:58:35 Considering that 2.3 was released over a dozen years ago, I'm not bothered. 2015-03-20 02:01:10 configure seems to be failing a test for XFS_XFLAGS_EXTSIZE when testing xfs.h, even though xfs.h calls that definition from xfs_fs.h. Odd. 2015-03-20 02:04:27 Anyone know how to recusively check for the presence of all referenced headers? I bet cpp can do it. 2015-03-20 02:51:17 Leaving aside the XFS header problem for now, I can't make it 2.0 seconds into a ceph compile before it falls over. Sigh. 2015-03-20 03:06:59 ACTION tries again with the git HEAD. 2015-03-20 03:09:31 Same compile-time error. Pretty bizarre for well-used, even if immature, software. 2015-03-20 04:21:51 oh, @fabled: http://www.zdnet.com/article/you-need-to-apply-the-openssl-patches-today-not-tomorrow/ 2015-03-20 06:38:41 musl malloc is quite fast, at least for mariadb 2015-03-20 06:39:10 you need more than 4 cores to actually make use of jemalloc, probably similiar about tcmalloc 2015-03-20 06:40:55 morning 2015-03-20 06:50:47 ACTION does a ceph compile run on Debian just to confirm his sanity. 2015-03-20 06:55:11 mjones_: not that bizarre I guess, they just don't care about anything other than glibc 2015-03-20 06:56:37 bathalion: this doesn't seem to be musl. At least not directly. The compiler is simply not liking the code, and my terrible C skills are a lot better than my nonexistent C++ skills. 2015-03-20 06:57:26 It's failing the asserts of funtion prototypes, or something like that. Can't paste from a spice session. 2015-03-20 06:57:35 s/funtion/function/ 2015-03-20 06:58:15 Almost like a c++11 dialect-selection issue. 2015-03-20 06:59:31 It's building fine on Debian with gcc 4.7.2. 2015-03-20 07:01:48 I suppose the autotools they're using could have hit a failwhale on gcc 4.9.2. 2015-03-20 07:05:13 mjones_, the zdnet article is not to the point of it 2015-03-20 07:05:21 actually that would be alpine's autotools. 2015-03-20 07:05:22 yes, 1.0.2 is major issue 2015-03-20 07:05:36 @fabled: I haven't read it yet. ;) 2015-03-20 07:05:38 but for 1.0.1, the issue he's talking about was fixed earlier 2015-03-20 07:06:04 hah 2015-03-20 07:06:14 in fact, he says update to 1.0.1k 2015-03-20 07:06:22 it's the *previous* version 2015-03-20 07:06:24 we have it already 2015-03-20 07:06:35 1.0.1m does not have anything with high priority in it 2015-03-20 07:06:40 1.0.2a is urgent 2015-03-20 07:06:52 though we have 1.0.2 in edge only, and it was updated already yesterday 2015-03-20 07:07:06 Understood. Alpine was already on 1.0.1k, previous current. 2015-03-20 07:09:17 s/Alpine/3.1/ 2015-03-20 08:08:01 is it posible to do sys install but afterwards to mount it as ro, and just before installing update to remounte it? Somehow i am not good with RAM install and sill have on apu4c old version as i dont wanna lose data on it with reboot :) 2015-03-20 08:14:10 I used to do it on my thinclients 2015-03-20 08:19:18 crow, what issues did you have with RAM installs 2015-03-20 08:20:19 crow: there were some issues IIRC with particular services, so I had to symlink some directories to tmpfs 2015-03-20 08:30:42 off to bed! 2015-03-20 08:34:46 vkrishn well to have RAM install always up2date, without afraid to lose data :/ 2015-03-20 08:35:13 barthalion i see, will that be pushed to alpine? :) 2015-03-20 08:38:22 bah, I never upstreamed it, and I lost these scripts since 2015-03-20 08:38:27 you could almost get crazy with `lbu include` 2015-03-20 08:46:46 vkrishn i tried it every time i somehow lost something, 2015-03-20 08:46:51 barthalion to bad :/ 2015-03-20 08:47:26 well i hope the "KINGSTON 30GB SSDNow mSATA 6Gbps " will run for long time. 2015-03-20 08:47:46 do we have trim option enabled with alpine? 2015-03-20 10:05:41 re libressl: this is something to think of: http://undeadly.org/cgi?action=article&sid=20150319145126 2015-03-20 10:06:27 ncopa: have you seen the eclipse? 2015-03-20 10:06:42 is it now? 2015-03-20 10:06:45 yeah 2015-03-20 10:06:50 im going out. brb 2015-03-20 10:11:05 can't see shit here :P 2015-03-20 10:13:14 confirmed, nothing to see there, move along. 2015-03-20 10:14:39 isnt that the whole idea of an eclipse :p 2015-03-20 10:14:44 haha 2015-03-20 10:15:28 its really misty here 2015-03-20 10:15:42 we have sunny weather today 2015-03-20 10:18:05 but still was -6°C in the morning. 2015-03-20 10:18:34 seems the moon devoured ncopa 2015-03-20 10:18:53 lel 2015-03-20 10:20:20 i saw it 2015-03-20 10:20:28 awesome! 2015-03-20 10:20:35 haha 2015-03-20 10:20:42 ACTION jelly 2015-03-20 10:21:20 :) 2015-03-20 10:21:33 dont worry, you'll see it next time 2015-03-20 10:21:37 its on 100 yrs... 2015-03-20 10:21:43 :) 2015-03-20 10:21:46 6 years ;-) 2015-03-20 10:21:51 6? 2015-03-20 10:21:55 depends where I guess 2015-03-20 10:21:57 11 in Poland 2015-03-20 10:22:20 some dj just said 6, thought im not sure thats a good reference :) 2015-03-20 10:22:36 thought/though 2015-03-20 10:23:01 it's a trap, the sun is out to get us out of our basements and burn us slowly to death. and humans usually forget in 5-6 years. so of course the sun tries again in about 5-6 years time when people already forget again the fallen victims of today. 2015-03-20 10:23:50 cunningly devilish this sun 2015-03-20 10:24:06 my eyes hurt 2015-03-20 10:24:09 maybe youre right.. 2015-03-20 10:24:10 :) 2015-03-20 10:24:39 you should have using only one eye and save the other for next eclipse 2015-03-20 10:24:46 it's a trap i tell you, also the opto-industry is it's ally 2015-03-20 10:24:52 :D 2015-03-20 11:46:48 trying to protect AL login, applying patch, running, su -n screen 2015-03-20 11:46:51 ;) 2015-03-20 12:32:26 can something be done with distorted octa-logo on, http://wiki.alpinelinux.org/wiki/Main_Page 2015-03-20 12:50:15 maybe a 3d logo, tilted on its axis, facing north-west 2015-03-20 13:11:30 ok? http://wwwtest.alpinelinux.org/posts/Alpine-3.1.3-released.html 2015-03-20 13:12:47 ncopa: can we verify if nfs is working? then we can announce it. 2015-03-20 13:13:18 i though someone tested it and confirmed that it worked? 2015-03-20 13:13:34 but yes, would be nice to doublecheck 2015-03-20 13:13:41 i saw that somebody mentioned the newer version of nfs-utils was not working 2015-03-20 13:13:49 i thought that was edge 2015-03-20 13:13:54 but i guess thats edge 2015-03-20 13:13:59 as i understood it works in v3.1 but not in edge 2015-03-20 13:14:22 if somebody can confirm, its fine by me. 2015-03-20 14:27:28 I use nfs from AL to AL to PXEboot quite a few KVM and I use it as client once booted 2015-03-20 14:28:51 mostly to lauch some post.install scripts but it is still a real usage ;) 2015-03-20 14:30:14 I was not able to use it from inside a LXC even if it is possible with other distro 2015-03-20 14:33:06 yeah that was on edge (I think I was the one who mentioned it) 2015-03-20 14:33:18 not tested on stable and did not yet have to time investigate what happens on edge 2015-03-20 14:33:36 s/to time/time to/ 2015-03-20 14:39:26 please git pull http://git.alpinelinux.org/cgit/alacerda/aports 2015-03-20 14:39:41 It includes some mate packages upgrade and one new aport 2015-03-20 14:40:13 nanomsg has new owners 2015-03-20 17:07:18 ? 2015-03-20 17:11:11 ? 2015-03-20 17:17:28 . 2015-03-20 17:32:14 https://www.freelists.org/post/nanomsg/New-fork-was-Re-Moving-forward,1 2015-03-20 17:32:19 https://www.freelists.org/post/nanomsg/New-fork-was-Re-Moving-forward,10 2015-03-20 17:53:58 <_mjones> Now for the silliness. After a proper-seeming ./configure, make fails to produce object one, with the following errors that I do not understand: 2015-03-20 17:54:20 <_mjones> CXX cls/version/cls_version_client.o In file included from cls/version/cls_version_client.cc:3: In file included from ./include/types.h:21: In file included from ./include/uuid.h:8: In file included from ./include/encoding.h:22: ./include/buffer.h:208:7: error: use of undeclared identifier '__STRING' assert(_raw); ^ ./include/assert.h:141:26: note: expanded from macro 'assert' : __ceph_assert_fail (__STRIN 2015-03-20 17:54:53 <_mjones> Same error with clang. It's obviously something fundamental, but c++ is unfamiliar territory. 2015-03-20 19:02:04 <_mjones> clear 2015-03-20 21:37:53 <_mjones> configure:13744: g++ -c D__cplusplus conftest.cpp >&5 g++: error: D__cplusplus: No such file or directory configure:15507: g++ -c D__cplusplus conftest.cpp >&5 g++: error: D__cplusplus: No such file or directory configure:16024: g++ -c D__cplusplus -DPIC -DPIC conftest.cpp >&5 g++: error: D__cplusplus: No such file or directory 2015-03-20 21:38:46 <_mjones> je ne plus pas de patience. 2015-03-20 21:44:10 <_mjones> "Be aware that, "out of the box", the wrapper only supports C applications, not C++. This is because the C++ libraries and headers are missing from the musl include/library path. The existing libstdc++ is actually compatible with musl in most cases and could be used by copying it into the musl library path, but the C++ header files are usually not compatible. One option may be rebuilding just libstdc++ against musl; however, if 2015-03-20 22:40:44 <_mjones> ok, this shows up in config.log: 2015-03-20 22:40:52 <_mjones> configure:13744: g++ -c D__cplusplus conftest.cpp >&5 g++: error: D__cplusplus: No such file or directory 2015-03-20 22:43:27 <_mjones> The -D isn't being called by ./configure. 2015-03-20 23:28:58 <_mjones> /usr/include/x86_64-linux-gnu/sys/types.h defines loff_t, but there is no such definition of loff_t in Alpine headers. Is this a musl issue? I can fix this occurence; it's included through xfs/linux.h. 2015-03-20 23:29:08 <_mjones> I guess it's usually the same as off_t on 64-bit. 2015-03-20 23:32:46 <_mjones> #define off64_t off_t 2015-03-23 07:16:16 morning 2015-03-23 07:16:23 i think I have found our first fortify error 2015-03-23 07:16:31 compile error 2015-03-23 07:17:18 http://sprunge.us/jAfi 2015-03-23 07:21:51 fabled: how do i disabel FORTIFY_SOURCE? 2015-03-23 07:22:03 ncopa, add -U_FORTIFY_SOURCE to CFLAGS 2015-03-23 07:23:22 i have a feeling gnulib is messing things up 2015-03-23 07:24:46 i think so too, gnulib probably tries to also replace sprintf 2015-03-23 07:24:46 hum 2015-03-23 07:24:52 http://sprunge.us/GdhJ 2015-03-23 07:27:18 ok 2015-03-23 07:27:32 ncopa, how about just disable gnulib sprintf replacing? 2015-03-23 07:27:41 renaming /usr/include/fortify makes it compile again 2015-03-23 07:29:32 fabled: my experience with gnulib is that it is normally simpler to fix anything around gnulib that trying to fix gnulib itself 2015-03-23 07:40:15 weird that -U_FORTIFY_SOURCE did not help 2015-03-23 07:40:20 does -D_FORTIFY_SOURCE=0 help? 2015-03-23 07:41:39 ncopa, ah 2015-03-23 07:41:46 libvirt enables _FORTIFY_SOURCE itself it's not enabled 2015-03-23 07:41:51 #if !defined _FORTIFY_SOURCE && defined __OPTIMIZE__ && __OPTIMIZE__ 2015-03-23 07:41:51 # define _FORTIFY_SOURCE 2 2015-03-23 07:41:51 #endif 2015-03-23 07:41:54 that's why -U does not work 2015-03-23 07:42:00 -D_FORTIFY_SOURCE=0 should work 2015-03-23 07:50:32 ok, this is a bit weird 2015-03-23 07:51:23 so util/virnetlink.h 2015-03-23 07:51:25 so util/virnetlink. 2015-03-23 07:51:41 so util/virnetdev.h pulls in /usr/include/netlink/netlink.h from libnl-dev 2015-03-23 07:51:49 which again pulls in stdio.h 2015-03-23 07:51:59 which will pull in gnulib/lib/stdio.h 2015-03-23 07:52:02 which breaks things 2015-03-23 07:52:42 virenetlink.h 2015-03-23 07:53:37 if i in virnetlink.h pull in stdio.h before the #include 2015-03-23 07:53:39 then it works 2015-03-23 07:54:39 https://gist.github.com/ncopa/0afd7a9779218837830d 2015-03-23 07:54:52 i think it is the define inline hack they do that break things 2015-03-23 07:56:28 https://gist.github.com/ncopa/0afd7a9779218837830d#file-gistfile1-c-L15 2015-03-23 07:56:29 yp 2015-03-23 07:56:30 yup 2015-03-23 07:56:36 that is what breaks things 2015-03-23 08:08:06 http://www.redhat.com/archives/libvir-list/2012-August/msg01699.html 2015-03-23 08:09:20 ncopa, so that breaks gnulib? 2015-03-23 08:09:33 yup 2015-03-23 08:09:34 because fortify-headers uses __inline 2015-03-23 08:09:57 i thought i could use libnl3 instead 2015-03-23 08:09:58 but no 2015-03-23 08:10:34 so next thing would be to check if the libnl is broken in first place so that workaround is needed 2015-03-23 08:11:12 morning 2015-03-23 08:12:29 is AL trying to move from krb5 towards heimdal ? or still uses it ? 2015-03-23 08:13:22 just noticed a krb5 shim https://github.com/heimdal/MKShim, maybe useful for pkgs that still uses krb5 2015-03-23 08:13:51 vkrishn: i think we use krb5 for nfs and heimdal for samba 2015-03-23 08:15:19 ok, then there is v1.13.1 2015-03-23 08:15:40 and samba has v 4.2.0 2015-03-23 08:20:02 tools here seems nice, https://cwrap.org/ 2015-03-23 08:20:43 https://cwrap.org/uid_wrapper.html 2015-03-23 09:05:43 https://github.com/scadu/aports bumped syncthing, lxc, sudo and wine. 2015-03-23 09:07:17 I tried to upgrade Firefox to 31.5.3, but I get this: ../libjs_static.a(jsdtoa.o): In function `_strtod': 2015-03-23 09:07:20 /home/builder/firefox/src/mozilla-esr31/js/src/dtoa.c:1663: undefined reference to `__flt_rounds()' 2015-03-23 09:26:52 cifs-utils uses krb5, and samba uses heimdal, hope its ok 2015-03-23 09:40:00 scadu: thanks! 2015-03-23 09:43:34 ncopa: np 2015-03-23 09:53:34 undefined reference to `__flt_rounds()' 2015-03-23 09:53:37 i got that too now 2015-03-23 09:56:26 hm… 2015-03-23 09:58:27 I'd to try again, but previously it failed after 1-2h so probably you'll fix it quicker ;f 2015-03-23 09:58:49 you had time to think about your life 2015-03-23 09:58:53 what's really important 2015-03-23 09:59:10 barthalion: I was sleeping :< 2015-03-23 09:59:30 yeah, with those 24 core build monsters i dont get so much time to think about my life... 2015-03-23 09:59:49 i wonder if its fortify source that triggers it 2015-03-23 10:00:02 I was wondering the same thing 2015-03-23 10:00:07 error sounds unlikely 2015-03-23 10:00:23 scadu: is your build env up to date? 2015-03-23 10:00:31 i am looking at it 2015-03-23 10:00:33 barthalion: yep. 2015-03-23 10:00:40 trying to build with -D_FORTIFY_SORUCE=0 2015-03-23 10:00:57 yeah, I hoped that it is not 2015-03-23 10:01:12 edge – update and upgrade done before abuild. 2015-03-23 10:12:09 ncopa: hm 2015-03-23 10:15:11 stateless: it happens with -D_FORTIFY_SOURCE=0 too 2015-03-23 10:15:32 but i also get warnings about redefinition 2015-03-23 10:15:44 i will try rebuild ff 36.0.1 2015-03-23 10:15:55 which i know built before fortify source 2015-03-23 10:19:22 ok 2015-03-23 10:19:50 stateless: https://gist.github.com/ncopa/8ece8eed358776abdba1 2015-03-23 10:20:00 that is with ff 36.0.1 which previously built 2015-03-23 10:21:27 it smells like it was introduced with fortify 2015-03-23 10:26:55 media/webrtc/signaling/signaling.gyp: '_FORTIFY_SOURCE=2', 2015-03-23 10:28:34 thats probably it 2015-03-23 10:31:57 hm no, that is only added for OSX 2015-03-23 10:34:38 ok 2015-03-23 10:34:42 interesting 2015-03-23 10:34:56 ncopa: well, I had some error about webrtc. what about --disable-webrtc? 2015-03-23 10:35:05 it happens even if i rename /usr/include/fortify 2015-03-23 10:35:07 or whatever switch like this. 2015-03-23 10:36:54 not fortify then 2015-03-23 10:37:25 i think it is this: http://git.musl-libc.org/cgit/musl/commit/?id=559de8f5f06da9022cbba70e22e14a710eb74513 2015-03-23 10:39:57 ncopa, how mqtt-exec can take an argument from the topic sent by mosquitto_pub? 2015-03-23 10:40:08 e.g. mosquitto_pub -d -t test -m "go" 2015-03-23 10:40:24 mqtt-exec -v iirc 2015-03-23 10:40:27 this "go" can become an argument that the script executed by mqtt-exec can take? 2015-03-23 10:41:06 -v is verbose 2015-03-23 10:41:34 the -v will make $1 to be the topic and $2 the message 2015-03-23 10:41:53 without -v $1 is message and executed program does not know the topic 2015-03-23 10:42:08 ah 2015-03-23 10:42:09 ok 2015-03-23 10:42:20 it is similar to mosquitto_sub 2015-03-23 10:43:01 so /usr/bin/mqtt-exec -h localhost -v -t test -- /tmp/test 2015-03-23 10:43:25 with mosquitto_pub -d -t test -m "go", it runs /tmp/test 2015-03-23 10:43:49 i'd like that "go" it's an argument that /tmp/test can take 2015-03-23 10:57:53 Got it ncopa. gotta pass $2 to /tmp/test to have "go" as argument 2015-03-23 10:57:57 thx for the hint 2015-03-23 11:29:53 ncopa: does libvirt build with fortify now? or do we have a workaround for this? 2015-03-23 11:30:00 I can test when I get home 2015-03-23 11:33:18 please git pull http://git.alpinelinux.org/cgit/alacerda/aports 2015-03-23 11:33:18 It includes some mate packages upgrade and one new aport 2015-03-23 11:33:21 ncopa, ^^ 2015-03-23 12:13:38 stateless: libvrit builds with fortify now 2015-03-23 12:23:17 alacerda: thanks! 2015-03-23 12:23:51 ncopa, Thank *You* 2015-03-23 13:07:00 https://github.com/scadu/aports upgraded exo, xfwm4, libxfce4ui, xfce4-session, xfce4-power-manager and xfdesktop 2015-03-23 13:14:41 also, there's new busybox, but probably I shouldn't touch it. 2015-03-23 13:16:47 http://tinyurl.com/kwnxazf 2015-03-23 13:20:49 kek 2015-03-23 13:21:13 destroying the whole system is possible :> 2015-03-23 13:21:31 in case if something went wrong :f 2015-03-23 13:24:41 ncopa: would that be possible to enable CONFIG_PAX_XATTR_PAX_FLAGS in the config of linux-grsec ? 2015-03-23 13:25:07 coredumb, not yet 2015-03-23 13:25:10 we are working on that 2015-03-23 13:25:17 any reason ? 2015-03-23 13:25:35 coredumb, our .apk packages provide ELF binaries which are pax marked 2015-03-23 13:25:43 i means *not easy reason* 2015-03-23 13:25:46 we don't want user to do it. we want packages to ship the correct markings 2015-03-23 13:26:03 if you change to xattr flags, the elf flags stop working -> broken system 2015-03-23 13:26:09 fabled: what about non packaged apps that can't be ELF modified ? 2015-03-23 13:26:32 i know tricky :D 2015-03-23 13:26:36 anyway 2015-03-23 13:26:43 we are migrating to xattr pax flags 2015-03-23 13:26:52 ok 2015-03-23 13:26:54 see http://bugs.alpinelinux.org/issues/3027 2015-03-23 13:26:59 any deadline ? 2015-03-23 13:27:17 to know when i'll be able to use skype :P 2015-03-23 13:27:23 it's missing apk audit support, and the build system changes (commits are ready but waiting) 2015-03-23 13:27:30 i hope to finish it this week 2015-03-23 13:27:30 ok 2015-03-23 13:27:40 go to hear :) 2015-03-23 13:27:44 good* 2015-03-23 13:27:56 basically, it's just: 2015-03-23 13:28:01 1. implement xattr audit support in apk-tools 2015-03-23 13:28:08 2. tag new apk-tools and update it in edge 2015-03-23 13:28:15 3. push the abuild + paxmark changes 2015-03-23 13:28:23 4. rebuild all packages with pax marked binaries 2015-03-23 13:28:34 5. change to CONFIG_PAX_XATTR_PAX_FLAGS 2015-03-23 13:28:54 i think we'll keep the ELF markings for backwards compatibility for a little while 2015-03-23 13:29:34 of the above steps, #1 and #2 needs doing; #3 is pending push; and #4 and #5 are relatively trivial 2015-03-23 13:30:42 but yes, plan is to finish it this week. and definitely to have it done for 3.2-release 2015-03-23 13:30:52 cool 2015-03-23 13:31:13 skype will wait a bit 2015-03-23 13:31:22 i'll update glibc wiki page 2015-03-23 13:31:38 coredumb, so skype works with pax markings? 2015-03-23 13:31:59 supposed to seeing gentoo hardened mailing list 2015-03-23 13:32:24 ELF marking makes the binary exit nicely as skype doesn't like running with modified binary 2015-03-23 13:32:36 i see 2015-03-23 13:32:49 it works on linux-vanilla + alpine ? 2015-03-23 13:33:58 have not tried actualyl 2015-03-23 13:34:23 you could try that to make sure there's no musl issues 2015-03-23 13:35:22 fabled: i run it in a debian chroot 2015-03-23 13:35:29 indeed it can run on musl directly 2015-03-23 13:36:38 so when ELF amrking i can see it running and exiting 2015-03-23 13:48:22 coredumb, but that's not guarantee that all works when running with paxmark and/or on linux-vanilla 2015-03-23 13:48:38 skype might might some other assumptions that do not work 2015-03-23 13:48:51 but it's a good start that the binary starts... means there should not be any missing symbols 2015-03-23 13:49:38 fabled: indeed but while i'm in remote i can't reboot on linux-vanilla ;) 2015-03-23 13:51:27 ncopa: ah nice :) 2015-03-23 13:51:34 you had me there for a second :P 2015-03-23 14:45:21 scadu: big thanks for those 2015-03-23 14:54:00 ncopa: you're welcome. 2015-03-23 15:12:46 ncopa: my APKBUILD for spectrwm is completely buggy and doesn't work at all, i guess it's better to jsut remove it completely from testing. what's the procedure ? should i send a patch of a git rm ? 2015-03-23 15:17:33 coredumb: I can purge it for you 2015-03-23 15:17:48 sometimes we move things to unmaintained 2015-03-23 15:17:54 but probably better just remove it 2015-03-23 15:32:28 yep better 2015-03-23 15:32:36 would be a hassle to debug 2015-03-23 15:32:53 it's segfaulting everywehre when exec apps 2015-03-23 15:33:03 better just remove it 2015-03-23 15:34:16 ncopa: care to apply http://sprunge.us/OSac ? 2015-03-23 15:44:13 clandmeter: http://wwwtest.alpinelinux.org/ 2015-03-23 15:44:24 i wonder if we should sorten it to "pkgs" 2015-03-23 15:46:55 packages seems more clear to me, dont know about the space. 2015-03-23 15:47:25 i think "packages" is good 2015-03-23 15:48:27 i think next ff error is due to fortify 2015-03-23 15:49:48 http://sprunge.us/QhOd 2015-03-23 15:53:10 ncopa: ok if you can put it live, then ill send some info to the ml. 2015-03-23 15:53:55 done 2015-03-23 15:53:57 thanks 2015-03-23 15:54:00 good job! 2015-03-23 15:54:25 thx! 2015-03-23 16:07:49 ok, fortify+firefox=broken 2015-03-23 16:07:56 will look at that tm 2015-03-23 16:08:03 have a nice evening 2015-03-23 18:20:58 ncopa: copyright is outdated on new pkgs page 2015-03-23 18:36:09 barthalion: its in sync with www ;-) 2015-03-23 18:38:19 wiki's correct. :P 2015-03-23 18:42:20 somebody can send me a git patch and ill update it :p 2015-03-23 21:59:31 clandmeter: it's really cool, good job 2015-03-24 10:29:50 i get fortify warnings when compiling inkscape 2015-03-24 10:29:52 CXX ui/tools/measure-tool.o 2015-03-24 10:29:52 :0:0: warning: "_FORTIFY_SOURCE" redefined 2015-03-24 10:29:52 : note: this is the location of the previous definition 2015-03-24 10:29:52 CXX ui/tools/mesh-tool.o 2015-03-24 10:29:52 :0:0: warning: "_FORTIFY_SOURCE" redefined 2015-03-24 10:29:53 : note: this is the location of the previous definition 2015-03-24 10:46:33 stateless: i'm building libreoffice with fortify now... 2015-03-24 10:48:07 https://github.com/scadu/aports updated mariadb, git, git-manpages and gitolite. 2015-03-24 10:54:15 how would I test oscam package with older revision? i was always doing apkbuild snapshot but is there way to specify date/revision ? 2015-03-24 11:08:51 is this posible at all? 2015-03-24 11:09:32 crow: http://git.alpinelinux.org/cgit/abuild/tree/abuild.in#n1822 2015-03-24 11:17:02 huh :) 2015-03-24 11:17:08 checking 2015-03-24 11:17:36 ncopa, i've just sent 2 patches to fix a bug in MATE environment. If you can apply them as soon as them reach the mail list i'd appreciate. 2015-03-24 11:18:19 I am almost finishing the work on MATE Desktop 2015-03-24 11:18:23 thanks 2015-03-24 11:32:22 http://lists.alpinelinux.org/alpine-aports/0313.html 2015-03-24 11:32:28 http://lists.alpinelinux.org/alpine-aports/0314.html 2015-03-24 11:32:32 ncopa, ^^ please 2015-03-24 11:32:41 lunch 2015-03-24 11:33:01 alacerda: sprunge them 2015-03-24 11:33:53 is it HEAD~2 ? 2015-03-24 11:34:57 http://sprunge.us/afAN 2015-03-24 11:35:02 clandmeter, ^^ 2015-03-24 11:37:02 alacerda: ok im building them local 2015-03-24 11:37:15 thanks :) 2015-03-24 11:48:23 ncopa: I do not understand why programs keep redefining _FORTIFY_SOURCE themselves 2015-03-24 11:49:23 ncopa, stateless : it's harmless warning. and yes, they should have #ifndef guards 2015-03-24 11:49:30 though 2015-03-24 11:49:40 if redefine comesfrom command line 2015-03-24 11:49:42 hmm 2015-03-24 11:50:10 hum 2015-03-24 11:50:14 that might be my mistake 2015-03-24 11:50:26 ~ $ gcc -U_FORTIFY_SOURCE -c a.c 2015-03-24 11:50:26 ~ $ gcc -D_FORTIFY_SOURCE=2 -c a.c 2015-03-24 11:50:26 :0:0: warning: "_FORTIFY_SOURCE" redefined 2015-03-24 11:50:26 : note: this is the location of the previous definition 2015-03-24 11:50:29 onemptyfile 2015-03-24 11:50:40 i think the gcc patch from gentoo is incorrect 2015-03-24 11:51:14 http://git.alpinelinux.org --> The favico is still the old one 2015-03-24 11:52:25 fabled: ah likely the case yes 2015-03-24 11:52:48 http://git.alpinelinux.org/cgit/aports/tree/main/gcc/10_all_default-fortify-source.patch 2015-03-24 11:53:52 is it possible to do this conditionally? 2015-03-24 11:54:25 i think not easily 2015-03-24 11:54:30 the built-ins are defined first 2015-03-24 11:54:35 right yeah 2015-03-24 11:54:37 and only after those cmdline 2015-03-24 11:54:45 that's why -U_FORTIFY_SOURCE works 2015-03-24 11:54:58 yeah makes sense 2015-03-24 11:55:32 I suppose to get rid of the warnings, we should patch the programs with ifndef guards like you said (if we care at all for the warnings that is) 2015-03-24 11:55:45 well 2015-03-24 11:56:00 i think the warning comes when program adds that to CFLAGS 2015-03-24 11:56:07 ah 2015-03-24 11:56:18 :0:0: warning: "_FORTIFY_SOURCE" redefined 2015-03-24 11:56:23 yeah 2015-03-24 11:56:28 perhaps we should just define it as =2 2015-03-24 11:56:31 or =1 2015-03-24 11:56:44 you don't get warning if they redefinition is identical 2015-03-24 11:56:57 and the headers check for __OPTIMIZE__ too 2015-03-24 11:58:10 in your case the redefinition was identical but you still got a warning? 2015-03-24 12:03:58 Thanks. 2015-03-24 12:04:27 >>> libreoffice-impress*: Tracing dependencies... 2015-03-24 12:04:27 >>> ERROR: libreoffice-impress*: libbasegfxlo.so: path not found 2015-03-24 12:04:34 fabled: seen that^^ 2015-03-24 12:04:49 ncopa, libreoffice uses rpath 2015-03-24 12:05:04 same as openjdk7 2015-03-24 12:05:10 usr/lib/libreoffice/program 2015-03-24 12:05:16 the .so files are in rpath, but abuild nags 2015-03-24 12:05:29 y 2015-03-24 12:05:37 only that path? 2015-03-24 12:05:44 there is a variable to help abuild 2015-03-24 12:06:08 ldpath 2015-03-24 12:07:21 oh 2015-03-24 12:07:31 i think only that path 2015-03-24 12:08:07 i'll rebuild again 2015-03-24 12:08:10 and fix it 2015-03-24 12:08:27 i am upgrading poppler, which requires libreoffice rebuild 2015-03-24 12:08:32 so i am upgrading libreoffice 2015-03-24 12:08:45 amazing 2015-03-24 12:08:48 it built 2015-03-24 12:08:49 :) 2015-03-24 12:09:56 re adding -D_FORTIFY_SOURCE=2 on cmdline 2015-03-24 12:10:05 inkscape unconditionally adds it 2015-03-24 12:10:14 the workaround: http://sprunge.us/IDaR 2015-03-24 12:10:59 aha :) 2015-03-24 12:12:41 ha 2015-03-24 12:12:50 i wondered what gentoo does 2015-03-24 12:13:03 the answer is: http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/media-gfx/inkscape/files/inkscape-0.91_pre3-cppflags.patch?revision=1.1&view=markup 2015-03-24 12:13:09 they do the same 2015-03-24 12:13:50 basically, i think the compile flags is the packagers business 2015-03-24 12:14:15 and upstream should respect whatever compile options packagers prefers 2015-03-24 12:29:28 where does this $version comes from? http://git.alpinelinux.org/cgit/abuild/tree/abuild.in#n1846 2015-03-24 12:30:49 http://git.alpinelinux.org/cgit/abuild/tree/abuild.in#n1841 2015-03-24 12:30:56 local _version 2015-03-24 12:31:06 but hum, that is wrong 2015-03-24 12:31:14 abuild shouldnt use _ prefixed variables 2015-03-24 12:31:27 not even local 2015-03-24 12:31:37 local _version=${verbase:-0}_git${_date} 2015-03-24 12:32:49 i would like to fix the APKBUILD format 2015-03-24 12:33:00 ncopa as seems you wrote it, is there way to spacify revision for the snapshot 2015-03-24 12:33:11 i didnt 2015-03-24 12:33:31 i would like to checkout this using apkbuild snapshot r10389 2015-03-24 12:33:43 iirc it was clandmeter, but only git blame will tell you the truth 2015-03-24 12:33:50 sounds like svn and not git 2015-03-24 12:34:28 erm. abuild snapshot r10389 2015-03-24 12:34:47 yes i did 2015-03-24 12:34:49 i think you need set reporev in APKBUILD 2015-03-24 12:34:50 blame it on me 2015-03-24 12:34:58 git blames clandmeter :) 2015-03-24 12:35:02 :p 2015-03-24 12:35:14 ah :) 2015-03-24 12:35:44 crow: specify $reporev 2015-03-24 12:45:23 clandmeter great Checked out revision 10389 :) 2015-03-24 13:06:05 shared-mime-info is still dog slow 2015-03-24 13:06:21 fabled: wasnt the 1.4 release suppose to include something that made it faster? 2015-03-24 13:06:58 dunno 2015-03-24 13:07:18 iirc, it's the fdatasync() issue you looked at 2015-03-24 13:16:03 arg, i missed the polarssl renaming. 2015-03-24 13:16:12 ACTION goes of to rebuild 2015-03-24 13:17:13 crow: i updated oscam in edge, if you are interested. 2015-03-24 13:30:00 stateless: firefox fails with: https://gist.github.com/ncopa/62049f773aa572405808 2015-03-24 13:35:44 this is how fim.o is built: http://sprunge.us/bSiO 2015-03-24 13:49:41 clandmeter thanks but i am compiling my own as i add and disable some unneeded feature 2015-03-24 14:02:49 ncopa: having a look now 2015-03-24 14:02:54 i think i found it 2015-03-24 14:02:56 and 2015-03-24 14:03:04 i think it is exactly same issue 2015-03-24 14:03:33 /home/ncopa/aports/testing/firefox/src/mozilla-release/media/webrtc/signaling/./src/sipcc/core/sdp/sdp_os_defs.h:32:2: error: #error INLINE!!!! 2015-03-24 14:03:34 #error INLINE!!!! 2015-03-24 14:03:34 ^ 2015-03-24 14:03:44 i added that #error INLINE!!! 2015-03-24 14:03:55 #ifndef __GNUC_STDC_INLINE__ 2015-03-24 14:03:55 #define inline 2015-03-24 14:03:55 #error INLINE!!!! 2015-03-24 14:03:55 #endif 2015-03-24 14:04:30 i think undefing __inline in fortify-headers will solve it 2015-03-24 14:04:35 hm yes 2015-03-24 14:04:43 can you apply the patch? 2015-03-24 14:04:51 if this works, maybe we can apply it in fortify-headers 2015-03-24 14:05:02 we can also remove the libvirt patch then I suppose 2015-03-24 14:09:16 i would be ok with patching firefox too for now 2015-03-24 14:09:23 ok :) 2015-03-24 14:09:24 to see if there are more of them 2015-03-24 14:09:33 yeah 2015-03-24 14:09:48 apparently there are some that does #define inline 2015-03-24 14:10:08 i applied the fortify-headers patch and are building firefox now 2015-03-24 14:10:16 would be interesting to see if firefox 36 will start :) 2015-03-24 14:10:20 i believe it will complete build 2015-03-24 14:10:28 yeah 2015-03-24 14:10:42 does libreoffice run? 2015-03-24 14:10:55 it faild build on build server 2015-03-24 14:10:57 dunno why 2015-03-24 14:11:01 it could actually be same issue 2015-03-24 14:11:15 probably other issue 2015-03-24 14:11:17 http://bld2.alpinelinux.org/buildlogs/build-edge-x86_64/testing/libreoffice/libreoffice-4.4.2.1-r0.log 2015-03-24 14:12:15 that made my browser freeze.. lol 2015-03-24 14:12:35 ok back to life 2015-03-24 14:14:59 fabled: does that tell you anything? http://bld2.alpinelinux.org/buildlogs/build-edge-x86_64/testing/libreoffice/libreoffice-4.4.2.1-r0.log 2015-03-24 14:16:35 ncopa, no 2015-03-24 14:16:41 stateless: ff built 2015-03-24 14:16:43 recent changes in musl? 2015-03-24 14:16:49 no 2015-03-24 14:16:51 Test name: testMathRad::Import_Export_Import 2015-03-24 14:16:51 equality assertion failed 2015-03-24 14:16:51 - Expected: sqrt{4} 2015-03-24 14:16:51 - Actual : nroot{3}{x+1} 2015-03-24 14:16:56 hum 2015-03-24 14:17:05 ooxmlexport2.cxx:334:Assertion 2015-03-24 14:17:09 what is weird is that it built in my build env 2015-03-24 14:17:20 which should run latest musl 2015-03-24 14:18:08 CHECK_FORMULA( "sqrt {4}", getFormula( getRun( getParagraph( 1 ), 1 ))); 2015-03-24 14:18:08 CHECK_FORMULA( "nroot {3} {x + 1}", getFormula( getRun( getParagraph( 1 ), 2 ))); 2015-03-24 14:18:12 check some sort of off by one 2015-03-24 14:18:16 smells* 2015-03-24 14:18:34 i think i might have had the fortify patch applied on dev box 2015-03-24 14:18:58 http://sprunge.us/TUgg 2015-03-24 14:19:18 Trutz sent me that on email 2015-03-24 14:19:45 due to http://git.alpinelinux.org/cgit/aports/tree/main/libvirt/fix-fortify-virnetlink.patch 2015-03-24 14:20:11 apparently firefox's webrtc also has a #define inline 2015-03-24 14:20:21 and the same fortify-headers patch fixes firefox build too 2015-03-24 14:22:44 just curious, why does git depend on lua? 2015-03-24 14:23:10 gitweb? 2015-03-24 14:25:00 ncopa: yeah I was CC-ed as well that patch 2015-03-24 14:25:43 ScrumpyJack: probably due to curl and ca-certificates 2015-03-24 14:25:58 I will sit on that patch until maybe we uncover more programs that do this like you said 2015-03-24 14:26:15 ok 2015-03-24 14:26:19 i'll patch firefox then 2015-03-24 14:26:44 we should probably create a testcase for it 2015-03-24 14:33:09 building firefox again, now with patch 2015-03-24 14:33:32 interesting, libreoffice built on x86 2015-03-24 14:33:35 but not x86_&4 2015-03-24 14:33:50 i wonder if it is some resource related issue 2015-03-24 14:33:56 out of /tmp or memory 2015-03-24 14:34:06 or parallel issue of some sort 2015-03-24 15:29:51 hmm, sudo doesn't seem to pick up proxy.sh in /etc/profile.d 2015-03-24 15:30:14 should i expect it sudo to source proxy.sh somehow? 2015-03-24 15:32:24 sounds like /etc/profile should source /etc/profile.d 2015-03-24 15:33:06 your /etc/profile should look something like this: http://sprunge.us/CBRO 2015-03-24 15:46:50 ncopa: that's exactly what it looks like :) 2015-03-24 16:02:23 here comes libressl 2015-03-24 16:02:30 i hope it does not break anything 2015-03-24 16:02:36 i think it should be ok 2015-03-24 16:05:00 hmm 2015-03-24 16:05:10 how can i force a downgrade of a package from repo? 2015-03-24 16:05:18 oh shit 2015-03-24 16:05:22 its still has the old pkg 2015-03-24 16:07:20 still same, anyone? 2015-03-24 16:20:20 please pull http://git.alpinelinux.org/cgit/alacerda/aports/ 2015-03-24 16:20:29 I moved MATE Desktop to main 2015-03-24 16:20:58 and added myself as Maintainer 2015-03-24 16:28:55 not sure i told anybody 2015-03-24 16:29:05 ruby sucks 2015-03-24 16:37:32 ACTION can confirm 2015-03-24 16:38:49 i wonder why people who depend on bloat come to a bloatless distro like alpine, and then want immediately to port the bloat over. 2015-03-24 16:39:49 ? 2015-03-24 16:40:21 maybe this: http://www.wussu.com/fractals/images/wd950112.gif 2015-03-24 16:44:27 probably that was about MATE, but I have no idea why it's bloated. 2015-03-24 16:45:13 oh everyone should use tty or zomg lightweight twm. 2015-03-24 16:47:41 no! sorry, no! it was the ruby thread i was alluding to 2015-03-24 16:50:31 also there was this skype thing today... /o\ 2015-03-24 16:52:48 I see. 2015-03-24 16:52:56 shame on me ;f 2015-03-24 16:56:27 actually i have no clue what MATE could be that makes sense to package into alpine instead of a gourd. i guess some x11 env? 2015-03-24 16:58:04 MATE is a fork of the old gnome (gtk2) desktop - was made popular by linux mint 2015-03-24 16:58:29 that sounds relatively unbloated 2015-03-24 16:59:38 it has a wikipedia page 2015-03-24 17:00:24 someone theorized that we're getting people fleeing from systemd 2015-03-24 17:00:58 so they don't care about bloat. They just want an os systemd-free, because AC on /. said so 2015-03-24 17:01:57 no systemd is a major attraction 2015-03-24 17:02:07 it was also for me 2015-03-24 17:02:16 but also the purity of alpine 2015-03-24 17:02:39 I'm trying to make xpra work so X apps can be run in LXC containers in alpine 2015-03-24 17:04:06 hum... it seems ncopa has already left 2015-03-24 17:04:21 clandmeter can you pull please? 2015-03-24 17:05:09 yeah, I'm perfect example of people fleeing from systemd 2015-03-24 17:05:15 2015-03-24 17:06:58 if anyone has any ideas why I get this relocation error in musl I would be most grateful http://xpra.org/trac/ticket/820 2015-03-24 17:10:54 barthalion: +1 2015-03-24 19:25:15 http://sprunge.us/NeIN 2015-03-24 19:25:22 ncopa, ^^ 2015-03-24 21:51:50 hi, I'm trying to compile the broadcom wl driver for alpine. What packages do I need install to build kernel modules? 2015-03-24 21:54:34 I've install alpine-sdk and linux-grsec, but the Makefile for the wl driver is complaining about the lack of a build directory at /lib/modules/`uname -r`, and then complains about the lack of a Makefile at that path if I create the directory 2015-03-24 22:05:34 linux-headers 2015-03-24 22:05:43 oops 2015-03-25 07:12:24 winmillwill: linux-grsec-dev 2015-03-25 10:35:45 stateless: gir-scanner gives those errors/warnings: https://dpaste.de/PP0j 2015-03-25 10:35:52 i dont know if they break anything though 2015-03-25 10:46:28 clandmeter: i'm already here :) 2015-03-25 10:46:53 :) 2015-03-25 10:47:39 im currently trying to package redmine (again) on edge, but its giving me a headache 2015-03-25 10:47:56 im not sure its worth it 2015-03-25 10:48:03 as usual :P 2015-03-25 10:48:21 I wonder if we could just use omnibus for that 2015-03-25 10:48:36 kunkku: you are also maintaining ruby pkgs right? 2015-03-25 10:49:51 mosez: i would prefer to package it, but our current way of packaging its a nightmare. 2015-03-25 10:50:10 i'm not a fan of the omnibus style. but for maintainance it should be easier to handle for sure 2015-03-25 10:50:57 currently alpine supports only single rubygem version, right? 2015-03-25 10:51:04 that's already a nightmare to handle :) 2015-03-25 10:51:19 correct 2015-03-25 10:52:42 yeah, omnibus is far from perfect 2015-03-25 10:52:45 i'm not totally familar with the apk format yet. i would suggest something like version suffixed packages. if apk is able to do that it shold include a "provides gemname" but the package is named gemname-1_2 or something like that 2015-03-25 10:52:51 violates probably every possible packaging policy 2015-03-25 10:53:10 but we can drop sotorm of gems from aports and be happy 2015-03-25 10:53:44 at suse it's currently handled that the latest version has no version suffix and older versions get this suffix so that it is still installable 2015-03-25 10:53:56 clandmeter: yes, I am using this to deal with the deps: http://git.alpinelinux.org/cgit/abuild/tree/apkbuild-gem-resolver.in 2015-03-25 10:54:49 repackaging a "package manager" is always a pain 2015-03-25 10:55:07 mosez: in alpine, the only way to support multiple versions is adding new packages with new name 2015-03-25 10:55:26 ncopa: for suse it's always a new package as well 2015-03-25 10:55:26 so ruby-foo1, ruby-foo2 2015-03-25 10:55:56 the above script should be improved to automatically update the APKBUILD files 2015-03-25 10:56:04 https://build.opensuse.org/project/show/devel:languages:ruby:extensions just search for rubygem-rails- 2015-03-25 10:56:30 kunkku: deal with, as in what exactly? 2015-03-25 10:56:44 we are using gem2rpm to avoid writing all the spec files by hand 2015-03-25 10:57:24 dependency hell 2015-03-25 10:57:31 this script fetches the gem from rubygems.org and reads the metadata like the dependencies 2015-03-25 10:59:22 https://build.opensuse.org/package/view_file/devel:languages:ruby:extensions/rubygem-rails-4_2/rubygem-rails-4_2.spec?expand=1 2015-03-25 10:59:47 will that ship all rails deps in the same rubygem-rails-4_2 package? 2015-03-25 11:01:14 mosez: we have similiar tool in Arch 2015-03-25 11:01:21 maybe we could reuse it? 2015-03-25 11:01:40 got to go for lunch, my colleagues are waiting 2015-03-25 11:01:44 https://github.com/anatol/gem2arch 2015-03-25 11:01:48 will be bakc in something like 30 minutes 2015-03-25 11:02:06 ncopa: no, it's just the single gem without the deps 2015-03-25 11:02:46 ncopa: but it got an internal dependency that installs the other gems as rpm 2015-03-25 11:03:05 and each depend gem will have its own rpm, right? 2015-03-25 11:03:25 yeah 2015-03-25 11:04:38 but you have also rubygem-rails-3_2 2015-03-25 11:05:01 which might depend on same gem as rails-4_2, but different versions 2015-03-25 11:05:37 ncopa, thanks for pulling my aports 2015-03-25 11:05:43 what do you of this: http://sprunge.us/NeIN ? 2015-03-25 11:06:03 i've done nothing yet 2015-03-25 11:06:25 and i have still not done anything with the setup-desktop thingy you did earlier 2015-03-25 11:06:59 hm 2015-03-25 11:07:13 alacerda: i wonder if it would be n idea to ship those in separate package 2015-03-25 11:07:28 and separate git repo 2015-03-25 11:07:49 are you talking about MATE or setup-pentest ? 2015-03-25 11:07:52 so we dont end up with 100 setup- on every alpine install 2015-03-25 11:07:53 both 2015-03-25 11:08:22 those are kind of install templates, right? 2015-03-25 11:08:39 yes 2015-03-25 11:08:48 setup-pentest will give you an alpine box designed for pentesting 2015-03-25 11:09:01 setup-desktop will give you an alpine box to be used as desktop 2015-03-25 11:09:18 setup-gparted will give you an alpine box for running gparted 2015-03-25 11:09:30 i think we should move those to a separate package 2015-03-25 11:09:51 probably alos a separate git repo 2015-03-25 11:09:58 hum... it sounds good 2015-03-25 11:10:05 what about more generic scripts? 2015-03-25 11:10:10 with yaml parsing? 2015-03-25 11:10:19 yes 2015-03-25 11:10:23 would be nice 2015-03-25 11:10:34 now I only need time to actually write it 2015-03-25 11:10:41 :) 2015-03-25 11:10:55 whoops, i forgot lunch 2015-03-25 11:10:56 bbl 2015-03-25 11:11:04 hehe 2015-03-25 11:35:24 gem2rpm -> rpm2aport (ncopa's hidden script), would save some effort 2015-03-25 11:47:58 please git pull git://git.alpinelinux.org/fab/aports 2015-03-25 11:48:09 there are a couple of updates 2015-03-25 11:53:36 back 2015-03-25 11:54:27 ncopa: sure, because of that all the gems have to be prepared with a version suffix. and a single gem can get installed in multiple versions of course. 2015-03-25 11:55:19 a colleague suggested to implement apk build capabilities into the open buildservice while our hackweek :D 2015-03-25 12:04:52 open buildservice? 2015-03-25 12:10:05 ncopa: can you give me a new url for the gir-scanner errors/warnings? the other has expired 2015-03-25 12:12:03 stateless: http://sprunge.us/GWDQ 2015-03-25 12:12:48 clandmeter: i think we should rename the ruby-redmine-* packages to ruby-rails-* 2015-03-25 12:12:56 that is a start 2015-03-25 12:13:14 we could try have ruby- as the latest 2015-03-25 12:13:27 and use kunkku's script to detect dep breakages 2015-03-25 12:14:10 if upgrade introduces a breakage we can add a suffixed ruby- pkg 2015-03-25 12:14:21 and we should have gem2apkbuild 2015-03-25 12:14:44 i can start clean up the .../bin/* symlinks 2015-03-25 12:16:59 do we have other rails apps except redmine? 2015-03-25 12:18:28 renaming them is ok, but it doesnt solve the problem. redmine needs specific versioning besides rails 2015-03-25 12:21:58 ncopa: then, for while we suspend the setup- works untill barthalion or someone else starts the new setup tool, isn't it? Or did I missunderstand? 2015-03-25 12:22:46 alacerda: i tihnk you can go on 2015-03-25 12:22:52 nobody has time to do it properly 2015-03-25 12:23:06 and i don't know yaml enough =/ 2015-03-25 12:23:42 the idea is that you define templates in some markup language 2015-03-25 12:23:47 jaml, json xml whatever 2015-03-25 12:24:02 and write a script that simpli reads the definition from some data file 2015-03-25 12:24:16 instead of hanwrite a script for every type of setup 2015-03-25 12:24:42 what package can I use as an examplo should I learn how to do this? 2015-03-25 12:25:21 nah, i think you can simply collect the setup-gparted, setup-desktop, setup-pentest scripts to a separate git repo 2015-03-25 12:25:28 and try clean up those 3 2015-03-25 12:25:40 in the process you'll notice that they all do very similar things 2015-03-25 12:25:54 in fact, they all do same thing more or less 2015-03-25 12:26:13 and you can try figure out how to normalize the code, remove duplicate code 2015-03-25 12:26:37 in that remove dupe code you'll end up doing something that we are after 2015-03-25 12:27:41 then we have cases like ruby-rack-test and ruby-redmine-rack-test 2015-03-25 12:28:10 ok... I'll finish the MATE Desktop, Vino and XRDP documentation in the wiki. Then i'll start this 2015-03-25 12:28:12 both are at version 0.6.2, but the difference is in the dependencies 2015-03-25 12:29:09 the former depends on ruby-rack (v1.6.0) and the latter on ruby-redmine-rack (v1.4.5) 2015-03-25 12:29:55 mosez: how do you handle cases like that^^ 2015-03-25 12:31:01 name your package like ruby-rack-1_4_5-test-0_6_2 ? 2015-03-25 12:31:03 i suppse we need a ruby-rails42-rack-test-0.6.2 and ruby-rack-test-0.6.2 2015-03-25 12:32:51 hm 2015-03-25 12:33:11 we have ruby-gollum in testing 2015-03-25 12:33:19 which also is a rails app 2015-03-25 12:35:34 ugh 2015-03-25 12:35:43 gollum requires ruby 1.9 2015-03-25 12:37:22 lol 2015-03-25 12:37:33 a bit more automation would be nice 2015-03-25 12:37:43 kunkku: yes 2015-03-25 12:37:49 i think gem2apkbuild would help 2015-03-25 12:37:57 but the gem ecosystem is so messy that implementing that is not trivial 2015-03-25 12:38:22 see the rack-test example above... 2015-03-25 12:38:28 yeah 2015-03-25 12:38:39 we cannot support multiple versions of ruby itself 2015-03-25 12:38:47 i think that will be too messy 2015-03-25 12:38:55 we will have 1000+ ruby-* packages 2015-03-25 12:39:04 for every single combination 2015-03-25 12:39:40 redmine is a high-profile app 2015-03-25 12:39:44 since we use it ourselves 2015-03-25 12:39:53 so i am thinking: 2015-03-25 12:40:00 1) upgrade to ruby-2.2 2015-03-25 12:40:07 2) upgrade redmine to 3.0.x 2015-03-25 12:40:19 3) try clean up deps 2015-03-25 12:40:26 4 remove stuff thats not fixable from there 2015-03-25 12:42:34 alacerda: you just need to think about good yaml structure to have the data in sane order 2015-03-25 12:42:53 alacerda: there is plenty of ready yaml parsing libraries for sh I guess 2015-03-25 12:43:36 or it could be written in lua 2015-03-25 12:43:57 but as barthalion says, the task is to create a sane data structure 2015-03-25 12:44:22 gem2apkbuild does not suffice 2015-03-25 12:45:06 in order to achieve an acceptable automation level, the tools needs to consider all the supported gems and their inter-dependencies at the same time 2015-03-25 12:45:58 sounds non-trivial 2015-03-25 12:46:26 the tool also needs to be able to automatically fork apkbuild files as a last resort when version depedencies cannot otherwise be satisfied 2015-03-25 12:46:57 i am thinking of making it semi-automatic 2015-03-25 12:47:22 below "acceptable automation level" 2015-03-25 12:47:26 at least as a first step 2015-03-25 12:47:37 because currently, we no automation 2015-03-25 12:48:30 alacerda: but it doesn't need to be yaml, use whatever suits you; people who write the code decide :p 2015-03-25 12:48:42 barthalion: +1 2015-03-25 12:48:59 well, with my resolver script you can check if upgrading a set of gem apks to versions will break anything 2015-03-25 12:49:51 i think of use that tool + some semi automatic 2015-03-25 12:50:17 things like the rack-test example will probably need manual decisions 2015-03-25 12:50:40 such as: "you know what... lets remove ruby-something-blah...." 2015-03-25 12:51:38 I have developed somewhat working methodology to upgrade stuff with the help of the resolver script 2015-03-25 12:51:51 that should probably go into the script 2015-03-25 12:52:32 then some additional tools for easily forking or creating new gem apks 2015-03-25 12:53:06 and some tool to easily create new gem from scratch 2015-03-25 12:53:24 as you said 2015-03-25 12:54:38 I have nothing against if mosez wanted to improve the gem packaging system :) 2015-03-25 12:56:31 kunkku: i'd also like a tool that checks version in APKBUILD and compare with latest available gem version 2015-03-25 12:56:53 something that could spit out a list for me: those gems could potensially be upgraded 2015-03-25 12:57:09 that's in the script already but you need to call it manually per each apk 2015-03-25 12:57:21 cool 2015-03-25 12:57:27 how do i call it for that? 2015-03-25 12:57:38 i run it in the correct dir? 2015-03-25 12:57:59 let me check. I haven't used it for a while... 2015-03-25 13:00:17 re: "tool that checks version in APKBUILD" , we have not pkgs.a.o, mayb it provides yaml,json 2015-03-25 13:00:31 oh, that's in the script already ;) 2015-03-25 13:00:50 we have now* pkgs.a.o 2015-03-25 13:03:23 ncopa: you need to be in the aports root dir 2015-03-25 13:05:42 apkbuild-gem-resolver {-}... 2015-03-25 13:06:54 example: apkbuild-gem-resolver ruby-rails-minitest ruby-rails 2015-03-25 13:07:59 hmpf 2015-03-25 13:08:15 ruby-augeas appears to be broken with ruby-2.2 2015-03-25 13:08:41 http://sprunge.us/YJaU 2015-03-25 13:09:03 its ruby-2.2.1 + a rebuilt ruby-augeas 2015-03-25 13:09:26 does it work on 2.1? 2015-03-25 13:09:44 tit didnt fail 30 mins ago 2015-03-25 13:09:47 so i suppose so 2015-03-25 13:11:26 oh... 2015-03-25 13:11:54 you need to add the version information of the bundled gems to the script 2015-03-25 13:12:05 the array on line 13 2015-03-25 13:12:57 ok 2015-03-25 13:13:15 this is btw one factor of extra complexity: some gems come with the ruby distro 2015-03-25 13:13:33 yeah... 2015-03-25 13:13:41 often with an outdated version incompatible with other gems 2015-03-25 13:13:42 doesnt make things simpler 2015-03-25 13:13:57 and you cannot programmatically extract this information anywhere 2015-03-25 13:14:07 ncopa-desktop:~/aports$ ruby ~/abuild/apkbuild-gem-resolver.in 2015-03-25 13:14:07 ncopa-desktop:~/aports$ 2015-03-25 13:14:16 it works with ruby-2.1 2015-03-25 13:14:32 but apparently there is a bunder 1.9 out 2015-03-25 13:15:00 which ofcourse would break something 2015-03-25 13:15:27 but it would be nice to see those gems that have new version but are not possible to upgrade 2015-03-25 13:16:09 ncopa: just dont build this kind of packages like redmine-gemname. build the option to version suffix the gem packages instead. 2015-03-25 13:16:09 $ ruby ~/abuild/apkbuild-gem-resolver.in ruby-bundler 2015-03-25 13:16:09 ruby-bundler-1.9.1 2015-03-25 13:16:19 ^^^ means upgrading should be ok 2015-03-25 13:16:29 aha 2015-03-25 13:16:30 cool 2015-03-25 13:16:35 perfect! 2015-03-25 13:17:24 now you need to manually edit the version in apkbuild 2015-03-25 13:17:44 abump ruby-bundler-1.9.1 2015-03-25 13:17:51 but it wouldn't be difficult to add a switch like --update 2015-03-25 13:18:04 and open buildservice is like build.opensuse.org. this software is oss :) 2015-03-25 13:18:30 the difficulties come when something would fail and you need to start figuring out the corrective actions 2015-03-25 13:19:02 detecting missing dependencies is really a pain in the ass as well :) 2015-03-25 13:19:11 or breaking dependencies :) 2015-03-25 13:19:21 mosez: you work at SUSE? 2015-03-25 13:19:26 barthalion: yes 2015-03-25 13:19:44 suse cloud 8) 2015-03-25 13:19:55 the resolver script compares the current apkbuild dependencies vs. gem metadata of the suggested (or latest) version, and complains if necessary 2015-03-25 13:20:02 mosez: openstack brofist 2015-03-25 13:20:21 barthalion: i'm working for the cloud team but on the installer (ruby on rails) :D 2015-03-25 13:20:57 I feel your pain :D 2015-03-25 13:21:13 crowbar is currently the most flexible solution to install a openstack cloud 2015-03-25 13:21:50 ACTION is waiting for TripleO 2015-03-25 13:21:53 barthalion, ok... I'll decide but probably i'll listen to your advices and ncopa's 2015-03-25 13:21:53 =] 2015-03-25 13:22:20 tripleO will die :) 2015-03-25 13:22:53 hp stoped working on tripleo... and they have been the biggest contributors. everybody is doing his own stuff 2015-03-25 13:23:18 crowbar doesn't seem to be active either 2015-03-25 13:23:40 really? When was that? 2015-03-25 13:24:21 barthalion: it will be more active when i'm done with the opensuse integration. than it will be easier for others to activate ubuntu or whatever else :) 2015-03-25 13:24:26 kunkku: a few months ago 2015-03-25 13:25:05 barthalion: and crowbar is quite active, of course not that many contributors as for openstack but ~30 people working actively on it 2015-03-25 13:25:17 why not stackforge, btw? 2015-03-25 13:25:34 or am I looking at wrong repository? 2015-03-25 13:26:14 https://www.openhub.net/p/crowbar-v1 2015-03-25 13:26:36 ah, ok, I was looking at opencrowbar 2015-03-25 13:26:49 stackforge is using puppet. crowbar is built around chef :) 2015-03-25 13:26:54 kunkku: $ ruby ~/abuild/apkbuild-gem-resolver.in ruby-bundler 2015-03-25 13:27:00 not very fast... 2015-03-25 13:27:02 stackforge is not about puppet or chef 2015-03-25 13:27:05 or any other tool 2015-03-25 13:27:10 opencrowbar is the fork where some people are working on that have been employed by dell 2015-03-25 13:27:13 maybe i should add another 24 cpu cores at the problem? :D 2015-03-25 13:27:40 ah, right... we are using some stackforge cookbooks :D 2015-03-25 13:27:48 but they are not that feature complete 8) 2015-03-25 13:27:59 kaniini: i'm still happy that it is faster than manually searching all the deps... 2015-03-25 13:28:11 i mean kunkku, not kaniini 2015-03-25 13:28:26 ncopa: version suffixes for the gem packages is not an option? 2015-03-25 13:28:41 version suffix in the package name? 2015-03-25 13:28:44 yes 2015-03-25 13:28:51 thats an option 2015-03-25 13:28:54 ncopa: yes, that's what i mean 2015-03-25 13:29:13 like ruby-rails-4_2 refers to rails 4.2.x 2015-03-25 13:29:34 but i dont want things like: ruby-rack-1_4_5-test-0_6_2 2015-03-25 13:29:35 ncopa: the gem metadata will get cached so its faster next time (though not fast enough) 2015-03-25 13:29:44 chef is rather disappoiting choice, I even like puppet more, but I guess it's flamewar territory 2015-03-25 13:29:45 with that you can build nearly every scenario :) 2015-03-25 13:30:01 why ruby-rack-1_4_5-test-0_6_2? 2015-03-25 13:30:48 ruby-rack-test-0_6, but how do you get ruby-rack-1_4_5-test-0_6_2? 2015-03-25 13:30:58 then we have cases like ruby-rack-test and ruby-redmine-rack-test 2015-03-25 13:30:58 both are at version 0.6.2, but the difference is in the dependencies 2015-03-25 13:30:58 the former depends on ruby-rack (v1.6.0) and the latter on ruby-redmine-rack (v1.4.5) 2015-03-25 13:31:29 i think ruby-redmine-rack is wrong 2015-03-25 13:31:44 ruby-redmine-rack is an alpine only package 8) 2015-03-25 13:31:53 i inted to fix it 2015-03-25 13:31:55 in fact it should be a ruby-rack 2015-03-25 13:32:02 exactly 2015-03-25 13:32:20 and with the suffix you can remove the project specific fake packages :) 2015-03-25 13:32:35 it does not remove the core problem 2015-03-25 13:32:46 kunkku: which core problem? 2015-03-25 13:33:18 than it's still not automated but you can install different versions of a single gem which solve the biggest core problem :) 2015-03-25 13:34:01 i'd like similar to what suse does, and have rails-* packages 2015-03-25 13:34:13 projects like redmine normally depend on a rails version 2015-03-25 13:34:34 i think that will solve the majority 2015-03-25 13:34:35 now the ruby-rack-test package is forked because there are multiple versions of the dependency (ruby-rack) 2015-03-25 13:35:07 kunkku: where is it forked? you mean in the package repo? 2015-03-25 13:35:23 ruby-rack-test 0.6.2 will work with rack 1.4.5 and rack 1.6.0 2015-03-25 13:35:47 but still there are two versions of ruby-rack-test 2015-03-25 13:36:08 https://rubygems.org/gems/rack-test rack-test works with rack >= 1.0 2015-03-25 13:36:18 yes 2015-03-25 13:36:35 just a basic apk question 2015-03-25 13:36:57 is there something like "provides" in apk like on rpm? 2015-03-25 13:37:03 yes 2015-03-25 13:37:19 hum 2015-03-25 13:37:20 yes, that's a way around the problem 2015-03-25 13:37:34 so all ruby-rack-* packages provide ruby-rack and apk will try to install the latest version, right? 2015-03-25 13:37:35 but I would not like to manage those manually... 2015-03-25 13:37:44 if the dependencies are resolvable 2015-03-25 13:37:50 mosez: correct 2015-03-25 13:38:35 kunkku: what do you mean? building the suffixed packages? 2015-03-25 13:39:10 btw, ruby-rack-test is not needed by anything 2015-03-25 13:39:15 http://pkgs.alpinelinux.org/package/x86/ruby-rack-test 2015-03-25 13:39:48 hum 2015-03-25 13:39:53 i wonder if thats broken 2015-03-25 13:40:13 ruby-redmine-rack-test has also Required by(0) 2015-03-25 13:40:46 maybe just build dependency? 2015-03-25 13:41:59 mosez: you would need to invent a new 'provides' tag every time you encounter this situation 2015-03-25 13:42:31 kunkku: why? it always provides the gem name? 2015-03-25 13:43:00 no 2015-03-25 13:43:07 i dont get it 2015-03-25 13:43:43 https://gist.github.com/tboerger/6c77a3b62012f2040e7f these are the automatically built provides of rubygem rpms on suse 2015-03-25 13:44:11 the tag represents not the gem itself but a set of gem versions compatible with whoever asks for the tag 2015-03-25 13:44:29 kunkku: than you are talking about a different thing :) 2015-03-25 13:45:02 ncopa: can you put something like "depends=foo>=1.0<2.0" in apkbuild? 2015-03-25 13:45:58 depends="foo>=1.0<2.0" hum 2015-03-25 13:46:09 i know that depends="foo>=1.0" works 2015-03-25 13:46:28 and depends="foo<2.0" 2015-03-25 13:46:32 the gem deps are almost always like ranges 2015-03-25 13:46:34 or maybe constraints like on ruby? that would be much easier :D 2015-03-25 13:46:57 thats a question for fabled 2015-03-25 13:46:59 depends="foo~>2.0" 2015-03-25 13:47:22 that means foo 2.0 and foo 2.0.x right? 2015-03-25 13:47:30 2.x 2015-03-25 13:48:11 depends="foo~>2.0.0" would mean 2.0.x, depends="foo~>2.0" means 2.x 2015-03-25 13:48:19 some gem dependencies are like "greater than 2.4 but less than 3.0 but still not 2.6.4 etc." 2015-03-25 13:48:21 i dont think apk currently does depends="foo>=1.0<2.0" 2015-03-25 13:49:00 does foo>=1.0 foo<2.0 work? 2015-03-25 13:49:02 and what about depends="foo>=1.0 foo<2.0"? 2015-03-25 13:49:06 xD 2015-03-25 13:49:09 however, i think you can do: depends="foo>=1.0 !foo>=2.0" 2015-03-25 13:49:25 true 2015-03-25 13:49:36 foo-2.1 would satisfy foo>=1.0 2015-03-25 13:49:36 if that gets combined correctly? 2015-03-25 13:49:37 if apk does not support that, you need to but version numbers in the provides tags 2015-03-25 13:49:41 does someone know how hard it would be to integrate a "yum whatprovides" function to apk ? 2015-03-25 13:49:49 ^^^ which I don't want to manage manually 2015-03-25 13:50:03 i guess APKINDEX would become at least 10x larger 2015-03-25 13:50:36 seems building the APKINDEX with pkg file list should be pretty trivial 2015-03-25 13:50:37 coredumb: might use the new pkgs.alpinelinux.org 2015-03-25 13:50:48 what about in apk 2015-03-25 13:50:54 coredumb: this is something I started to work on as a separate tool 2015-03-25 13:51:04 I don't think this belongs to apk, to be honest 2015-03-25 13:51:13 as i said, these are the provides of a single rubygem package on suse/rpm https://gist.github.com/tboerger/6c77a3b62012f2040e7f 2015-03-25 13:51:13 we dont want index all files by default 2015-03-25 13:51:15 barthalion: why not ? 2015-03-25 13:51:19 as you said, it blows up APKINDEXes size 2015-03-25 13:51:26 ncopa: i can understand that 2015-03-25 13:51:46 http://wiki.alpinelinux.org/wiki/MATE 2015-03-25 13:51:55 done 2015-03-25 13:52:03 but again, I have no time 2015-03-25 13:52:14 ncopa: pkgs.alpinelinux.org doesn't show up files contained in pkg 2015-03-25 13:52:28 barthalion: what's your idea ? 2015-03-25 13:52:43 http://pkgs.alpinelinux.org/contents?filename=ruby&pkgname=&arch=x86 2015-03-25 13:52:56 basically waht pkgfile does 2015-03-25 13:53:07 but I see that we have it on pkgs.alpinelinux.org already 2015-03-25 13:53:17 which is really cool, cheers clandmeter 2015-03-25 13:53:29 and what about a tool that communicates with http://pkgs.alpinelinux.org on the command line? :) 2015-03-25 13:53:39 thats next step 2015-03-25 13:53:46 did i mention we have a new package db? :) 2015-03-25 13:53:47 good one 8) 2015-03-25 13:53:53 we need provide an api 2015-03-25 13:53:59 and then we're done 2015-03-25 13:54:01 seems to me like a really really needed feature 2015-03-25 13:54:03 :) 2015-03-25 13:54:05 yup 2015-03-25 13:54:20 and clandmeter got the ball rolling 2015-03-25 13:54:27 pkgs.alpinelinux.org is written with which languages/framework? :) 2015-03-25 13:54:36 turbolua 2015-03-25 13:54:37 turbolua 2015-03-25 13:54:44 eh... ok :) 2015-03-25 13:54:47 http://git.alpinelinux.org/cgit/clandmeter/aports-turbo/ 2015-03-25 13:54:56 or lua-turbo in aports 2015-03-25 13:55:07 it could be optimized 2015-03-25 13:55:35 i'd like to normalize it - avoid data duplication (copy to sql db) 2015-03-25 13:55:41 but its a start at leasts 2015-03-25 13:55:44 lua for web development... that's new to me :D 2015-03-25 13:55:46 and its a good thing 2015-03-25 13:56:07 we kinda like lua :) 2015-03-25 13:56:09 its small 2015-03-25 13:56:11 its fast 2015-03-25 13:56:13 oh would be nice to have partial search on content :) 2015-03-25 13:56:23 its not like ruby 2015-03-25 13:56:25 :) 2015-03-25 13:56:37 like %asterisk/version.h% 2015-03-25 13:56:39 coredumb: use %a% 2015-03-25 13:56:48 clandmeter: doesn't work in content 2015-03-25 13:56:49 lua is cool... i have done lua coding in computercraft on minecraft :D 2015-03-25 13:56:57 but i prefer ruby 8) 2015-03-25 13:57:00 first thing i tried ;) 2015-03-25 13:57:27 coredumb: i think you only can search on last element in path 2015-03-25 13:57:36 for now... 2015-03-25 13:58:02 clandmeter: ncopa mmmh apparently results tend to vary 2015-03-25 13:58:13 something's fishy 2015-03-25 13:58:19 coredumb: yes you can only search on files 2015-03-25 13:58:35 ncopa: 2015-03-25 13:59:01 is alpine backed by a company? or made by volunteers only? :) 2015-03-25 13:59:22 I don't know what this gir-scanner is at all but it seems that some other tool is parsing those header files (giscan in this case?) 2015-03-25 13:59:30 well no it's me the fishy result always the same :D 2015-03-25 13:59:45 clandmeter: please make it possible on the rest ^^ 2015-03-25 14:00:12 stateless: i suspect it parses the header files yes 2015-03-25 14:00:34 coredumb: i never had a need for it, so i didnt add it. :) 2015-03-25 14:00:52 it might make sense to start using 'provides' on the ruby packages 2015-03-25 14:00:53 clandmeter: search version.h 2015-03-25 14:01:02 then you have too much results 2015-03-25 14:01:25 kunkku: i think so too 2015-03-25 14:01:44 ncopa: it doesn't seem to understand these gcc extensions, from what I've seen from a quick google search there is a way to ignore parse failures by using regexes 2015-03-25 14:01:50 shouldnt be too hard make abuild autoadd a provides 2015-03-25 14:02:09 kunkku, ncopa:and how should they look like? 2015-03-25 14:02:27 coredumb: ill look into it. first need to cleanup parts of our infra, and update our redmine. 2015-03-25 14:02:39 I'm not sure, but I'm nor concerned about apk capabilities 2015-03-25 14:02:42 in worst case you can generate the provides with gem2apkbuild?! 2015-03-25 14:02:47 clandmeter: sure :) 2015-03-25 14:03:12 and i dont understand that you are still running on redmine :P 2015-03-25 14:03:12 can I use the apk dependency resolver with plain apkbuild files? 2015-03-25 14:03:31 i think not 2015-03-25 14:03:44 I would not like to abuild ruby-* just to be able to verify deps 2015-03-25 14:04:20 then we need generate the provides and stuff it into the apkbuild 2015-03-25 14:04:37 as i said... with gem2apkbuild :) 2015-03-25 14:04:42 yup 2015-03-25 14:04:48 i just echo :) 2015-03-25 14:04:50 no 2015-03-25 14:05:11 it would bloat the apkbuild a little bit... but you can see directly everything without big magic 2015-03-25 14:05:19 we need a means to verify that nothing breaks when an individual gem i s updated 2015-03-25 14:05:54 :D 2015-03-25 14:06:02 with gem2apkbuild only, you discover that when it's too late 2015-03-25 14:06:05 do you have some kind of infrastructure to test things like that automated in the background? 2015-03-25 14:06:11 i am updating some stuff and i am 100% sure i break things :) 2015-03-25 14:06:18 ruby-2.2 upgrade breaks it all 2015-03-25 14:06:33 so i need rebuild it all 2015-03-25 14:06:54 I don't think so 2015-03-25 14:07:55 the problem is that you might put a non-existent "provides" tag the "depends" list 2015-03-25 14:08:23 kunkku: we do post-checks with shared objects 2015-03-25 14:08:25 .so files 2015-03-25 14:08:36 we dont know if an upgrade break anything til after its built 2015-03-25 14:08:49 then i run a checkapk that will diff the package 2015-03-25 14:09:08 and tell me that it changed SONAME 2015-03-25 14:09:21 then i will ahve to manually search for packages needing that SONAME 2015-03-25 14:09:23 and rebuild those 2015-03-25 14:09:56 i am kinda ok that we test ruby packages after they are built 2015-03-25 14:10:51 is there anything where i can give some hands on? if not i'm out of the discussion and get back to work 8) 2015-03-25 14:11:24 hwo do you guys earn your money? alpine is only volunteer stuff or not? :D 2015-03-25 14:11:29 wanna take over the maintenance of ruby-*? :) 2015-03-25 14:12:44 yes please take over, and we split our earnings :D 2015-03-25 14:13:51 haha 2015-03-25 14:14:14 mosez: thanks alot for your input! 2015-03-25 14:14:41 ncopa: you are welcome. if i can help on somthing beside my job ping me :) 2015-03-25 14:14:50 i really love alpine for my docker images :) 2015-03-25 14:15:06 mosez: afaik there is no commercial support for alpine 2015-03-25 14:15:10 my next step will be an elasticsearch package 8) 2015-03-25 14:15:41 what pretty much means that we are not backed per se :p 2015-03-25 14:16:21 no one to sue if your systems fail :) 2015-03-25 14:16:21 so there is no company behind alpine? so found one and pay me xD 2015-03-25 14:16:57 it's always algitbot fail 2015-03-25 14:17:03 fault* 2015-03-25 14:17:11 :D 2015-03-25 14:18:33 algitbot: why do you always fail? 2015-03-25 14:18:41 hehe 2015-03-25 14:23:49 why not introduce rubyprovides or rprovides 2015-03-25 14:27:22 maybe rdepends also 2015-03-25 14:28:33 which is correct in commit message: rebuild for/with/against ruby 2.2 2015-03-25 14:29:36 against? 2015-03-25 14:29:47 at least this is what I use 2015-03-25 14:30:19 +1 against 2015-03-25 14:36:44 https://github.com/scadu/aports bumped libmicrohttpd, youtube-dl and goaccess. 2015-03-25 14:37:04 maybe hostapd in a while. 2015-03-25 14:39:04 https://paste.xinu.at/eD1sEZ/plain or maybe not :f 2015-03-25 14:39:11 merged, thanks 2015-03-25 14:39:18 thanks 2015-03-25 14:43:56 clandmeter: ruby-redmine-unicorn-4.8.3-r0.apk vs ruby-unicorn-4.8.3-r0.apk 2015-03-25 14:44:22 did you check the deps or did you just ruby-redmine-* it all 2015-03-25 14:45:19 :D 2015-03-25 14:45:39 i intend to rm ruby-redmine-* and create a ruby-rails4.2-* prefix 2015-03-25 14:46:28 so i wonder if rails depends on a specific version of unicorn 2015-03-25 14:46:37 or redmine depends on specific version of unicorn 2015-03-25 14:47:10 does it make any sense: http://hastebin.com/ejoyoxuyer.avrasm ? 2015-03-25 14:47:28 ncopa: :) 2015-03-25 14:47:44 alacerda: something needs a rebuild 2015-03-25 14:47:55 i dont remember why i did that. probably brain farthed after so many depcrap 2015-03-25 14:48:20 thats what i suspect 2015-03-25 14:48:26 i bumped into it also yesterday 2015-03-25 14:48:50 i still got a gazzilion changes in my tree because of redmine. 2015-03-25 14:48:50 i am upgrading ruby 2.2 2015-03-25 14:49:12 and intend to do redmine 3.0 2015-03-25 14:49:24 and clean up a bit while at it 2015-03-25 14:49:32 ncopa: dont do that :( 2015-03-25 14:49:37 aw! 2015-03-25 14:49:39 now what? 2015-03-25 14:49:48 ruby-rails4.2-* feels totally wrong 2015-03-25 14:49:58 i was about to say: i got the needed moral support.. 2015-03-25 14:50:07 ruby-rails42-*? 2015-03-25 14:50:24 why ruby-railsxx-? 2015-03-25 14:50:41 in case other app needs rails 2015-03-25 14:50:52 just ruby-rails-4.2 or ruby-rails-4_2, ruby-activesupport-4_2 and so on 2015-03-25 14:51:11 the other gems are not deps of a specific rails version 2015-03-25 14:51:23 4_2 everywhere 2015-03-25 14:51:51 do you wanna do ruby-rails4.2-activesupport? or what? :) 2015-03-25 14:51:59 ok im going to make changes to our vlan's, so if all goes down, it wasnt me, it was ncopa upgrading ruby 2015-03-25 14:52:00 so ruby-activesupport4.2-4.2.1 2015-03-25 14:52:07 we do lua5.1 and lua5.2 2015-03-25 14:52:15 clandmeter: lol 2015-03-25 14:52:18 ie a dot instead of _ 2015-03-25 14:52:34 so ruby-activesupport4.2 2015-03-25 14:52:38 clandmeter: no!! its algitbots fault! 2015-03-25 14:53:04 algitbot: what are you doin? 2015-03-25 14:53:22 folks... I gotta go... see you tm afternoon 2015-03-25 14:53:30 see u 2015-03-25 14:53:55 so ruby-activesupport4.2 2015-03-25 14:53:57 makes sense 2015-03-25 14:56:41 felt kinda good: https://dpaste.de/DqXi 2015-03-25 14:56:52 now lets see if i can make redmine work again :) 2015-03-25 14:57:10 why it feels so good to destroy my life's work? 2015-03-25 14:57:11 good one :) 2015-03-25 14:57:15 haha 2015-03-25 14:57:25 burn it! 2015-03-25 14:57:30 sorry clandmeter 2015-03-25 14:57:31 <__mjones> @ncopa: I was looking at building mongodb-3.0.1. from an updated version of the unmaintained APKBUILD. Any particular reason that one is unmaintained? 2015-03-25 14:57:52 does it build? 2015-03-25 14:57:55 clandmeter: how do we always ennd up like that? you build it up i tear it down... 2015-03-25 14:58:22 __mjones: it probably does not build an d nobody cared enough to fix it 2015-03-25 14:58:25 ncopa: its a dirty job... 2015-03-25 14:59:41 <__mjones> ncopa: I wondered if that was why. It definitely does not build, and the scons build system is revised. I've put some effort into it but I'm looking to finish something else before I go back and try to finish. 2015-03-25 14:59:57 __mjones: i think i tried to build it when we switched to musl, but failed horrible. 2015-03-25 15:00:03 it applies to all unmaintained packages 2015-03-25 15:00:08 yeah 2015-03-25 15:00:11 and I believe mongodb is terribly unportable 2015-03-25 15:00:26 moved to unmaintained means someone gave up 2015-03-25 15:00:45 and that it does not build 2015-03-25 15:00:59 its even worse then ruby's dep hell 2015-03-25 15:01:03 <__mjones> It builds for a long time without any patches, and using the included third-party libs, before it falls over some GNU headers. :-/ 2015-03-25 15:02:29 <__mjones> And if I can ask another, is there a policy on the non-free tree? I did an update of intel-ucode, but new kernels now need a separate initrd to load microcode so it turns out it needs to be redone anyway. 2015-03-25 15:02:44 <__mjones> I just didn't know if there was any policy about updates to non-free. 2015-03-25 15:05:26 we will merge them 2015-03-25 15:05:35 these packages aren't distributed anyway 2015-03-25 15:06:15 yeah 2015-03-25 15:06:25 somebody got an example how the dependencies for a javaproject should look like? 2015-03-25 15:06:47 we barely have openjdk 2015-03-25 15:06:56 not many java projects 2015-03-25 15:06:58 if any 2015-03-25 15:07:38 uh... and do i need to build it from source? 2015-03-25 15:07:48 openjdk? 2015-03-25 15:07:48 no 2015-03-25 15:07:52 no 2015-03-25 15:07:55 https://www.elastic.co/downloads/elasticsearch 2015-03-25 15:08:10 there are sources available but than i need to build everything with maven -.- 2015-03-25 15:08:30 i think you dont need build from sources 2015-03-25 15:08:41 i got fop running locally 2015-03-25 15:08:45 waste of time to use maven 2015-03-25 15:08:51 you will kill yourself due to that 2015-03-25 15:08:53 using the precompiled java binary 2015-03-25 15:09:03 ok, so i will use this prebuilt stuff and just remove the windows and mac stuff 2015-03-25 15:09:08 yeah 2015-03-25 15:09:19 cool :) 2015-03-25 15:10:08 kunkku: ruby-rails-minitest, why the rails in there while the gem name is just "minitest" 2015-03-25 15:10:16 why not ruby-minitest? 2015-03-25 15:10:26 um 2015-03-25 15:10:32 i know the answer already 2015-03-25 15:10:37 ruby bundles it 2015-03-25 15:11:26 :) 2015-03-25 15:12:55 where should i put the lib folder with all the required jars? 2015-03-25 15:13:54 have no idea to be honest :) 2015-03-25 15:14:31 ruby 2.2.1 comes with minitest 5.4.3 2015-03-25 15:15:01 but activesupport needs minitest ~> 5.1 2015-03-25 15:15:10 /usr/lib/elasticsearch maybe? 2015-03-25 15:15:23 mosez sounds good to me 2015-03-25 15:15:30 will java find it there? 2015-03-25 15:15:45 /var/lib/elasticsearch should be the index data... like the database data of mariadb or postgres 2015-03-25 15:16:29 %ES_HOME%/lib/elasticsearch-1.5.0.jar :) 2015-03-25 15:16:51 i need to define ES_HOME with that path 2015-03-25 15:17:17 hum? 2015-03-25 15:17:19 just sed it 2015-03-25 15:17:59 yeah, that's the other option. sed the path fixed within the 2 scripts that use that variable 2015-03-25 15:20:04 damn... i really need an ES_HOME 2015-03-25 15:20:18 beside the path to the libs 2015-03-25 15:20:41 ah... home will be /var/lib/elasticsearch and libraries will be /usr/lib/elasticsearch :) 2015-03-25 15:23:53 would it make sense to simply ship custom versions of the elasticsearch and plugin script? that way i can remove all mongw checks as well 2015-03-25 15:24:30 elasticsearch is 250 lines bash including comments. plugin got 48 lines... 2015-03-25 15:25:15 will minitest-5.4.3 satisfy the dependency: minitest ~> 5.1 ? 2015-03-25 15:25:39 yes 2015-03-25 15:26:00 minimized plugin version is 28 lines :D 2015-03-25 15:26:39 maybe as a patch? than we can see that something in a new version have been changed 2015-03-25 15:27:29 kunkku: i think I'll remove ruby-rails-minitest since ruby-2.2 provided minitest should work 2015-03-25 15:28:29 and elasticsearch will heave 200 lines 2015-03-25 15:28:32 i will patch it :) 2015-03-25 15:28:34 sure. I added it because this was not the case with 2.1 2015-03-25 15:50:44 https://gist.github.com/tboerger/4b2614064cacbc223989 a tiny patch i would apply to elasticsearch 8) 2015-03-25 15:53:38 any news about alpine on raspberry pi 2 ? 2015-03-25 15:56:25 hum 2015-03-25 15:56:38 i wonder if we shoudl start use provides for rails 2015-03-25 15:57:10 that is, ruby-activesupport4.2 package provides ruby-activesupport 2015-03-25 16:01:11 ncopa: that's what i said before xD 2015-03-25 16:01:29 yes 2015-03-25 16:01:42 i am wondering if we should follow that advice 2015-03-25 16:01:48 my question is 2015-03-25 16:02:10 should ruby-activeview4.2 depend on ruby-activesupport4.2 2015-03-25 16:02:26 yes, because it's a hard dependency 2015-03-25 16:02:44 = 4.2 and not ~> 4.2 2015-03-25 16:03:10 or should it depend on ruby-activesupport and ruby-activesupport4.2 provide ruby-activesupport? 2015-03-25 16:03:22 ok 2015-03-25 16:03:28 i suppose the former 2015-03-25 16:04:03 According to gemspec it should be the same 2015-03-25 16:05:13 They have one VERSION file for all rails projects 2015-03-25 16:07:43 mosez: how to handle redmine itself? 2015-03-25 16:08:33 clandmeter: with gloves 2015-03-25 16:08:38 gotta go see u 2015-03-25 16:08:54 why do we use redmine, btw? 2015-03-25 16:09:13 which alternative we have? 2015-03-25 16:09:26 anything easier to package? 2015-03-25 16:09:35 which is? 2015-03-25 16:09:43 do we rely on something particular in redmine? 2015-03-25 16:09:52 bugzilla is mprobably the only candidate 2015-03-25 16:10:27 because? 2015-03-25 16:10:44 <__mjones> probably the best candidate, true. Redmine is supposed to be great, though. 2015-03-25 16:10:47 there is mantis for example 2015-03-25 16:13:38 What is wrong with redmine? 2015-03-25 16:14:04 <__mjones> Sounds like problematic dependencies. 2015-03-25 16:14:45 ncopa has some issues with redmine 2015-03-25 16:14:59 tagging to multipel releases i think 2015-03-25 16:15:17 I can help 2015-03-25 16:15:57 every one that i know using redmine have issues with it ^^ 2015-03-25 16:16:36 dinner time. nice evening. 2015-03-25 16:17:03 I suspect that all this problems comes from OS package managers 2015-03-25 16:19:52 <__mjones> ptico: funny, I often find all those problems come from language package managers. ;) 2015-03-25 16:21:12 __mjones: have no idea why. Rubygems is very simple tool and most of projects uses bundler which removes all the dependency pain 2015-03-25 16:22:20 Most of rails projects can be delivered with simple one-liner 2015-03-25 16:24:16 <__mjones> ptico: as a mostly-ops person, I only mix dist and lang repos under existential threat. 2015-03-25 16:24:49 <__mjones> ptico: and since almost all dists have their own package managers, that means we just say no to lang repos. 2015-03-25 16:26:02 <__mjones> ptico: there are exceptions and workarounds, but I'm very disinclined to go for those merely because language people insist on ubiquitously pointing devs at lang repos, and deprecating or denigrating dist repos. 2015-03-25 16:32:12 __mjones: it differs 2015-03-25 16:36:18 But generally i had problems only with dist repos instead. Especially when waiting for security releases of rubygems can take about a month 2015-03-25 16:36:39 I mean ruby gems 2015-03-25 16:42:56 <__mjones> ptico: one definitely wants the repo to be fast to patch, and to contain those components that one needs. 2015-03-25 16:43:19 <__mjones> Does the ruby repo allow picking specific releases, or only the current release? 2015-03-25 16:45:58 __mjones: anything you want, even lock specific tag or commit in git 2015-03-25 16:52:23 __mjones: that's why i think repackaging of ruby gems is a bad idea. Redmine for example, could use old rails and other deps, Chef instead could use the same gems, but newer versions. So you will be able to package only 2 years old chef distribution or remove redmine at all. 2015-03-25 16:54:19 <__mjones> ptico: this is true of any language with a lot of optional components. It's especially acute with interpreted, dynamic languge proliferation. 2015-03-25 16:57:45 Also this is just wasting maintainer's time 2015-03-25 16:58:12 <__mjones> ptico: I seek to minimise the proliferation of runtimes, and to minimise the proliferation of optional or third-party additions to those runtimes. 2015-03-25 16:58:50 <__mjones> If you have a system that builds images from packages, it's necessary and useful time. 2015-03-25 16:59:07 <__mjones> quit proliferating modules. 2015-03-25 17:01:21 <__mjones> viz.: http://libguestfs.org/virt-builder.1.html 2015-03-25 17:01:22 What about to bundle libs inside the package? 2015-03-25 17:01:44 <__mjones> Sometimes they need to be in a separate package, sometimes not. 2015-03-25 17:02:17 <__mjones> But when you have more than 5 or 6 packages for a system/application-core/runtime, that 2015-03-25 17:02:24 <__mjones> that's a sign of a problem. 2015-03-25 17:04:01 <__mjones> # apk search ruby- | wc -l 2015-03-25 17:04:02 This is very specific question, in ruby (and i believe in js) word, it is ok to extract useful code to small, one-file libs 2015-03-25 17:04:02 <__mjones> 132 2015-03-25 17:04:21 <__mjones> # apk search ruby-redmine | wc -l 2015-03-25 17:04:28 rubygems.org -> 6219 gems 2015-03-25 17:04:28 <__mjones> 20 2015-03-25 17:04:41 <__mjones> ruby-redmine-tzinfo-0.3.42-r0 2015-03-25 17:04:58 <__mjones> ruby-redmine-tzinfo? Systems already have tzdata for a reason, damnit. 2015-03-25 17:05:21 <__mjones> 132 is far more than enough. 6219 is right out. 2015-03-25 17:05:48 <__mjones> Coders need to use a subset. 6219 are not going to be supported indefinitely. 2015-03-25 17:06:31 <__mjones> But to avoid declaring a strongly-supported subset and hurting someone's feelings, or making any commitment, the answer is always: 2015-03-25 17:06:38 <__mjones> "use our language repo." 2015-03-25 17:06:49 Most scripting langs supports windows too :(. I'm not sure how to retreive tzdata on windows 2015-03-25 17:06:58 <__mjones> Use system calls. 2015-03-25 17:07:14 <__mjones> Nobody should be doing their own date functions unless they're looking to create wrong code. Full stop. 2015-03-25 17:08:02 <__mjones> I don't mean system calls literally; I mean date functions native to the platform if not already abstracted by the language. 2015-03-25 17:08:11 Nice idea btw, will add proposal to ruby core 2015-03-25 17:08:17 <__mjones> time is easy to get right: let system libs handle it. 2015-03-25 17:08:40 <__mjones> You're saying nobody thought of that before? That's hard to imagine. 2015-03-25 17:09:43 Imagine you develop your own language. There are many scopes that should be covered 2015-03-25 17:10:10 <__mjones> I wouldn't in a million years develop my own language. 2015-03-25 17:10:16 And if you support microsoft shit - it can hurt 2015-03-25 17:10:40 <__mjones> I'm a rather bad programmer. But being a mostly-ops person who is a bad programmer gives me a rather different perspective, I find. 2015-03-25 17:11:07 Word is not perfect) 2015-03-25 17:11:15 <__mjones> I lean heavily on automation to help me produce adequate code, or to turn adequate code into polished code. 2015-03-25 17:11:56 <__mjones> I know a numbe rof things not to do, that dedicated devs often don't know. One of those things is never do your own time/date/calendar stuff. 2015-03-25 17:12:26 Why linux OSes have so many init systems or package managers instead of using one? 2015-03-25 17:12:29 <__mjones> Know when to outsource. ;) 2015-03-25 17:12:49 <__mjones> Because package managers are often sources of pain, standardisation, and politics. 2015-03-25 17:13:05 <__mjones> Init systems were until a few years ago always SysV init on Linux. That only changed recently. 2015-03-25 17:13:16 Lang guys have the same problems 2015-03-25 17:13:54 <__mjones> I say for the long term service developers can produce a systemd init script/config, one traditional SysV init script, and let everyone else make their own. 2015-03-25 17:14:31 <__mjones> In the meantime, an upstart script would still be nice until Ubuntu moves everything over including LTS. launchd for the mac folks is nice if you don't mind. 2015-03-25 17:14:32 stdlib for example can have minimal implementation, so one guy creates more reach api, then another guy fork it to add some functionality which first guy disagree 2015-03-25 17:14:37 etc, etc 2015-03-25 17:14:51 <__mjones> I judge the mateurity of a package and the worldliness of its developers and maintainers by these sorts of metrics. 2015-03-25 17:15:47 <__mjones> ptico: I know how it happens, but it's incumbent upon devs to push back and reign in to those which were necessary and proper, while ignoring those which weren't. 2015-03-25 17:16:26 <__mjones> ptico: also comments in code help a lot. I spent a lot of time trying to port from a very-unpopular and deprecated XML library to libxslt and couldn't get it to work, nor produce any error message. 2015-03-25 17:17:02 <__mjones> Only later did I realise that it's quite possible the original dev started with libxslt, couldn't get it to work either, and managed to cobble up something working with this unpopular (but then still supported) library. 2015-03-25 17:17:29 <__mjones> If the dev had left comments I might have saved myself a dozen hours and not had to endure a lifetime of not knowing whose fault it was. 2015-03-25 17:18:44 btw, ruby have very good development culture and, i think, 40% of modern gems have good test coverage and comments 2015-03-25 17:18:57 <__mjones> But yes, if you have influence to get rub or any other language to revert to system time/date routines, you should 100% do that, always. 2015-03-25 17:19:11 Sorry, need to go, thanks for discussion 2015-03-25 17:19:54 <__mjones> I'm amenable to change. openoffice->libreoffice, openssl->*, gethostbyname() -> getaddrinfo(), etc. 2015-03-25 17:20:08 <__mjones> Bye. 2015-03-25 21:21:30 ncopa, clandmeter: what about gogs instead of redmine? 8) http://gogs.io/ it's freaking easy to package because the tool itself is written in go :) 2015-03-25 21:21:54 go git service... including issue tracker, wiki and pretty soon pull requests :) 2015-03-25 21:22:30 and maybe you can find somebody in the gitter chat who is interested in writing an importer for redmine :) 2015-03-25 21:36:11 <_mjones> mosez: intriguing, but semver < 1.0.0, and majority uptake by users whose charset is CKJV. 2015-03-25 21:46:46 i really like gogs. and the devs are quite nice and open for any help :) 2015-03-25 22:07:39 gogs and its race conditions 2015-03-25 22:07:40 good luck 2015-03-25 22:07:47 and we don't look for code hosting solution 2015-03-25 22:07:52 for one, I use redmine a lot with different projects/teams. Needed feature for me are the issue tracker (with custom workflows), the time tracker, the per sub-project wikis, the documenst, the git/subversion/bazaar integration. Gantt charts are usefull for some of the projects as the ldap link and I suppose other features I forget to mention here... 2015-03-25 22:08:26 if we ever change our git setup, it will be gitolite 2015-03-25 22:10:24 barthalion: race conditions? 2015-03-25 22:11:29 barthalion: so still no option for pull request integrated solution hum? :) 2015-03-25 22:11:42 we have github for that :) 2015-03-25 22:11:59 but aports is not managed through github :) 2015-03-25 22:12:19 as intended 2015-03-25 22:17:59 23:07 < barthalion> gogs and its race conditions which race conditions? 2015-03-25 22:20:31 at least the one in DeliverHooks 3 months ago 2015-03-25 22:20:37 also some silly things with avatars 2015-03-25 22:21:08 also it used to return 404 if refreshed too often 2015-03-25 22:28:50 hum... can not confirm that 8) 2015-03-25 23:29:31 hexchat code is based on xchat, and is currently active maintained, maybe move xchat to unmaintained ? 2015-03-25 23:30:53 ok, makedepends seems different 2015-03-25 23:31:41 gtk+ vs. gtk+2.0 2015-03-25 23:51:09 hmm, gtk+ "metapackage for ... " 2015-03-26 06:59:50 morning 2015-03-26 07:00:28 gogs looks nice indeed 2015-03-26 07:25:16 morning :) 2015-03-26 07:25:54 kunkku: been thinking of the provides for gems 2015-03-26 07:26:58 for gem-x.y.z we do: provides="rb:gem.x rb:gem.x.y rb:gem.x.y.z" 2015-03-26 07:28:09 then if foo depends on gem ~> x.y, we do depends="rb:gem.x.y" 2015-03-26 07:29:11 depends="rb:gem.x 2015-03-26 07:30:59 other option might be to introduce ~> OPERATOR TO APK 2015-03-26 07:31:04 whoops.. caps 2015-03-26 07:41:47 rnalrd: Hi i've found why virt-manager doesn't save sessions 2015-03-26 07:42:08 it requires either dconf or gconf in order to save them 2015-03-26 07:42:36 shouldn't one of these be added in depends ? 2015-03-26 07:42:47 hi 2015-03-26 07:42:50 yes 2015-03-26 07:43:15 i'll add it right away 2015-03-26 07:44:26 cool thx 2015-03-26 07:44:46 i'll add dconf since it's the new replacement for gconf 2015-03-26 07:44:53 ok 2015-03-26 07:44:53 any objection? 2015-03-26 07:44:55 k 2015-03-26 07:45:07 was gonna say that i didn't know which one was better :D 2015-03-26 07:45:28 ACTION either until I googled for it :D 2015-03-26 07:45:48 ^^ 2015-03-26 07:46:25 there may be something fishy as well with the way it handles keys in ssh-agent... 2015-03-26 07:46:30 not sure if it's related 2015-03-26 07:48:29 btw is dconf started automatically when virt-manager starts or should it be started another way ? 2015-03-26 07:50:13 apk does not try to be smart take care of starting services 2015-03-26 07:50:20 taking care* 2015-03-26 07:50:38 so you have to start it manually or add it to autostart 2015-03-26 07:52:57 ok so there's a service to run or it's gnome internal stuff ? 2015-03-26 07:54:32 not sure since i don't use gnome here 2015-03-26 08:27:02 rnalrd: i don't either hence the question ^^ 2015-03-26 08:27:12 ncopa: is it possible to use version constraints with 'provides' tags in apk? 2015-03-26 08:28:23 rnalrd: thx :) 2015-03-26 08:28:31 yw 2015-03-26 08:28:54 fabled: ^^^ re provides and version constraints 2015-03-26 08:29:18 kunkku, what you mean? 2015-03-26 08:29:31 you can do "provides=foo=1.0" 2015-03-26 08:29:36 but that changes the sematics 2015-03-26 08:29:53 it'll also implicitly conflict with foo after that 2015-03-26 08:31:20 depends=foo>1.0 2015-03-26 08:31:35 foobar-1.1 has a provides=foo 2015-03-26 08:32:51 how will apk desl with the depends=foo>1.0 if foo is only available via provides? 2015-03-26 08:32:56 ok, so using them does not work for multiversion support 2015-03-26 08:33:46 i think we will not deal with it before v3.2 anyways 2015-03-26 08:34:02 i have only ~5 pkgs to manually verify 2015-03-26 08:34:06 rubygems often have version constraint like ~> x.y, which means > x.y && < (x+1) 2015-03-26 08:34:26 no.. >= x.y && < (x+1) 2015-03-26 08:34:26 i wonder if we should introduce that to apk 2015-03-26 08:34:55 but at the same time, you need to be able to install multiple versions of the same gem simultaneously 2015-03-26 08:36:27 (or at least mosez wants to do so) 2015-03-26 08:40:33 rnalrd: i see virt-manager in dconf paths and in gsettings schemas 2015-03-26 08:40:38 so i guess it's OK 2015-03-26 08:40:44 i just restarted dbus 2015-03-26 08:41:15 if apk does not allow multiple concurrently installed packages to provide the same tag, gem versions need to be encoded in the tags themselves 2015-03-26 08:41:18 can't access virt-manager remotely from here will confirm this evening 2015-03-26 08:41:31 if it's good I can fix 3.1-stable too, coredumb 2015-03-26 08:41:37 k 2015-03-26 08:42:20 but then the ~> operator becomes problematic 2015-03-26 08:44:07 if you encode version infomation to the tag name, you cannot leverage apk's version constraint logic 2015-03-26 08:45:24 normally, libraries are versioned such that when backward compatbility breaks, the major version number is incremented 2015-03-26 08:45:44 but ruby developers do not follow this practise 2015-03-26 08:46:50 they apparently cast lots to determine which part of the version number to increment, and outsource the version management problem to the users of the library 2015-03-26 08:51:06 kunkku: i have renamed the rails 4.2 packages to ruby-${_gemname}${_railsver} where _railsver=4.2 2015-03-26 08:51:28 + a handful of dependencies 2015-03-26 08:51:57 there are some deps like ruby-mail which has newer version that cannot be upgraded 2015-03-26 08:52:15 https://rubygems.org/gems/mail 2015-03-26 08:52:30 isn't 4.2 the latest version? 2015-03-26 08:52:40 4.2.1 i think 2015-03-26 08:52:44 it is 2015-03-26 08:52:49 which already blocks redmine 2015-03-26 08:52:54 redmine = 4.2.0 2015-03-26 08:52:58 :| 2015-03-26 08:53:12 it does not, redmine 3.0 appears to support rails 4.2 2015-03-26 08:53:20 no 2015-03-26 08:53:31 http://www.redmine.org/projects/redmine/wiki/RedmineInstall#Ruby-interpreter 2015-03-26 08:53:39 oh ruby-mail 2015-03-26 08:53:53 i was talking about rails sorry 2015-03-26 08:54:20 rails 4.2 depends on actionmailer 4.2 2015-03-26 08:54:35 https://rubygems.org/gems/actionmailer 2015-03-26 08:54:56 which depends on mail >= 2.5.4, ~> 2.5 2015-03-26 08:55:06 and ruby-mail 2.6 is out 2015-03-26 08:55:24 why do they depend so strick on minor version numbers? 2015-03-26 08:55:36 does anyone know why rails has been split into zillions of gems that anyway need to be installed/updated simultaneously? 2015-03-26 08:55:51 action* gems? 2015-03-26 08:55:59 and active* 2015-03-26 08:56:06 kunkku: yes it sucks 2015-03-26 08:56:38 i was thinking that we could ship rails in one single package since they all share same VERSION 2015-03-26 08:57:02 but i think the reason is to break circular deps 2015-03-26 08:57:08 i suspect thats the reason 2015-03-26 08:57:37 surely there would be no circular deps if it was shipped in a monolithic package? 2015-03-26 08:58:01 i htink they might have some external deps 2015-03-26 08:58:23 I would keep the rails packaging as is 2015-03-26 08:58:25 for example ruby-sprockets-rails 2015-03-26 08:58:27 ncopa: can i trash this now? http://sprunge.us/bbcc 2015-03-26 08:58:42 otherwise we need more exceptions in the upcoming automation system... 2015-03-26 08:59:13 sprockets-rails depends on activesupport 2015-03-26 08:59:36 kunkku: you mean i should not rename rails to rails4.2? 2015-03-26 08:59:54 I meant we should not combine the rails apks 2015-03-26 08:59:58 ah 2015-03-26 09:00:03 no, i will not do that 2015-03-26 09:00:05 i thought of it 2015-03-26 09:00:13 but decided to not 2015-03-26 09:00:26 as it would create circular deps 2015-03-26 09:01:16 will there be a metapackage named 'ruby-rails' which depends on the latest version? 2015-03-26 09:11:29 what's the easier way to move from the arch from x86 to x86_64 on a disk install? 2015-03-26 09:11:37 s/easier/faster 2015-03-26 09:12:11 boot from a disk, mount the / and unpack the apk ? 2015-03-26 09:16:15 clandmeter: yes you can trash http://sprunge.us/bbcc 2015-03-26 09:17:05 kunkku: i didnt think of adding a meta rails pkg 2015-03-26 09:17:27 since i expect that you will always want pick a specific rails version 2015-03-26 09:17:34 but i could add a meta if you want 2015-03-26 09:22:04 clandmeter: so redmine-3.0.1 has a hard dependency on rails 4.2.0? 2015-03-26 09:22:12 yes 2015-03-26 09:22:37 it it possible to see the gemfile for redmine-3.0.1? 2015-03-26 09:23:05 http://www.redmine.org/projects/redmine/repository/entry/trunk/Gemfile 2015-03-26 09:23:13 thats the trunk version? 2015-03-26 09:24:05 http://www.redmine.org/projects/redmine/repository/entry/branches/3.0-stable/Gemfile 2015-03-26 09:24:18 stable now has 4.2.1 2015-03-26 09:24:24 but the 3.0.1 release doesnt 2015-03-26 09:24:48 i dont understand why they depends on minor 2015-03-26 09:25:12 but then again, i dont understand anything ruby related. 2015-03-26 09:25:58 hard dependcy... that sucks. for rails versions i always write ~> 4.2.0 2015-03-26 09:26:08 what is the expected output of apk search -v --description 2015-03-26 09:26:10 no need to upgrade anything if a patch comes out 2015-03-26 09:26:50 how could one generate a dep list for redmine? 2015-03-26 09:27:05 i guess the tools avaiable wont work? 2015-03-26 09:27:13 cd ~/aports/main && ap recursdeps redmine 2015-03-26 09:27:32 where is that dep stored in the .apk package? 2015-03-26 09:27:52 http://sprunge.us/YLhS 2015-03-26 09:28:17 i mean regular pkgs have gems, but redmine doesnt. 2015-03-26 09:29:02 pkg/redmine/usr/share/webapps/redmine/.Gemfile 2015-03-26 09:29:27 gem "rails", "4.2.0" 2015-03-26 09:29:29 yup 2015-03-26 09:29:31 hum 2015-03-26 09:29:48 i also wonder how to properly patch ruby-* files 2015-03-26 09:30:00 i'd like to patch that to ~> 4.2.0 2015-03-26 09:30:27 nice to do critical security updates when you depend on minor version :) 2015-03-26 09:30:34 \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ 2015-03-26 09:30:41 ///////////////////////////////////////// 2015-03-26 09:30:50 not nice to do security updates... 2015-03-26 09:31:23 that actually stop rollout of security dates til app has pushed update 2015-03-26 09:31:33 apk search -v -d 'anything' always retuns the same three packages 8: 2015-03-26 09:31:55 is that just me? 2015-03-26 09:32:15 is the above gemfile actually used in alpine? 2015-03-26 09:32:24 yes 2015-03-26 09:32:31 well 2015-03-26 09:32:36 no 2015-03-26 09:32:39 i renamed it 2015-03-26 09:32:45 i think bundler uses it 2015-03-26 09:32:45 and added the deps myself 2015-03-26 09:32:58 and it was a headache 2015-03-26 09:33:12 # this should disable bundler dependencies 2015-03-26 09:33:12 mv "$pkgdir"/usr/share/webapps/redmine/Gemfile \ 2015-03-26 09:33:12 "$pkgdir"/usr/share/webapps/redmine/.Gemfile 2015-03-26 09:34:28 ah, redmine is not a .gem 2015-03-26 09:34:28 redmine is not available via rubygems.org? 2015-03-26 09:34:32 apparently not 2015-03-26 09:35:05 means my dependency resolver happily ignores redmine deps :) 2015-03-26 09:35:10 last time i just installed redmine with bunlder and mark the deps down 2015-03-26 09:37:51 but there is a gemfile 2015-03-26 09:38:12 kunkku: can you feed your gem resolver with a gemfile? 2015-03-26 09:38:20 or does it only eat things from net? 2015-03-26 09:39:10 the set of dependencies depends on config/database.yml: http://www.redmine.org/projects/redmine/repository/revisions/14135/entry/trunk/Gemfile 2015-03-26 09:39:22 isnt that a completely diff format? 2015-03-26 09:40:30 currently it uses only rubygems.org 2015-03-26 09:40:36 he uses ruby's gem module and i expect ruby gems be able to read gemfiles 2015-03-26 09:40:54 i still think its useless due to it reads the config/database.yaml 2015-03-26 09:41:12 oh... 2015-03-26 09:41:22 warn("Please configure your config/database.yml first") 2015-03-26 09:41:31 should just ignore those deps if its unconfigured 2015-03-26 09:41:32 this is a big mess 2015-03-26 09:41:36 yes :) 2015-03-26 09:41:48 big mess is an understatement :) 2015-03-26 09:42:05 i need coffe 2015-03-26 09:42:07 bundler reads the Gemfile and feeds it to gem 2015-03-26 09:42:16 atleast thats what i understand 2015-03-26 09:42:48 how about a tool that reads a give gemfile and list out the dependencies? 2015-03-26 09:43:43 bundle viz(1) 2015-03-26 09:43:43 Generate a visual representation of your dependencies 2015-03-26 09:44:14 but i guess a png is difficult to parse :) 2015-03-26 09:44:20 what a bright idea to implement config files as a piece of executable code... 2015-03-26 09:45:05 ncopa: such a function exists in the gem lib 2015-03-26 09:46:33 or maybe even better, run bundler --simulate or similar 2015-03-26 09:46:33 but it does not help if the dependencies depend on database.yml or other run-time aspects 2015-03-26 09:47:07 correct 2015-03-26 09:47:11 i have an idea 2015-03-26 09:47:16 in redmine APKBUILD 2015-03-26 09:47:43 use bundler to verify if all deps are there 2015-03-26 09:47:57 and error out if deps are satisfied 2015-03-26 09:48:21 it will not help the opposite way (can i upgrade ruby-foo without breaking redmine?) 2015-03-26 09:48:48 but at least it will be a step forward 2015-03-26 09:49:56 bundle check 2015-03-26 10:13:12 bundle check is useless 2015-03-26 10:13:26 $ bundle check 2015-03-26 10:13:27 Please configure your config/database.yml first 2015-03-26 10:13:27 Bundler can't satisfy your Gemfile's dependencies. 2015-03-26 10:13:27 Install missing gems with `bundle install`. 2015-03-26 10:13:43 does not tell which is missing 2015-03-26 10:24:10 bunlder list does 2015-03-26 10:24:16 Could not find gem 'request_store (= 1.0.5) ruby' in the gems available on this 2015-03-26 10:24:16 machine. 2015-03-26 10:24:26 i think i would need that gem2apkbuild now... 2015-03-26 10:25:00 I use cp -r ruby-foo ruby-bar :) 2015-03-26 10:25:15 yeah.. 2015-03-26 10:25:25 i see that some has done that already 2015-03-26 10:25:29 hehe 2015-03-26 10:25:35 and forgotten to update pkgdesc 2015-03-26 10:25:35 ACTION hides 2015-03-26 10:25:40 :) 2015-03-26 10:28:04 gem "request_store", "1.0.5" 2015-03-26 10:28:06 whatta 2015-03-26 10:28:12 must be 1.0.5 2015-03-26 10:28:17 cannot be 1.0.6 2015-03-26 10:28:37 or 1.0.8 2015-03-26 10:29:13 suddenly I realize ruby-redmine-* might actually be a good idea 2015-03-26 10:29:55 :) 2015-03-26 10:30:18 we will need ruby--* 2015-03-26 10:30:21 for every app 2015-03-26 10:30:26 horrible... 2015-03-26 10:30:43 yes i told you :) 2015-03-26 10:30:48 and i had it all ready 2015-03-26 10:30:49 i think I'll just patch the gemfile 2015-03-26 10:30:55 but now its rm -rf :p 2015-03-26 10:31:16 actually 2015-03-26 10:31:25 ruby- is a horrible idea :) 2015-03-26 10:35:51 talking with freebsd ppl in #redmine 2015-03-26 10:35:55 then patch the gemfile 2015-03-26 10:36:00 they* 2015-03-26 10:36:03 i'll do the same 2015-03-26 10:36:21 force redmine relax a bit 2015-03-26 13:15:26 phew 2015-03-26 13:15:33 finally im done with all ruby-* 2015-03-26 13:15:41 ... in main/ 2015-03-26 13:16:06 git rm -r testing 2015-03-26 13:16:46 apkbuild-gem-resolver.in:107:in `block in depends': Dependency for ruby-rails-deprecated_sanitizer4.2 does not exist: ruby-activesupport (RuntimeError) 2015-03-26 13:16:50 nice 2015-03-26 13:16:55 it caught mistakes 2015-03-26 13:18:56 hm 2015-03-26 13:19:09 i acutally thing its a gem resolver limitation/bug 2015-03-26 13:19:32 kunkku: apkbuild-gem-resolver does not deal with: depends="ruby ruby-activesupport${_railsver}" 2015-03-26 13:20:55 did you add the 4.2 suffix to depends in ruby-rails-deprecated_sanitizer4.2/APKBUILD? 2015-03-26 13:23:24 i did:http://sprunge.us/ZMNI 2015-03-26 13:24:06 to make it easier to migrate to rails 4.3 2015-03-26 13:24:22 ah... 2015-03-26 13:24:44 the gem resolver does not evaluate the file in shell 2015-03-26 13:25:04 that is one of the reasons that i would like to change the APKBUILD format 2015-03-26 13:25:07 to somethign new 2015-03-26 13:25:16 the only way to parse it is via shell 2015-03-26 13:25:29 you need to execute it in a shell 2015-03-26 13:25:44 now it uses augeas to parse it 2015-03-26 13:26:12 I thought it would be easier to update the file programmatically this way 2015-03-26 13:26:36 e.g. auto-update dependencies 2015-03-26 13:26:54 but this becomes difficult if it needs to understand shell variables... 2015-03-26 13:27:38 yes 2015-03-26 13:27:39 hm 2015-03-26 13:28:01 i suppose i could sed $_railsver 2015-03-26 13:28:32 or you would need to parse it via shell 2015-03-26 13:28:51 my vision was to add gem2apkbuild functionality to the resolver 2015-03-26 13:29:21 in which case using shell variables to reduce manual work becomes pointless 2015-03-26 13:29:58 ok i'll sed _railsver 2015-03-26 13:38:14 and rebase -i 2015-03-26 13:46:57 gem resolver found alot of mistakes i did 2015-03-26 13:51:22 kunkku: i have a problem 2015-03-26 13:51:35 ruby 2.2.1 ships io-console 2015-03-26 13:51:43 version 0.4.3 2015-03-26 13:51:59 however, the 0.4.3 version does not exist online: https://rubygems.org/gems/io-console 2015-03-26 13:53:34 https://gist.github.com/ncopa/7fd228b4aa5c62da8357 2015-03-26 13:59:25 i'll just say that io-console is 0.4.2 even if that is a lie 2015-03-26 14:00:03 \o/ 2015-03-26 14:00:22 gemresolver is happy 2015-03-26 14:13:12 not understanding this, why are gems installed under, /usr/lib/ruby/gems/2.1.0/gems 2015-03-26 14:13:22 and not just, /usr/lib/ruby/gems/ 2015-03-26 14:14:47 the gems pkgs have versions eg actionmailer-4.1.8 2015-03-26 14:16:33 is there plan to have multiple version of say pkg actionmailer-4.1.8 for different gem version ? 2015-03-26 14:19:55 if a pkg is a gem why not ruby-gem-.apk ? 2015-03-26 14:20:42 coz, there are also ruby extentions, which will distinguish it from gem 2015-03-26 14:22:22 unless there is plan to ship for 2 ruby version 2015-03-26 14:23:07 it is theorietically possible to shipe 2 different ruby versions 2015-03-26 14:23:22 i dont think a ruby module installed with ruby 2.1 will work in ruby 2.2 2015-03-26 14:23:30 or, more 2015-03-26 14:23:45 there is no guarantee that a gem installed with ruby 2.1 will work in ruby 2.2 2015-03-26 14:23:53 so we need rebuild all ruby-* packages anyways 2015-03-26 14:24:05 I thought gems were rubyVersion agnostic 2015-03-26 14:24:05 and ruby 2.2 comes here: .... 2015-03-26 14:24:22 i have seen no evidence that they are rubyver agostic 2015-03-26 14:24:40 seems like they are very much not anythingversion agnostic 2015-03-26 14:25:04 once a gem pkg is installed, the scripts calling it does not go through gem, it just looks for gem-install-dir/pkg, i guess 2015-03-26 14:25:12 it seems that the common practice is to install whatever version that the developer happened to have a that specific time 2015-03-26 14:25:48 so you dont need test all kind of version combinations 2015-03-26 14:25:54 here comes ruby 2.2 2015-03-26 14:28:07 still think gem pkg name be kinda ruby-gem-.apk 2015-03-26 14:38:21 we should upgrade python to python 3 too... 2015-03-26 14:39:17 if there should be rubyVersion, then it would be like /usr/lib/ruby/gems/... 2015-03-26 14:39:31 and proper environment var set 2015-03-26 14:39:58 i am not really in the mood for rebuildign evey ruby-* again... 2015-03-26 14:40:04 :) 2015-03-26 14:40:12 is current approach broken? 2015-03-26 14:40:35 I don't know, will be clear in next redmine update 2015-03-26 14:41:05 ncopa: did you try redmine, if it works with those deps? 2015-03-26 14:41:22 or someone doing extensive/edgy development on ruby 2015-03-26 14:42:05 clandmeter: nope 2015-03-26 14:42:16 but bundler does not complain 2015-03-26 14:42:29 you should give it a try :) 2015-03-26 14:43:06 create the session and generate and sqlite db should be enough to make it work 2015-03-26 14:47:40 and unlikely to have 2gem version under same rubyVersion 2015-03-26 14:59:17 ok, got it , /usr/lib/ruby/gems/2.1.0/gems , 2.1.0 is ruby version 2015-03-26 14:59:27 not gem version !! 2015-03-26 15:00:37 nope, ruby-2.1.5-r1 2015-03-26 15:01:00 and /usr/lib/ruby/gems/2.1.0/gems/actionmailer-4.1.8/ 2015-03-26 15:07:28 module Gem 2015-03-26 15:07:29 VERSION = '2.2.2' 2015-03-26 15:07:29 end 2015-03-26 15:08:08 ncopa: well, we have python3 2015-03-26 15:08:21 im mean all the py-* packages 2015-03-26 15:08:21 ncopa: and IIRC two modules in testing that are built against it 2015-03-26 15:08:35 so, technically, we support py3 :P 2015-03-26 15:08:40 :) 2015-03-26 15:08:45 http://permalink.gmane.org/gmane.comp.security.oss.general/16288 fun 2015-03-26 15:08:54 i was more thinking of getting rid of python2 2015-03-26 15:09:09 I don't think we should do it 2015-03-26 15:09:13 ok 2015-03-26 15:09:18 thats the selinux thingy? 2015-03-26 15:09:20 cool :) 2015-03-26 15:09:45 py2 is still used by some big projects, so I think it makes sense to have it for a release or two 2015-03-26 15:17:51 nice, my mSata for APU4C pc arrived, finnaly i can do sys install :) 2015-03-26 15:19:26 any simple guide to set up minimal redmine with sqlite? 2015-03-26 15:19:58 apk add redmine 2015-03-26 15:20:00 done 2015-03-26 15:20:09 ncopa about the edge release iso are they also with vanilla kernel http://nl.alpinelinux.org/alpine/edge/releases/x86_64/ from naming convetion i dont see it 2015-03-26 15:20:39 bundle exec rake generate_secret_token 2015-03-26 15:20:50 RAILS_ENV=production bundle exec rake db:migrate 2015-03-26 15:20:58 both inside the redmine dir 2015-03-26 15:21:07 and setup the database config 2015-03-26 15:21:26 bundle exec ruby script/rails server webrick -e production 2015-03-26 15:21:35 that will start webrick to test 2015-03-26 15:22:47 I tried redmine with sqlite, awhile back, there is config/database.yml to be modified 2015-03-26 15:23:04 let me see if I still have the configs lying around 2015-03-26 15:24:14 you can just remove everything except the sqlite part 2015-03-26 15:24:30 it will generate the db in redmine/db 2015-03-26 15:24:31 hum it complained about the missing .Gemfile 2015-03-26 15:24:54 yes because you didnt use bundler to install the deps 2015-03-26 15:25:01 you can rename it back to Gemfile and try 2015-03-26 15:25:52 yeah 2015-03-26 15:25:54 that worked 2015-03-26 15:27:17 you can generate the db? 2015-03-26 15:27:21 whee \o/ 2015-03-26 15:27:22 yup 2015-03-26 15:27:26 nice 2015-03-26 15:27:31 i couldnt yesterday 2015-03-26 15:27:36 complained about openid or whatever 2015-03-26 15:27:37 and i can connect to localhost:3000 2015-03-26 15:27:46 i think i have fixed those 2015-03-26 15:28:06 and i have lobotomized the testing and development groups 2015-03-26 15:28:19 only production works 2015-03-26 15:28:29 nice 2015-03-26 15:28:31 it does work 2015-03-26 15:30:37 what happens when a new redmine version is out? 2015-03-26 15:30:43 any procedure? 2015-03-26 15:32:23 3.0.2? 2015-03-26 15:32:38 abump redmine-3.0.2 2015-03-26 15:32:47 you might need fix the patch 2015-03-26 15:32:54 i patch the gemfile 2015-03-26 15:32:57 to relax a bit 2015-03-26 15:33:15 then it will run bundler check 2015-03-26 15:33:25 to verify that all deps are properly satisfied 2015-03-26 15:33:35 if it is not, it will ERROR out 2015-03-26 15:33:54 currently 2015-03-26 15:34:11 if rails4.2 gets updated to rails 4.2.2 2015-03-26 15:34:15 redmine will still work 2015-03-26 15:34:21 (hopefully) 2015-03-26 15:34:41 ok 2015-03-26 15:34:51 i think we should keep the Gemfile there 2015-03-26 15:36:18 nice 2015-03-26 15:36:29 i wonder if we should make a docker file for redmine 2015-03-26 15:36:43 might be a nice way to test it 2015-03-26 15:53:45 ok i am pretty happy with redmine 2015-03-26 15:53:50 i think we have more or less success 2015-03-26 15:54:24 to update ruby-* packages in future, we can use apkbuild-gem-resolver to verify that we dont break anything 2015-03-26 15:54:38 and simply rebuilding redmine should verify that redmine didnt break 2015-03-26 15:57:41 alpine infra will be down tomorrow for periods 2015-03-26 15:57:50 that is bugs.a.o, www.a.o msg.a.o 2015-03-26 15:57:56 git.a.o should be up 2015-03-26 16:22:23 rnalrd: a friend of mine upgraded virt-manager and confirmed that settings are now saved :) 2015-03-26 17:03:03 ping ncopa 2015-03-26 17:05:05 http://pkgs.alpinelinux.org/contents?pkgname=ruby-unicorn&arch=x86 , nice, ruby version set ok 2015-03-26 17:05:35 confusion clear 2015-03-26 20:05:40 <_mjones> Looks like ftrace is enabled in kernel .36-0: 2015-03-26 20:05:52 <_mjones> CONFIG_HAVE_DYNAMIC_FTRACE=y 2015-03-26 20:06:13 <_mjones> ...but I would've sworn it was off in .30. Maybe there's something in the release notes. 2015-03-26 21:14:18 hmm. I build outside aports to test, and i moved my build directories somewhere new and now abuild tells me "fatal: No names found, can not track anything" at the end of every build 2015-03-26 21:15:26 before i poke around, this familiar? I guess i can't find my local packages direcrtory since i moved abuild directories. 2015-03-26 21:29:11 <_mjones> Yes, I guess it keeps state after you build. 2015-03-26 21:29:25 <_mjones> It doesn't like the absence of a 'packages' dir when you do that either, I've found. 2015-03-26 21:29:41 <_mjones> It does it per-package if it matters. Copy a package and compile it clean and you're ok. 2015-03-27 00:15:37 "apk add redmine" on a fresh edge LXC is running fine 2015-03-27 00:16:46 it does install 92 packages 2015-03-27 00:17:48 I will try to use my running instance (mysal DB and so on) tomorrow (today) morning... 2015-03-27 00:19:40 I wil update #3941 with results 2015-03-27 08:29:07 guys 2015-03-27 08:29:41 we are shutting down bugs.a.o, www.a.o, msg.a.o, forums.a.o pkgs.a.o for a while now 2015-03-27 08:29:56 upgrading the box 2015-03-27 08:30:02 moving from vserver to lxc 2015-03-27 08:30:15 internal network changes 2015-03-27 08:30:40 and nl.a.o and rsync.a.o 2015-03-27 09:48:32 building dictd fails here -> http://sprunge.us/NEiV 2015-03-27 09:49:07 and i sort of stuck. lex.yy.c is no where do be seen 2015-03-27 09:51:57 ScrumpyJack: sounds like parallel build issue 2015-03-27 09:52:09 does make -j1 help? 2015-03-27 09:52:18 lex.yy.c is probably generated from lex.yy 2015-03-27 09:52:20 or similar 2015-03-27 09:52:31 you will probalby need bison and flex for that 2015-03-27 09:53:01 yeah, bison is in makedeps 2015-03-27 09:55:02 i'll try flex instead (although confugure is happy with finding bison) 2015-03-27 09:56:23 that help, but more errors now :) 2015-03-27 11:10:46 algitbot went postal 2015-03-27 11:12:28 barthalion, we are working on it 2015-03-27 11:12:36 just some things we didnt think about :) 2015-03-27 11:12:39 like usual 2015-03-27 11:26:08 ncopa, something wrong in the build scripts, now that the rsync master is temporarily down? 2015-03-27 11:36:45 rsync master is down 2015-03-27 11:37:05 we lost the host keys 2015-03-27 11:37:07 ssh keys 2015-03-27 11:37:19 so i need mount the virtual disk and extract them 2015-03-27 11:37:34 or we need reconfig all builders 2015-03-27 11:37:42 update with new ssh host key 2015-03-27 11:40:01 and who said algibot does not do its job, look such adverse conditions but still working 2015-03-27 11:40:21 hrrr, buck up 2015-03-27 11:41:05 im stopping arm builder 2015-03-27 11:41:45 :) 2015-03-27 11:46:48 ncopa, when new ssh keys are generated, zip key+pub, encrypt using your ssh key and email it to your self 2015-03-27 11:49:50 or me, I can write to dvd, and put it under the carpet ;) 2015-03-27 11:56:39 anyone have idea how to mount a partition that is on a raw lvm? 2015-03-27 11:57:03 parted /dev/vg0/alpine-vserver print 2015-03-27 11:57:09 Number Start End Size Type File system Flags 2015-03-27 11:57:09 1 512B 105MB 105MB primary ext4 boot 2015-03-27 11:57:09 2 105MB 4313MB 4208MB primary linux-swap(v1) 2015-03-27 11:57:09 3 4313MB 419GB 415GB primary ext4 2015-03-27 11:57:27 mount -o offset appears to try losetup 2015-03-27 11:57:31 <_mjones> ACTION doesn't use much LVM. :( 2015-03-27 11:57:42 but losetup fails on the lvm device 2015-03-27 12:00:52 does "kpartx -l /dev/vg0/alpine-vserver" give you some meaningfull answer? 2015-03-27 12:01:06 ncopa, what error losetup gives? 2015-03-27 12:02:06 alpine-nl:~# mount -o ro,offset=512 /dev/mapper/vg0-alpine--vserver /mnt 2015-03-27 12:02:06 mount: wrong fs type, bad option, bad superblock on /dev/loop1, 2015-03-27 12:02:07 missing codepage or helper program, or other error 2015-03-27 12:02:25 alpine-nl:~# losetup -o512 /dev/mapper/vg0-alpine--vserver 2015-03-27 12:02:25 losetup: /dev/mapper/vg0-alpine--vserver: failed to use device: No error information 2015-03-27 12:02:45 can probably no losetup an lvm device 2015-03-27 12:02:49 check dmesg 2015-03-27 12:02:55 it has more msgs 2015-03-27 12:04:32 [ 9509.215737] EXT4-fs (loop1): INFO: recovery required on readonly filesystem 2015-03-27 12:04:32 [ 9509.215739] EXT4-fs (loop1): write access unavailable, cannot proceed 2015-03-27 12:04:42 maybe its the ro option 2015-03-27 12:05:07 ha 2015-03-27 12:05:08 yes 2015-03-27 12:05:11 vkrishn: thanks!!! 2015-03-27 12:05:30 :) 2015-03-27 12:06:03 wonderful 2015-03-27 12:25:28 http://nl.alpinelinux.org is up 2015-03-27 12:25:57 rsync://rsync.alpinelinux.org is up 2015-03-27 12:27:30 msg.alpinelinux.org is up 2015-03-27 12:30:19 huh 2015-03-27 12:33:19 does mqtt require persistence enabled ? 2015-03-27 12:33:35 mqtt does not require that 2015-03-27 12:33:44 but its handy for build servers 2015-03-27 12:33:56 ok now it should work 2015-03-27 12:33:57 does it report of each file separately ? (was asking myself if "files" must not be "file" in that case :) ) 2015-03-27 12:34:32 no, it faild due to ssh known_hosts verification failure 2015-03-27 12:34:35 and it will loop 2015-03-27 12:35:02 it shouldnt report success 2015-03-27 12:35:11 instead it should have reporeted failure 2015-03-27 12:35:41 package building should be back to normal now 2015-03-27 12:35:41 hu ok :) 2015-03-27 12:55:16 whee, http://alpinelinux.org is up again 2015-03-27 13:11:22 forum.alpinelinux.org/forum , still functional or not restarted ? 2015-03-27 13:12:16 non-funtional 2015-03-27 13:12:24 not sure if we will fix it either 2015-03-27 13:12:28 or move to something else 2015-03-27 13:12:48 do we want save the current messages there? 2015-03-27 13:13:00 I don't think so 2015-03-27 13:14:02 some stackoverflow clone would be good, or categoryless forum, but I'm not sure how would it work out 2015-03-27 13:14:32 im looking at flaskbb 2015-03-27 13:14:43 http://flaskbb.org/ 2015-03-27 13:14:47 looks kinda nice 2015-03-27 13:15:43 why not fluxbb? 2015-03-27 13:15:53 at least proven to work 2015-03-27 13:16:05 I don't recall any forum using flaskbb 2015-03-27 13:17:11 "FlaskBB is not yet stable so changes might occur without warning." 2015-03-27 13:17:36 "FlaskBB comes with a lot of inbuilt features.." 2015-03-27 13:18:46 newsboard.unclassified.de 2015-03-27 13:19:16 it was mostly due to its python instead php 2015-03-27 13:19:43 but fluxbb is probably a better alternative 2015-03-27 13:21:16 fluxbb 2.0 is dead upstream 2015-03-27 13:22:07 fluxbb 1.x will upgrade to flarum 2015-03-27 13:22:15 http://fluxbb.org/forums/viewtopic.php?id=8282 2015-03-27 13:25:10 I guess it's fine to use 1.x 2015-03-27 13:25:28 or wait for flarum 2015-03-27 13:25:47 i suppose so 2015-03-27 13:26:13 http://demo.flarum.org/ nice 2015-03-27 13:27:18 although I hope classical look will be also available 2015-03-27 13:36:12 ugh, so much js 2015-03-27 13:36:41 would be nice to be able to import old users and old messages 2015-03-27 13:36:46 but not sure if its worth it 2015-03-27 13:37:01 its kinda sad for those who have spend hours there... 2015-03-27 13:38:22 "Flarum is forum software for the future, a future where JavaScript web apps are the norm and Internet connections and mobile devices are fast enough to comfortably handle them" 2015-03-27 13:38:36 sad 2015-03-27 13:38:45 ncopa: what about read-only copy? 2015-03-27 13:38:48 the future is sad :) 2015-03-27 13:38:51 hum 2015-03-27 13:38:54 that might work 2015-03-27 13:39:11 maybe http slurp it 2015-03-27 13:39:15 as static pages 2015-03-27 13:39:24 or something 2015-03-27 13:39:27 yeah 2015-03-27 13:39:30 "Flarum’s JavaScript assets, when minified and gzipped, will be hopefully no more than about 400 kB." 2015-03-27 13:39:39 lol 2015-03-27 13:39:39 ok 2015-03-27 13:39:54 I can like systemd, but I'll never like tons of JS 8) 2015-03-27 13:40:05 ohhhhh 2015-03-27 13:40:08 great idea!!! 2015-03-27 13:40:15 implement systemd in JS 2015-03-27 13:40:19 for the WEB!!! 2015-03-27 13:40:50 sure, there is libmicrohttpd dependency for vieving logs online already 2015-03-27 13:40:51 then can all mobile device enjoy the systemd experience! 2015-03-27 13:43:03 "JavaScript Apps MAY result in a better user experience, IF (again the magic word) done right" wat 2015-03-27 13:43:29 "You're in a thread and click on something like "back to the overview". A good JavaScript App may instantly display the thread-overview" I don't think he heard about caching 2015-03-27 13:44:31 I get his point about easy development, but it's not what fluxbb was liked for 2015-03-27 14:16:19 <_mjones> hopefully be no more than 400k?! I'm all for ecmascript, but still. 2015-03-27 14:18:02 <_mjones> openid and/or google login and oauth2 would be quite nice to support. 2015-03-27 14:33:09 <_mjones> uh, tls compression was on by default? Oops. 2015-03-27 14:33:21 well, it's kinda upstream bug 2015-03-27 14:33:28 they suggest 'no-comp' but that breaks ABI 2015-03-27 14:33:38 most sane apps disable it 2015-03-27 14:33:40 but yeah 2015-03-27 14:33:44 <_mjones> oh, I didn't know about abi breakage 2015-03-27 14:33:57 thanks so much for 'good defaults' in openssl 2015-03-27 14:34:22 <_mjones> I think everybody knows the story of openssl by now, and why the code was always a nightmare. 2015-03-27 14:34:24 seem arch has same issue still 2015-03-27 14:34:38 all distributions seem to do this different 2015-03-27 14:34:46 I don't think anyone was bothered to fix it actually 2015-03-27 14:34:47 the patch i took is from ubuntu 2015-03-27 14:34:57 but debian switched to 'no-comp' at some point 2015-03-27 14:35:23 and given how pacman doesn't track sonames, things like that just go into repos unnoticed 2015-03-27 14:35:26 <_mjones> Not many people really knew exactly how bad until after heartbleed and the http://opensslrampage.org/ 2015-03-27 14:36:05 additionally I'm pretty sure that maintainer won't apply it once I ask about it 2015-03-27 14:36:26 <_mjones> openssl was the dustman of open-source: nobody else wanted to do the job. 2015-03-27 14:45:20 I cannot connect to bugs.a.o. Only read access; i wanted to add tests results for the redmine ticket with mysql (It fails on the rake db:migrate stuff complaining on timezone data missing) 2015-03-27 14:45:47 Jean-Scotch: sorry. i cannot log in either 2015-03-27 14:45:56 its a fw issues for https i think 2015-03-27 14:46:05 ok. so you are aware of this issue ;) 2015-03-27 14:46:09 might be we need wait til clandmeter is back on monday 2015-03-27 14:46:13 i emailed him about it 2015-03-27 15:04:02 <_mjones> So I've been interested in an x32 dist for a little while now, for smaller-mem VM guests, but now I see an additional target. 2015-03-27 15:04:21 <_mjones> Dell is shipping an awfully nice x86-64 tablet with 2-3GB of core. 2015-03-27 15:04:32 we dont support x32 2015-03-27 15:04:48 i dont feel its worth the effort 2015-03-27 15:05:49 <_mjones> It might not be. I'm interested in finding out if it is. 2015-03-27 15:08:00 <_mjones> ncopa: would you choose to run amd64 or i686 with less ram? 2015-03-27 15:09:06 dunno 2015-03-27 15:09:15 ncopa: I think #3941 is not totally resolved, see http://wiki.alpinelinux.org/wiki/User_talk:Jch#redmine (as I cannot add to b.a.o for now) 2015-03-27 15:10:18 Jean-Scotch: clandmeter mentioned that too 2015-03-27 15:10:25 thanks for an exact error message 2015-03-27 15:10:32 i think apk add tzdata solves it 2015-03-27 15:10:39 ;)it doesn't 2015-03-27 15:10:41 but it means that tzdata should be added as a dep 2015-03-27 15:10:59 I mean ruby-tzdata doen't 2015-03-27 15:11:25 i think its ruby-tzdata thats broken 2015-03-27 15:11:33 can you apk add tzdata 2015-03-27 15:11:37 and see if that solves it? 2015-03-27 15:11:44 yes tzdata do the trick 2015-03-27 15:12:00 ruby-tzinfo 2015-03-27 15:12:06 should depend on tzdata 2015-03-27 15:13:21 that should fix it 2015-03-27 15:28:41 Hello for everybody 2015-03-27 15:31:00 Sorry if I bother you, I have a little problem and I want to see if someone could replicate it 2015-03-27 15:32:36 I'm running 3.1.2 version of Alpinelinux and I installed Opensc 2015-03-27 15:33:16 I want to be able to read a Feitian ePass2300 usb token 2015-03-27 15:33:54 do you have an error message? 2015-03-27 15:34:01 but when I run command "opensc-tool -l" what I get is an error 2015-03-27 15:34:13 0x50ccf754 15:45:47.656 [opensc-tool] ctx.c:722:sc_context_create: =================================== 2015-03-27 15:34:31 Failed to establish context: Unable to load external module 2015-03-27 15:34:39 soryy for the firs part 2015-03-27 15:35:20 I switched on debug and I have more lines 2015-03-27 15:35:21 Laurentius77, you have pcsc and openct? 2015-03-27 15:35:30 grsec or vanilla kernel? 2015-03-27 15:36:00 pcsc and grsec kernel is not yet working by default- we have http://bugs.alpinelinux.org/issues/3280 open on that 2015-03-27 15:48:02 I have opensc + pcsc-lite 2015-03-27 15:48:17 Shoud I have openct 2015-03-27 15:48:20 ? 2015-03-27 15:51:35 hah, i wanted to say that it works for me, then i found out i'm the submitter of the bug and i posted a workaround \o/ 2015-03-27 15:53:28 so it doesn't work now? 2015-03-27 15:53:51 do you have an workaround 2015-03-27 15:53:57 it does. if that's your problem as described in the bug, the workaround should fix it. 2015-03-27 15:54:42 were the symptoms the same? 2015-03-27 15:55:01 beacause I read some bugs 2015-03-27 15:55:29 your symptoms are not very informative and don't quite remember the details anymore, it was 9 months ago that i looked at this. 2015-03-27 15:55:49 have you read this bug from earlier? http://bugs.alpinelinux.org/issues/3280 2015-03-27 15:56:03 I know that 2015-03-27 15:56:19 tried your workaround with no succes 2015-03-27 15:59:50 probably is something different 2015-03-27 16:01:35 rpu3uO8PEVZZ, just so we have not forgotten the bug. we should have soon the infrastructure code ready. fix is planned for alpine 3.2 stable due in few months. 2015-03-27 16:03:07 good 2015-03-27 16:03:26 if it will be fixed 2015-03-27 16:05:04 now is unusable for me 2015-03-27 16:05:50 I compiled the package myself but the results were the same 2015-03-27 16:06:22 tried 3280 bug workaround with no success 2015-03-27 16:06:38 mdev or udev? 2015-03-27 16:06:45 maybe the device nodes are not created right 2015-03-27 16:07:25 I will try again 2015-03-27 16:08:42 https://bugs.alpinelinux.org/ is not available for me this time 2015-03-27 16:11:15 the errors are these: 2015-03-27 16:11:33 0x50ccf754 15:45:47.656 [opensc-tool] ctx.c:722:sc_context_create: =================================== 2015-03-27 16:11:50 0x50ccf754 15:45:47.656 [opensc-tool] ctx.c:723:sc_context_create: opensc version: 0.14.0 2015-03-27 16:12:07 0x50ccf754 15:45:47.656 [opensc-tool] reader-pcsc.c:666:pcsc_init: PC/SC options: connect_exclusive=0 disconnect_action=1 transaction_end_action=0 reconnect_action=0 enable_pinpad=1 enable_pace=1 2015-03-27 16:12:24 0x50ccf754 15:45:47.656 [opensc-tool] ctx.c:795:sc_release_context: called 2015-03-27 16:12:37 0x50ccf754 15:45:47.656 [opensc-tool] reader-pcsc.c:745:pcsc_finish: called 2015-03-27 16:12:55 Failed to establish context: Unable to load external module 2015-03-27 16:13:18 do you have straces? 2015-03-27 16:13:29 do you have a udev rule for your device? 2015-03-27 16:13:41 are you doing this as root? 2015-03-27 16:13:50 yes, as root 2015-03-27 16:13:53 the last line: Unable to load external module 2015-03-27 16:13:57 which module? 2015-03-27 16:14:10 I didn't insert any device 2015-03-27 16:14:24 and if I insert it the rerror is the same 2015-03-27 16:14:45 is not mentioned anywhere 2015-03-27 16:14:48 which module 2015-03-27 16:15:11 I don't know what module fail to load 2015-03-27 16:15:33 have you straced it? 2015-03-27 16:15:38 have you enabled debugging? 2015-03-27 16:15:49 so the error is the same if I insert ePass2003 2015-03-27 16:16:04 yes, the errors are those from above 2015-03-27 16:16:38 ? 2015-03-27 16:16:57 I enabled debug mode 2015-03-27 16:17:00 how? 2015-03-27 16:17:16 LIBCCID_ifdLogLevel=0x00F pcscd -f -T -d -a 2015-03-27 16:17:23 from opensc config file 2015-03-27 16:17:44 I will try that 2015-03-27 16:17:45 hmm, i dunno about opensc. i use libccid + pcscd 2015-03-27 16:18:12 I have no experience with etokens 2015-03-27 16:18:16 i only need this in gpg, and that is enough. i guess you need opensc, if you want to do else than pgp 2015-03-27 16:18:27 give me an advice 2015-03-27 16:18:45 I want it for openvpn + pkcs11-helper 2015-03-27 16:18:50 no clue 2015-03-27 16:18:51 just for that 2015-03-27 16:18:54 sorry 2015-03-27 16:19:04 strace it, and find out what open fails 2015-03-27 16:19:23 how can I strace it? 2015-03-27 16:19:45 i guess opensc? dunno what complains about the missing module 2015-03-27 16:20:11 opensc-tool -l command 2015-03-27 16:20:33 strace -fo /tmp/opensc.strace opensc-tool -l command 2015-03-27 16:20:50 but i guess, if you never seen strace output, might be daunting at first. 2015-03-27 16:20:52 I expected other results but I got those errors from above 2015-03-27 16:20:57 ok 2015-03-27 16:21:01 <_mjones> strace 2015-03-27 16:21:10 I will try those tonight 2015-03-27 16:21:55 <_mjones> strace isn't so bad for newbies -- usually the last things before it fails are what you need anyway. 2015-03-27 16:21:59 but knowing strace is like knowing where the wd40 is 2015-03-27 16:22:21 <_mjones> Although I suppose if you don't know what fopen() means you might not know where to go from there. 2015-03-27 16:22:23 :) 2015-03-27 16:22:48 I will read more 2015-03-27 16:23:02 this is the only solution 2015-03-27 16:23:14 you gave me good clues 2015-03-27 16:23:24 if something doesn't work you have not applied enough strace/wd40/bruteforce 2015-03-27 16:23:35 ok 2015-03-27 16:24:18 I will play with strace 2015-03-27 16:24:24 tonight 2015-03-27 16:24:28 thank you 2015-03-27 16:25:47 <_mjones> There should be a web cartoon about debuggers. 2015-03-27 16:54:38 lo 2015-03-27 16:54:49 how is it going with ruby 2.2? =) 2015-03-27 18:24:58 mosez: seems it's now in edge 2015-03-27 21:04:37 Hi folks 2015-03-27 21:04:43 i'm packaging a new app 2015-03-27 21:05:02 where the dev had the nice idea to provide sources on a google drive 2015-03-27 21:05:05 https://drive.google.com/uc?export=download&id=0Bzghs6gQWi60LV9rM3RMQk85Z1E 2015-03-27 21:05:23 crappy link gives a crappy source name ... how should i workaournd that 2015-03-27 21:05:25 ? 2015-03-27 21:11:24 $pkgname-$pkgver.tar.gz::http://domain.tld/craPpyFilename 2015-03-27 21:11:30 coredumb: ↑ 2015-03-27 21:16:14 barthalion: nice thx ! 2015-03-28 05:56:28 oh 2015-03-29 09:22:57 please git pull git://git.alpinelinux.org/fab/aports 2015-03-29 09:23:37 it would be nice if those updates would be pulled thanks 2015-03-29 13:06:50 ncopa: it's sunday, go away and rest 2015-03-29 13:07:21 ok :) 2015-03-29 13:07:33 i am only picking up a charger i forgot... 2015-03-29 13:07:47 see u tm 2015-03-29 13:07:47 on irc? :P 2015-03-29 13:08:03 cu 2015-03-29 13:47:41 ircharger :p 2015-03-29 14:47:25 ncopa, thanks 2015-03-30 00:28:46 <_mjones> nothing in alpine depends on xfsprogs. This is probably why nobody has thought about the headers situation. 2015-03-30 06:01:40 morning 2015-03-30 06:02:08 _mjones: xfsprogs headers situation? 2015-03-30 06:03:51 <_mjones> right. 2015-03-30 06:04:28 <_mjones> Debian has an earlier version of xfsprogs, from the same source (oss.sgi.com), but dramatically fewer headers. 2015-03-30 06:05:53 <_mjones> The one that jumps out is Debian's xfs.h doesn't include . 2015-03-30 06:06:49 <_mjones> A few days ago when working on ceph port, I was having a lot of trouble with headers and got some feedback from #musl. Also at that time I noticed the copyright date on the xfsprogs headers from SGI stopped in 2005. 2015-03-30 06:07:30 <_mjones> Specifically, too, ceph wanted the typedef of loff_t (instead of off_t which I gather is usual) and it wasn't available on alpine. 2015-03-30 06:08:01 probably because off_t is posix and loff_t is not 2015-03-30 06:08:05 <_mjones> So I was going to look into what Debian did for headers, presuming it's not just a difference between the package versions. 2015-03-30 06:08:30 it it only a problem with loff_t 2015-03-30 06:08:32 ? 2015-03-30 06:10:21 <_mjones> I worked around that problem, but I was concerned that there might be others. 2015-03-30 06:11:12 common issues with musl are: 2015-03-30 06:11:30 well, when working with musl - issue is not musl... 2015-03-30 06:11:43 #include 2015-03-30 06:11:53 <_mjones> There are 52 files in /usr/include/xfs on alpine, but 7 files in /usr/include/xfs on Debian 7.8, even though they use the same upstream. 2015-03-30 06:12:06 <_mjones> I know a lot of it is gnu. 2015-03-30 06:12:07 glibc includes linux header stuff. musl does not 2015-03-30 06:12:23 so you might need add linux-headers to makedepends 2015-03-30 06:12:23 <_mjones> gnu defines loff_t somewhere after a complicated series of indirections. 2015-03-30 06:12:53 other common issues are bsd extensions like sys/queue.h 2015-03-30 06:13:04 we have bsd-compat-headers package for that 2015-03-30 06:13:12 <_mjones> It wasn't that one typedef, I don't understand the dramatic difference in files. But I haven't looked into the matter. 2015-03-30 06:13:59 <_mjones> bsd-compat-headers I will check out additionally. 2015-03-30 06:14:21 i think it also has the non-standaard sys/ctypes.h somehting 2015-03-30 06:14:32 for u_char typedef etc 2015-03-30 06:14:50 when it comes to ceph 2015-03-30 06:14:55 i started to look at that too 2015-03-30 06:15:08 i got an initial APKBUILD from Jean-Scotch 2015-03-30 06:15:14 <_mjones> It's conceivable that the file count difference is because AL is at 3.2.2 and Debian 7.8 is on 3.1.7, but it's a big difference. 2015-03-30 06:15:39 have you compared the upstream source packages? 2015-03-30 06:15:46 <_mjones> Did he get something working? great. I hadn't gotten a clean compile. 2015-03-30 06:15:55 <_mjones> No I hadn't gotten to the point of comparing the upstream yet. 2015-03-30 06:16:09 i bet its an upstream difference 2015-03-30 06:16:15 no, it does not compile yet 2015-03-30 06:16:21 <_mjones> My comment was about how any issues might have gone unnoticed because no other package in aports relies on xfsprogs. 2015-03-30 06:16:51 hum 2015-03-30 06:17:04 i dont even have xfsprogs deps here in my ceph build 2015-03-30 06:17:19 --without-libxfs \ 2015-03-30 06:17:57 still does not build 2015-03-30 06:18:10 ./include/buffer.h:224:74: error: '__STRING' was not declared in this scope 2015-03-30 06:18:10 assert(_raw); 2015-03-30 06:18:12 <_mjones> I haven't gotten it to finish compiling, but gotten through all of it so far with just --without-tcmalloc. 2015-03-30 06:18:22 <_mjones> I fixed that one. 2015-03-30 06:18:26 cool 2015-03-30 06:18:35 <_mjones> #define __STRING(x) #x 2015-03-30 06:18:42 i have a fix for libresolv 2015-03-30 06:18:49 <_mjones> It held me up for a couple of days, but it shouldn't have. 2015-03-30 06:19:17 <_mjones> I have that too, but only fixed it in configure so far. dalias says it should just work converting from res_nquery to res_query. 2015-03-30 06:19:18 libresolv-hack.patch: http://sprunge.us/OdcE 2015-03-30 06:19:28 <_mjones> Let's see where I left it -- I've forgotten. 2015-03-30 06:19:44 $ sprunge < stateless-resolver.patch 2015-03-30 06:19:45 http://sprunge.us/cjLE 2015-03-30 06:20:23 i also consulted dalias about that res_nquery stuff 2015-03-30 06:20:31 <_mjones> from erasure-code/jerasure/ErasureCodePluginJerasure.cc:19: /usr/include/c++/4.9.2/tr1/hashtable_policy.h: In member function 'std::pair std::tr1::__detail::_Prime_rehash_policy::_M_need_rehash(std::size_t, std::size_t, std::size_t) const': /usr/include/c++/4.9.2/tr1/hashtable_policy.h:464:24: error: expected unqualified-id before '(' token __min_bkts = std::max(__min_bkts, _M_gro 2015-03-30 06:20:41 <_mjones> that's awfully far into the build. 2015-03-30 06:22:59 what is __STRING? 2015-03-30 06:23:07 glibc localization stuff? 2015-03-30 06:23:20 <_mjones> It's just a macro. I'll look at my notes. 2015-03-30 06:23:36 <_mjones> http://sprunge.us/cjLE has the src patch I haven't done yet. 2015-03-30 06:23:51 <_mjones> http://sprunge.us/OdcE seems wrong because res_query has only 5 arguments, not 6. 2015-03-30 06:24:08 oh 2015-03-30 06:24:10 yes 2015-03-30 06:25:44 <_mjones> I just added "#define __STRING(x) #x" to ceph's include/assert.h and it's fine. 2015-03-30 06:25:47 do you have a patch for backtrace? 2015-03-30 06:25:49 yeah 2015-03-30 06:25:55 i added that too now, thanks :) 2015-03-30 06:25:57 <_mjones> It needs cdefs.h though 2015-03-30 06:26:08 apk add bsd-compat-headers 2015-03-30 06:26:21 <_mjones> yeah, I just realised that what you said would fix that. 2015-03-30 06:26:33 ./common/BackTrace.h:5:22: fatal error: execinfo.h: No such file or directory 2015-03-30 06:26:36 <_mjones> nothing for backtrace. Which is that? 2015-03-30 06:26:40 <_mjones> oh that, yes. 2015-03-30 06:26:42 <_mjones> one second. 2015-03-30 06:26:54 <_mjones> actually that one was ugly I think. 2015-03-30 06:27:16 <_mjones> I already have bsd-compat-headers. 2015-03-30 06:27:35 <_mjones> The issue was making sure it was included and turning off -Werror in the build so it wouldn't scream in every line: 2015-03-30 06:27:51 *nod* 2015-03-30 06:28:21 <_mjones> One second on the backtrace.h 2015-03-30 06:29:30 i suppose AC_CHECK_HEADERS should check for execinfo.h 2015-03-30 06:30:29 <_mjones> Yes, that was the one that's super-ugly. I was hoping to make a better one. 2015-03-30 06:31:57 <_mjones> It was so ugly I didn't keep a diff. Let me check something. 2015-03-30 06:32:08 :) 2015-03-30 06:34:38 <_mjones> Looks like I pasted straight out of Debian's execinfo.h into backtrace.h, and added "#include " and "#include 2015-03-30 06:34:57 <_mjones> There's also another function, I'm checking on where that came from. Might have already been in backtrace.h 2015-03-30 06:35:13 basically 2015-03-30 06:35:20 musl does not have any backtrace support 2015-03-30 06:35:30 so i think it could just segfault 2015-03-30 06:35:45 and you'd need use the core dump to do backtrace 2015-03-30 06:35:52 its only for debugging purposes 2015-03-30 06:36:08 so i think we can just lobotomize it out 2015-03-30 06:36:12 <_mjones> I didn't realise that. 2015-03-30 06:36:15 <_mjones> :P 2015-03-30 06:36:25 <_mjones> Anyway, the other thing I pulled in from somewhere was this:L 2015-03-30 06:36:27 <_mjones> #ifndef __nonnull # if LIBBSD_GCC_VERSION >= 0x0302 # define __nonnull(x) __attribute__((__nonnull__(x))) # else # define __nonnull(x) # endif #endif 2015-03-30 06:37:03 <_mjones> My goal was to get a compile and see if it was doable before I started cleaning it up and then cutting patches. 2015-03-30 06:38:45 <_mjones> at the bottom of memorymodel.cc you have to kill the macros and make mallinfo() unimplemented, because it is on musl 2015-03-30 06:38:56 $ sprunge < backtrace.patch 2015-03-30 06:38:56 http://sprunge.us/AcgF 2015-03-30 06:39:05 somthign like that should make it compile at leasts 2015-03-30 06:39:32 <_mjones> There's a compile for the monitor that will never complete unless you have >640mb ram. I bumped the VM from 640mb to 4GB and it works. 2015-03-30 06:39:41 <_mjones> That's as far as I got. 2015-03-30 06:40:20 aha 2015-03-30 06:40:27 here comes loff_t 2015-03-30 06:40:30 ./common/safe_io.h:44:34: error: 'loff_t' has not been declared 2015-03-30 06:41:48 <_mjones> Best fixed at: 2015-03-30 06:42:18 ncdev-edge-x86_64:~/aports/main$ grep loff_t */*.patch 2015-03-30 06:42:18 dosfstools/musl-fixes.patch:-#include /* for loff_t */ 2015-03-30 06:42:30 dosfstools/musl-fixes.patch:+#include /* for loff_t */ 2015-03-30 06:44:28 i suppose replace sys/types.h with fcntl.h in ./common/safe_io.h 2015-03-30 06:44:35 <_mjones> I put it in cephfs/libcephfs.h but it needs to go in include/types.h with #ifdef musl (not really) 2015-03-30 06:44:58 <_mjones> good catch. I didn't check other ports for that one. 2015-03-30 06:45:26 its even a bug for it i nglibc 2015-03-30 06:45:31 in glibc* 2015-03-30 06:45:33 https://sourceware.org/bugzilla/show_bug.cgi?id=14553 2015-03-30 06:46:03 <_mjones> lol. 2015-03-30 06:46:15 <_mjones> I think that's all I have. 2015-03-30 06:46:52 <_mjones> I was going to pick it up again in the morning (I'm in the states). 2015-03-30 06:46:59 ok 2015-03-30 06:47:05 better get some sleep then ;) 2015-03-30 06:47:14 now i bumped into mallinfo 2015-03-30 06:47:39 <_mjones> just prune it, musl doesn't have it. 2015-03-30 06:48:18 <_mjones> After you do that you'll be caught up, but with much cleaner patches. 2015-03-30 06:49:09 that only needed a s/__linux__/__GLIBC__/ 2015-03-30 06:49:22 <_mjones> hashtable_policy.h:464:24: error: expected unqualified-id before '(' token 2015-03-30 06:49:35 <_mjones> re:__GLIBC__: yes, much cleaner. 2015-03-30 06:49:52 <_mjones> unqualified-id, hmm. 2015-03-30 06:49:58 or configure.ac check for mallinfo 2015-03-30 06:50:14 and #ifdef HAVE_MALLINFO 2015-03-30 06:50:57 <_mjones> from ./common/debug.h:18, from erasure-code/jerasure/ErasureCodePluginJerasure.cc:19: 2015-03-30 06:51:06 <_mjones> I didn't see it in configure. 2015-03-30 06:52:07 no, we'll have to add it 2015-03-30 06:52:18 <_mjones> ok, this new error is stemming from my ugly hacks in BackTrace.h 2015-03-30 06:52:47 would be nice to upstream patches 2015-03-30 06:52:47 <_mjones> since backtrace is being pruned, this shouldn't apply 2015-03-30 06:52:50 as many as possible 2015-03-30 06:52:56 <_mjones> @ncopa: that was my plan. 2015-03-30 06:53:13 In file included from libcephfs.cc:22:0: 2015-03-30 06:53:13 include/cephfs/libcephfs.h:41:3: error: #error libceph: must define __USE_FILE_OFFSET64 or readdir results will be corrupted 2015-03-30 06:53:34 <_mjones> I did get a clue from sage weil the project bigwig, but generally the #ceph people on OFTC don't talk about compiling or libc support. 2015-03-30 06:53:53 <_mjones> right, I hacked that one. Let me get it. 2015-03-30 06:54:49 <_mjones> #if !defined(__FreeBSD__) && !defined(__APPLE__) && !defined(__USE_FILE_OFFSET64) /* # error libceph: must define __USE_FILE_OFFSET64 or readdir results will be corrupted */ /* 2015-03-25 mjones */ #define __USE_FILE_OFFSET64 1 #endif 2015-03-30 06:55:01 <_mjones> That can go in configure.ac 2015-03-30 06:55:04 :) 2015-03-30 06:55:30 yes 2015-03-30 06:55:32 <_mjones> Hey, this was my first pass with it. ;) 2015-03-30 06:55:42 configure should test it properly 2015-03-30 06:56:42 <_mjones> Most of this stuff should upstream easily, as long as upstream is receptive. 2015-03-30 06:59:24 <_mjones> Oh, I didn't even see that the loff_t bug was filed by Rich Felker. haha, figures. 2015-03-30 06:59:35 :) 2015-03-30 06:59:46 he has lots of those 2015-03-30 07:00:11 <_mjones> I know, I followed a bunch of links once. 2015-03-30 07:00:33 <_mjones> They're less memorable before they've bitten you individually. 2015-03-30 07:00:51 yup 2015-03-30 07:01:32 msg/async/AsyncMessenger.cc:313:28: error: 'sched_getcpu' was not declared in this scope 2015-03-30 07:01:32 cpuid = sched_getcpu(); 2015-03-30 07:01:36 this is weird 2015-03-30 07:01:50 because it should be enough with #include 2015-03-30 07:01:51 <_mjones> ok, this is further than I got. You're off the map now. 2015-03-30 07:01:54 which it appears to do 2015-03-30 07:02:21 _GNU_SOURCE and sched.h should do it 2015-03-30 07:02:24 but it doesnt... 2015-03-30 07:02:25 hum 2015-03-30 07:02:59 <_mjones> cplusplus wrapper thing maybe? 2015-03-30 07:03:04 oh 2015-03-30 07:03:07 that might be 2015-03-30 07:03:45 <_mjones> #ifdef __cplusplus extern "C" { #endif 2015-03-30 07:03:50 <_mjones> I 2015-03-30 07:04:12 <_mjones> I'm even less sensible in c++ than c, so I don't know. 2015-03-30 07:04:20 <_mjones> But the boilerplate seems ok. 2015-03-30 07:06:04 <_mjones> What version are you using? I have the git tree but I was working on 0.87.2 and I don't have msg/async/ 2015-03-30 07:06:32 <_mjones> Yeah, the newer version has it. 2015-03-30 07:07:19 sched_getcpu() is glibc-specific. 2015-03-30 07:07:22 hum 2015-03-30 07:07:22 ok 2015-03-30 07:08:20 <_mjones> It seems to be missing from HEAD now. 2015-03-30 07:08:45 <_mjones> That function isn't in that file in the master branch head. 2015-03-30 07:08:55 you have a link to git? 2015-03-30 07:09:11 https://github.com/ceph/ceph 2015-03-30 07:10:41 damn... i need containers for ethercalc and etherpad. nobody created these packages, right? :) 2015-03-30 07:10:49 <_mjones> https://github.com/ceph/ceph/blob/master/src/msg/async/AsyncMessenger.cc#L313 2015-03-30 07:11:42 <_mjones> This is funny. I had both an older and a newer version than you, and neither of my versions had that function. 2015-03-30 07:11:59 <_mjones> This code changes a lot. 2015-03-30 07:12:13 yeah 2015-03-30 07:12:21 wanted to see how they solved it in git 2015-03-30 07:12:24 and then backport 2015-03-30 07:12:29 maybe i should use the current git first 2015-03-30 07:13:07 https://github.com/ceph/ceph/commit/3d8c09c27f54a5cde774d32e9e17845ee99c3473 2015-03-30 07:13:28 <_mjones> qhat version were you using? The download link on http://ceph.com/resources/downloads/ is still set to 0.87.1, which I was using for the port. 2015-03-30 07:13:38 ceph-0.93 2015-03-30 07:13:53 found here: http://eu.ceph.com/download/ 2015-03-30 07:14:56 <_mjones> Wow. I know 0.93 is in the works, but still. weird. 2015-03-30 07:15:27 <_mjones> Bad comment because they don't say why the switch in strategy. 2015-03-30 07:16:03 yup 2015-03-30 07:16:34 and it fails 2015-03-30 07:16:36 msg/async/AsyncMessenger.cc:303:37: error: 'set_affinity' was not declared in this scope 2015-03-30 07:16:36 if (cid >= 0 && set_affinity(cid)) { 2015-03-30 07:18:09 <_mjones> this is pthreads stuff, so don't look at me. ;) 2015-03-30 07:18:38 i think they have some c++ implementation there 2015-03-30 07:18:42 oki 2015-03-30 07:18:53 i should try fix mail from bugs.a.o 2015-03-30 07:19:07 i will have to go back to ceph later 2015-03-30 07:21:40 <_mjones> ok, ping me some time if you work on it. 2015-03-30 07:23:35 <_mjones> oh, ncopa: the issues are still preventing login to bugs, and were related to upgrade, correct? Some visitors were asking. 2015-03-30 07:23:49 i saw 2015-03-30 07:24:02 which is why i need fix it 2015-03-30 07:24:20 <_mjones> ok. ttyl. 2015-03-30 07:52:43 morning 2015-03-30 08:10:16 i really need too much complicated packages -.- 2015-03-30 08:10:28 kolab, etherpad, ethercalc... 2015-03-30 08:38:12 bugs.a.o password reste should work again 2015-03-30 08:38:45 speaking of forums 2015-03-30 08:38:55 i wonder if we could just use redmine forum 2015-03-30 08:46:47 if we do that, we shouldnt speak about a diff tracker anymore. it should be forbidden ;-) 2015-03-30 08:47:04 :) 2015-03-30 08:47:30 i havent really seen any alternative for issue tracking 2015-03-30 08:47:42 i also find the web ui pretty nice to use 2015-03-30 08:48:15 the only feature i am missing is being able to have one single issue connected to multiple git branches 2015-03-30 08:48:49 eg same issue affects various git branches and should not be closed til all branches are fixed 2015-03-30 08:48:55 and if ruby plays nice in aports, thats one issue less. 2015-03-30 08:49:10 I still vote for good old fluxbb :p 2015-03-30 08:49:26 fluxbb+1 2015-03-30 08:49:30 ruby will not play nice, but i think we can get it bearable 2015-03-30 08:49:52 ok 2015-03-30 08:50:02 fluxbb is probably better than redmine forums 2015-03-30 08:50:52 i am just not too happy that we will have to replace it once fluxbb-1.x support is dropped 2015-03-30 08:51:20 but we can probably use it for a couple of years before that happen 2015-03-30 08:52:00 we can probably migrate to that new solution. everything will js anyways :) 2015-03-30 08:52:33 js everywhere 2015-03-30 08:52:47 js everything 2015-03-30 08:53:01 let's move #alpine-devel to slack, irc is so 70s :p 2015-03-30 08:53:12 we could write abuild in js 2015-03-30 08:53:25 and APKBUILD would be written in js as well 2015-03-30 08:53:27 man whats happening... /o\ 2015-03-30 08:53:31 ncopa: you wanted to replace it after all 2015-03-30 08:53:42 jsbuild and jsbuildfile 2015-03-30 08:53:52 :) 2015-03-30 08:54:02 speaking of 2015-03-30 08:54:14 the major issue i have with any new format 2015-03-30 08:54:14 boostrapping in firefox 2015-03-30 08:54:38 with systemd implemented in js!!! 2015-03-30 08:54:52 ok the apkbuild format 2015-03-30 08:55:00 clandmeter: lol 2015-03-30 08:55:09 the major issue is versino format transformation 2015-03-30 08:55:39 how do we transform the apk version to upstream version string 2015-03-30 08:56:21 currently we do things like: _ver=${pkgver/_p/p} 2015-03-30 08:56:46 we also need a way to define 2 first elements in version string 2015-03-30 08:56:51 eg x.y from x.y.z 2015-03-30 08:57:22 we need a way to describe those transformations 2015-03-30 08:58:20 and we need the parser be able to parse it without need to run it in a shell 2015-03-30 11:08:23 barthalion: fluxbb plugin system is really messy. it wants to patch core files 2015-03-30 11:09:08 seems a strange decission, simple solution without a proper way to extend it. 2015-03-30 11:09:44 ncopa: i already dont like it, and i just intalled it :) 2015-03-30 11:10:09 i'm not surprised. its php... :) 2015-03-30 11:10:26 lol 2015-03-30 11:10:33 like that has anything to do with it. 2015-03-30 11:11:07 php makes it easy to mess up 2015-03-30 11:11:17 actually i dont care about function enhancement 2015-03-30 11:11:24 but there is one thing i do care about 2015-03-30 11:11:26 spam 2015-03-30 11:11:42 and fluxbb has no option by default it seems. 2015-03-30 11:11:58 it looks like what smf does 2015-03-30 11:12:00 true, arch forums has some local modification 2015-03-30 11:12:01 patch core files 2015-03-30 11:12:10 for antispam 2015-03-30 11:12:21 but what other options do we have? mybb? 2015-03-30 11:12:25 thats what i like about drupal 2015-03-30 11:12:45 its made to be pluggable 2015-03-30 11:15:05 barthalion: i seriously dont know any alternative. 2015-03-30 11:15:27 why the h*ll do they not include something like SFS 2015-03-30 11:16:18 like every forum is used in an intranet 2015-03-30 11:16:51 vanilla forums maybe? 2015-03-30 11:16:58 still php though 2015-03-30 11:17:08 it doesnt matter which lang 2015-03-30 11:17:15 astlong its properly coded 2015-03-30 11:17:46 ncopa: which was that pything you mentioned? 2015-03-30 11:18:02 flaskbb iirc 2015-03-30 11:18:05 flaskbb 2015-03-30 11:18:31 it was first hit by google when searching for lightweight forum software 2015-03-30 11:19:11 google: lightweight forum software plus antispam 2015-03-30 11:19:27 algitbot: do your job! 2015-03-30 11:21:01 algitbot: please behave! 2015-03-30 11:21:09 algitbot: or you will be fired... 2015-03-30 11:21:34 algitbot: hi 2015-03-30 11:21:58 google lightweight forum software plus antispam 2015-03-30 11:22:05 algitbot: lunch time is 30-60 minutes only 2015-03-30 11:22:07 algitbot: thanks... 2015-03-30 13:55:00 ncopa: i think we have a small issue with abuild here: http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#license 2015-03-30 13:55:14 or a small issue with documentation 2015-03-30 13:55:43 license field will split a licence based on a space 2015-03-30 13:56:07 MPL 2.0 will result in 2 types, MPL and 2.0 2015-03-30 13:57:09 ok? 2015-03-30 13:57:20 when is that an issue? 2015-03-30 13:57:30 always? 2015-03-30 13:57:46 when is the license fied used? 2015-03-30 13:58:00 when you want to view what the license is? 2015-03-30 13:58:20 ie http://pkgs.alpinelinux.org/packages 2015-03-30 13:58:26 check mopidy 2015-03-30 13:58:27 so if you read that license is MPL 2.0 2015-03-30 13:59:00 http://pkgs.alpinelinux.org/package/x86/mopidy 2015-03-30 13:59:05 i see 2015-03-30 13:59:31 apkindex only has a single license 2015-03-30 13:59:50 but the apk can hold multiple lines 2015-03-30 14:00:55 license = ASL 2015-03-30 14:00:55 license = 2.0 2015-03-30 14:01:08 yes 2015-03-30 14:01:34 i guess in case of dual/multiple license 2015-03-30 14:01:49 yes, i think so too 2015-03-30 14:02:07 i am still kinda confused 2015-03-30 14:02:20 why did we implement it that wy? 2015-03-30 14:02:24 way* 2015-03-30 14:02:35 i mean, why is not the license a single string 2015-03-30 14:02:45 i suppose i'll have to ask fabled tomorrow 2015-03-30 14:02:49 i dunno :) 2015-03-30 14:03:00 it just looked weird on pkgs.a.o 2015-03-30 14:03:07 yes, you are right 2015-03-30 14:03:24 it is wrong in index 2015-03-30 14:03:56 wrong in .PKGINFO that is 2015-03-30 14:05:52 ok, not that its important, but what happends if we later want to query that data and there is no delimiter? 2015-03-30 14:13:33 clandmeter: we should fix it 2015-03-30 14:13:42 i just need figure out how 2015-03-30 14:20:54 is github down again? 2015-03-30 14:21:52 not here 2015-03-30 14:22:03 https://github.com/baskerville/sxhkd/archive/0.5.5.tar.gz 2015-03-30 14:24:11 sigh 2015-03-30 14:24:18 i didnt mean to push that sxhkd 2015-03-30 14:24:46 oh it works now 2015-03-30 14:25:26 what is current the way the handle backtrace in musl? just remove it? 2015-03-30 14:25:45 backtrace(3)? 2015-03-30 14:25:46 yes 2015-03-30 14:25:49 yep 2015-03-30 14:26:06 that is a debugging feature that is only used if application craches 2015-03-30 14:26:26 we can use core dumps instead 2015-03-30 14:36:14 git hub issue 2015-03-30 20:20:50 ncopa: Lua 5.2.4 was released March 7 2015-03-30 20:21:01 you want to update package, or should I? 2015-03-30 23:38:10 ncopa: here's my latest repo: https://github.com/stef/aports/commits/master 2015-03-30 23:38:49 bumped s6 pkgs, added some console mail utils, and w3m 2015-03-31 05:10:27 <_mjones> has phabricator been suggested for future use? 2015-03-31 05:11:25 too heavy I guess 2015-03-31 05:15:24 <_mjones> oh. 2015-03-31 06:02:42 morning 2015-03-31 06:02:59 <_mjones> morgen. 2015-03-31 06:15:41 rpu3uO8PEVZZ: >>> cjdns: Checking sha512sums... 2015-03-31 06:15:41 cjdns-master.zip: FAILED 2015-03-31 06:15:41 sha512sum: WARNING: 1 of 1 computed checksums did NOT match 2015-03-31 06:16:56 oh its a 'master' git snapshot 2015-03-31 06:17:07 so its not same thing as when you packaged it 2015-03-31 06:18:06 rpu3uO8PEVZZ: i cannot use your cjdns 2015-03-31 06:18:25 your cjdns-0.1-r1 is not the same thing as the cjdns-0.1-r1 that i have here... 2015-03-31 06:18:39 we cannot have inconsistencies like that 2015-03-31 06:20:48 >>> skalibs: Checking sanity of /home/ncopa/aports/testing/skalibs/APKBUILD... 2015-03-31 06:20:48 >>> ERROR: skalibs: skalibs-2.3.3.0.tar.gz is missing in checksums 2015-03-31 06:22:16 >>> execline: Checking sanity of /home/ncopa/aports/testing/execline/APKBUILD... 2015-03-31 06:22:17 >>> ERROR: execline: execline-2.1.1.1.tar.gz is missing in checksums 2015-03-31 06:25:10 grc should probably noarch 2015-03-31 06:31:06 rpu3uO8PEVZZ: we have openjpeg in main which should be same version as openjpeg2 2015-03-31 06:53:34 morning 2015-03-31 06:53:48 did somebody forget to close a window? 2015-03-31 06:54:56 algitbot again? 2015-03-31 06:55:23 not again... 2015-03-31 07:01:09 <_mjones> ok, I'm turning in for the night. 2015-03-31 07:01:15 <_mjones> oops wrong window. 2015-03-31 07:01:37 sorry algitbot left it open 2015-03-31 07:03:46 lol 2015-03-31 07:49:14 algitbot: you are pretty stupid 2015-03-31 07:49:52 algitbot: you are fired! 2015-03-31 07:49:59 algitbot: hi 2015-03-31 07:51:51 he fainted 2015-03-31 07:52:14 rebooting brb 2015-03-31 07:56:00 i really start to like go programming :) 2015-03-31 07:56:37 next hackweek project: learning go by developing an irc bot including plugin system 8) 2015-03-31 07:56:50 cool 2015-03-31 07:56:52 mosez: i really must start getting myself into it 2015-03-31 07:57:01 maybe we could retire algitbot when you are done :) 2015-03-31 07:57:06 :D 2015-03-31 07:57:30 http://www.golangbootcamp.com/?from=@ interesting free ebook for the go basics :) 2015-03-31 07:57:33 next hackweek: learning Go by programming a poliycy/greylisting postfix daemon using consul as a backend 2015-03-31 07:57:56 does anyone work on packaging newest go btw? 2015-03-31 07:57:58 pretty sure i'm gonna fail though :P 2015-03-31 07:58:16 because quite soon next release will be out, and we're still on 1.3 iirc 2015-03-31 08:02:50 how soon? 2015-03-31 08:03:34 coredumb: interesting :) 2015-03-31 08:04:32 ncopa: no idea, but read some articles about incoming features 2015-03-31 08:05:47 we'll have a feature freeze soon 2015-03-31 08:05:49 for v3.2 2015-03-31 08:07:08 mosez: i'll start with greylisting only to start ^^ 2015-03-31 08:23:50 https://hackweek.suse.com/12/projects/760 :) 2015-03-31 08:25:12 i am starting to like docker 2015-03-31 08:25:43 i like it quite long. but a friend of mine went away from docker because of the poor ipv6 support :) 2015-03-31 08:26:37 i'd like run abuild in a docker 2015-03-31 08:26:42 i love it... 1 or 2 times a year we are getting a week of hacking at suse :D 2015-03-31 08:26:54 cool! 2015-03-31 08:27:02 ncopa: take my image... or the one from andyshinn :) 2015-03-31 08:29:28 i am using gliderlabs alpine image 2015-03-31 08:30:09 docker is all new to me 2015-03-31 08:30:15 yeah, that's the image of andyshinn :) 2015-03-31 08:30:30 how can i run a 32 bit alpine image on my 64 bit host? 2015-03-31 08:30:32 or that's a good base image i'm using as well 2015-03-31 08:33:54 32 bit is unsupported 2015-03-31 08:34:11 there are hacks here and there, but still you're on your own 2015-03-31 08:34:18 huh? i cannot run 32 bit container? 2015-03-31 08:34:20 :-( 2015-03-31 08:34:31 there goes your dream 2015-03-31 08:34:41 really 2015-03-31 08:34:51 its just to add --arch=x86 to apk... 2015-03-31 08:40:57 actually, i think for abuild we would ned set the personallity to 32 bit too 2015-03-31 10:15:55 is this still available, http://www.packagekit.org 2015-03-31 10:31:43 freaking git snapshots 2015-03-31 10:31:54 algitbot: kill the snapshots 2015-03-31 11:01:27 seems empty, http://pkgs.alpinelinux.org/contents?pkgname=libjpeg&arch=x86 2015-03-31 11:02:02 ok ,libjpeg-turbo 2015-03-31 11:02:16 nope 2015-03-31 11:08:02 found it, http://pkgs.alpinelinux.org/contents?pkgname=libjpeg-turbo-utils&arch=x86 2015-03-31 11:09:13 hmm, http://pkgs.alpinelinux.org/contents?pkgname=libjpeg-turbo-dev&arch=x86 2015-03-31 11:09:28 has, /usr/lib/libjpeg.so 2015-03-31 11:09:41 and /usr/lib/libturbojpeg.so 2015-03-31 11:11:21 ah,. ok, libjpeg-turbo 2015-03-31 11:27:59 hey, is someone here developer of Alpine for Raspberry PI? 2015-03-31 11:36:22 no 2015-03-31 11:36:25 or not yet 2015-03-31 11:36:32 looks like fabled is not here atm 2015-03-31 11:36:54 i will set up arm build box soonish - at which point i become arm developer i suppose... :) 2015-03-31 11:37:09 I have a question about Raspberry Pi kernel 2015-03-31 11:37:22 who should I look for? 2015-03-31 11:37:38 fabled 2015-03-31 11:37:41 but he is not here now 2015-03-31 11:37:54 ok 2015-03-31 11:37:56 you can ask anyway 2015-03-31 11:38:05 since you got me curious... 2015-03-31 11:38:24 I've just noticed that you guys do not use Grsecurity patch for Raspberry Pi kernel 2015-03-31 11:38:45 I wonder why 2015-03-31 11:39:55 I run a pet project utilizing same 3.14 kernel you guys use, but with Grsecurity patch - all running on rPi 2015-03-31 11:40:18 problem is I have lost mainstream support for 3.14 branch recently, so I hope we could cooporate 2015-03-31 11:41:38 ok 2015-03-31 11:41:49 i think there was some issue with grsecurity on arm at that point 2015-03-31 11:41:57 i dont know what the status is now 2015-03-31 11:47:03 ncopa, didn't want to publish the cjdns thingy at all. sorry. 2015-03-31 11:48:10 and re s6+dependencies, i suck. i had my own APKs but someone published his own, so i merged only the version numbers, without also the hashsums. damnit i suck. will try to remember this next time. 2015-03-31 11:50:00 damnit 2015-03-31 11:50:14 im no good for packaging. sorry 2015-03-31 11:52:08 np 2015-03-31 11:52:20 thanks anyways for sharing 2015-03-31 11:52:22 i asked for it 2015-03-31 11:54:27 https://github.com/peerchemist/Peerbox/blob/master/PKGBLD/core/linux-raspberrypi-grsec/PKGBUILD 2015-03-31 11:54:39 it works actually 2015-03-31 11:55:02 whats that patch? 2015-03-31 11:55:10 grsec-rpi-3.14.28.patch 2015-03-31 11:55:18 so its a special arm patch? 2015-03-31 11:55:27 yes, I got it from upstream 2015-03-31 11:55:28 special rpi patch 2015-03-31 11:55:28 https://grsecurity.net/~spender/ 2015-03-31 11:55:44 so the normal grsecurity patch will not work? 2015-03-31 11:56:05 well yes and no, no as it fails to properly patch 2015-03-31 11:56:15 *fails on patching that boot logo stuff 2015-03-31 11:56:21 otherwise it works 2015-03-31 11:56:34 I tried against your in-house rpi patch but that fails 2015-03-31 11:57:18 https://grsecurity.net/~spender/grsec-rpi-3.14.28.patch 2015-03-31 11:57:35 https://github.com/raspberrypi/linux/tree/rpi-3.14.y 2015-03-31 11:57:44 those two combined work 2015-03-31 11:57:46 http://dev.alpinelinux.org/~tteras/linux-3.14.y-rpi-20140923.patch 2015-03-31 11:57:52 yes, that is the one 2015-03-31 11:58:21 its pretty big 2015-03-31 11:58:34 i suppose that is why we dont use grsecurity patch 2015-03-31 11:59:12 well I want to help you get grsec patch working with it 2015-03-31 11:59:21 so I can use it too with my project 2015-03-31 11:59:40 get it working is not the problem 2015-03-31 11:59:44 problem is maintain it 2015-03-31 11:59:54 that is my problem :) 2015-03-31 12:00:17 i doubt we have resources to maintain it 2015-03-31 12:00:31 if we keep our rpi kernel synced to upstream, grsec people will maintain the grsec for us 2015-03-31 12:00:42 that worked for me over the last year 2015-03-31 12:01:36 rpi kernel synced with upstream 2015-03-31 12:01:42 yes 2015-03-31 12:01:45 which upstream? 2015-03-31 12:01:59 rpi-3.14.y? 2015-03-31 12:02:00 mainline LTS kernel (3.14) 2015-03-31 12:02:02 yes 2015-03-31 12:02:09 or linux-3.14.y 2015-03-31 12:02:32 rpi-3.14.y is linux-3.14.y but with rpi specific patches 2015-03-31 12:03:04 http://dev.alpinelinux.org/~tteras/linux-3.14.y-rpi-20140923.patch 2015-03-31 12:03:08 your patch is that 2015-03-31 12:03:18 to get mainline 3.14 working on rpi 2015-03-31 12:03:37 apparently we have linux-3.14.37 with rpi 2015-03-31 12:03:51 yep 2015-03-31 12:03:54 so we are already maintaining the rpi part 2015-03-31 12:04:05 yes :) that is why I am here 2015-03-31 12:04:06 so you mean that grsecurity will maintain the grsec part? 2015-03-31 12:04:11 yes 2015-03-31 12:04:15 ok 2015-03-31 12:04:32 we just need to figure out why your patch fails with grsec 2015-03-31 12:04:32 then this sounds like its doable 2015-03-31 12:05:32 i suppose we should talk with fabled 2015-03-31 12:06:39 mupdf is kinda huge 2015-03-31 12:06:42 $ apk info -s mupdf 2015-03-31 12:06:42 mupdf-1.6-r1 installed size: 2015-03-31 12:06:42 44429312 2015-03-31 12:06:46 44MB 2015-03-31 12:07:14 with a patch that makes it build the lib as shared instead of static 2015-03-31 12:07:21 the size becomes: 2015-03-31 12:07:27 >>> mupdf*: Package size: 8.9 MB 2015-03-31 12:08:29 btw, why dont you use systemd? 2015-03-31 12:08:45 oh no! you said the s* word! 2015-03-31 12:08:49 oh my 2015-03-31 12:08:55 :) 2015-03-31 12:09:04 well, systemd is glibc-only 2015-03-31 12:09:07 we use musl libc 2015-03-31 12:09:20 so we cannot use systemd even if we wanted 2015-03-31 12:09:38 oh I did not know, ok 2015-03-31 12:09:54 they have specifically stated that they dont make systemd portable 2015-03-31 12:10:29 if you want to port it to your system, then you need modify your system to become a glibc system 2015-03-31 12:10:41 I would port my project to Alpine due to Musl, but I really need systemd 2015-03-31 12:10:55 maybe I will just compile some stuff with Musl 2015-03-31 12:11:08 then you really need glibc and can just forget musl libc 2015-03-31 12:11:26 may i ask "I really need systemd" what for? 2015-03-31 12:11:51 beside it is really good stuff? all is streamlined and well integrated 2015-03-31 12:12:07 I use it mostly for socket activation and namespace isolation 2015-03-31 12:12:26 isn't it what uselessd provides? 2015-03-31 12:12:38 I did not check it out yet 2015-03-31 12:12:42 oki, the namespace isolation i hope to contribute to soon with minijail 2015-03-31 12:12:47 they stripped a lot of glibc-specific code and many features as well 2015-03-31 12:12:59 but I think namespaces and sockets are there (yet :p) 2015-03-31 12:13:04 so it can work with Musl? 2015-03-31 12:13:07 ah, too bad 2015-03-31 12:13:19 take my "yet" with grain of salt 2015-03-31 12:13:22 namespace isolation with .service files is awesome 2015-03-31 12:13:26 no idea what's their plans are 2015-03-31 12:14:36 hmmm, i guess namespace isolation with s6+minijail will be even more amazing ;) 2015-03-31 12:15:32 there's even a possibility of a rootless Xort 2015-03-31 12:17:09 well in my experience it is best to stay close to upstream 2015-03-31 12:17:25 unless you have excess free time :) 2015-03-31 12:19:26 depends on the complexity of the system, that you want to stay close to upstream-wise. i rather have control myself, than to depend on the whims of upstream. 2015-03-31 12:19:52 and with s6 and minijail you barely have such complexity-vendor-lockin issues 2015-03-31 12:21:44 I do not have resources to experiment so much with it 2015-03-31 12:23:06 firefox v37 out 2015-03-31 12:30:15 getting https://gmplib.org/gmp6.0.html in main is still has issues ? 2015-03-31 12:30:35 i see, isn't always a compromise between short-term and long-term costs? 2015-03-31 12:31:10 where the longterm costs are externalized on 3rd parties. 2015-03-31 13:01:46 ACTION just ordered a HP U2515H and sets his line break to 800 chars :D 2015-03-31 13:16:50 rpu3uO8PEVZZ, try http://tasktools.org/projects/clog.html, this could colourize practically any text file 2015-03-31 13:17:19 I use it for mosquitto, logs, stdout from progs... etc 2015-03-31 13:17:52 let me see, this grc thingy is a historical tool for me, dunno how many years i'm a happy user. thx vkrishn 2015-03-31 13:18:05 I think this app can modularize git/log apps...etc 2015-03-31 13:18:31 i only use it to colorize my /var/log/* in a sticky window 2015-03-31 13:18:37 and user can customize using their own conf file 2015-03-31 13:18:46 ok 2015-03-31 13:19:42 I even use it to colourize mbox files 2015-03-31 13:19:45 :) 2015-03-31 13:20:09 mboxgrep ... | clog 2015-03-31 13:20:11 hehe, why would someone want to do that? 2015-03-31 13:20:12 aaah. 2015-03-31 13:20:33 i use maildirs, and thus maildir-utils as above refered to as 'mu' 2015-03-31 13:20:42 for the task of "greping mails" 2015-03-31 13:20:55 :) 2015-03-31 13:23:41 you can look back in the irclog and will find that whenever someone advertises a wm i always ask does it do sticky windows? 2015-03-31 13:25:10 :) ok 2015-03-31 13:25:18 what app is this, http://www.djcbsoftware.nl/code/mu/mu4e-splitview.png 2015-03-31 13:26:11 looks like emacs 2015-03-31 13:26:19 with mu-bindings 2015-03-31 13:26:35 aah, mu4e i guess means mu 4 emacs? 2015-03-31 13:26:39 ok, got it, an emacs-based e-mail client based on mu (see screenshot). 2015-03-31 13:31:34 ncopa: reviewing your fixes to my patches in an extreme-facepalming session. hope to reduce the shame next time. 2015-03-31 13:38:42 rpu3uO8PEVZZ: i'll be more careful for what i ask for next time :) 2015-03-31 13:42:48 so i'll remove linux-headers, libtool and auto* from my buildenv. and i hope to remember to check the checksums 2015-03-31 13:43:02 next time before committing. sorry 2015-03-31 13:45:01 i should write a script to offset my limitations 2015-03-31 13:47:34 btw i have a minimal setup for a client-host based on openrc for the init stuff (like hw and such), and s6 for the service supervision. basically it's a /etc/services in a targz, would it make sense to package it up? or should i get of my lazy ass and get a wiki account on al.o? 2015-03-31 13:48:25 yeah, better wiki 2015-03-31 13:49:04 also i'm gonna try to go rootless in this, using minijail for all services in this package. basically it's acpid, cpufreqd, dnscrypt, unbound, udevd, syslog, glogd, havegd, fcron, ntpclient 2015-03-31 13:50:42 i have taken minijail from chromeos, and now it runs and compiles standalone, and runs as expected, even the tests 2015-03-31 14:13:56 damn, i only slept 2.5 hours, so im adding autocrap to depends... 2015-03-31 14:16:20 ncopa, bump apache2 to v2.4.12 pls, 2015-03-31 14:16:30 also there are lots of changes, http://www.apache.org/dist/httpd/CHANGES_2.4.12 2015-03-31 14:16:47 does this mean rebuilding pkgs with 'apache2-dev' ? 2015-03-31 14:16:55 possibly 2015-03-31 14:17:02 i hope not 2015-03-31 14:17:31 there are 6/4 main/testing pkgs 2015-03-31 14:17:40 ok 2015-03-31 14:30:31 ncopa: one question re: s6, you did not use the dependency versions that i provided, should i drop them in general or was there some problem with them? 2015-03-31 14:30:45 i didnt investigate what that was about 2015-03-31 14:30:53 i just noticed it didnt build 2015-03-31 14:30:59 due to missing checksums 2015-03-31 14:30:59 because of the versions? 2015-03-31 14:31:02 ah 2015-03-31 14:31:19 so i skiped those patches and ran abump - 2015-03-31 14:31:37 those are specified by the devs as minimum versions for the deps. so i don't drop them. 2015-03-31 14:31:37 i didnt notice that you had versioned deps 2015-03-31 14:31:43 no probs 2015-03-31 14:31:59 i'm working on the wikipage, and in the mean time make the packages more rounded. 2015-03-31 14:32:21 instead of writing the wiki and make people copy paste stuff, it's easier if it's packaged well i guess. 2015-03-31 14:32:28 yes 2015-03-31 14:32:39 better fix the packages 2015-03-31 14:32:52 on it 2015-03-31 14:32:54 vkrishn: could you please test if apache2 breaks any modules? 2015-03-31 14:33:07 ok 2015-03-31 14:49:26 apache-mod-auth-kerb 2015-03-31 14:49:26 apache-mod-auth-ntlm-winbind 2015-03-31 14:49:26 apache-mod-auth-radius 2015-03-31 14:49:26 apache-mod-fcgid 2015-03-31 14:49:32 seems ok 2015-03-31 14:54:10 error like, Hunk #1 succeeded at 227 (offset -1 lines). 2015-03-31 14:54:19 would mean, empty line ? 2015-03-31 14:59:02 I guess this is not an issue, 2015-03-31 14:59:04 #warning redirecting incorrect #include to 2015-03-31 15:03:02 so no packages needs rebuild? 2015-03-31 15:03:04 nice 2015-03-31 15:05:42 yabadabadoo, php builds on newer apache2 ok 2015-03-31 15:05:45 rock solid AL 2015-03-31 15:06:23 would it be ok to add 'grep' in makedepends to php ? 2015-03-31 15:06:46 why? 2015-03-31 15:06:49 does it build wrong? 2015-03-31 15:07:34 just saw the grep -E msg fly by 2015-03-31 15:07:45 did not capture the msgs though 2015-03-31 15:07:58 but build all ok 2015-03-31 15:08:16 more in the list, 2015-03-31 15:08:17 subversion 2015-03-31 15:08:17 xapian-bindings 2015-03-31 15:08:17 php-apcu 2015-03-31 15:08:17 php-mailparse 2015-03-31 15:08:18 php-memcache 2015-03-31 15:08:20 php-phalcon 2015-03-31 15:08:22 php-xcache 2015-03-31 15:08:27 test them ? 2015-03-31 15:08:47 i dont think its needed 2015-03-31 15:08:52 ok 2015-03-31 15:08:56 if one runs, then they probably all run 2015-03-31 15:09:16 also leaving testing/... 2015-03-31 15:09:20 ok, i think redmine should be ok now 2015-03-31 15:09:25 need to go 2015-03-31 15:09:28 see u 2015-03-31 15:10:39 yabadabadoo, should be read as, yaba-daba-doo ;) 2015-03-31 15:10:59 not ya-bad-a-bad-oo, 2015-03-31 15:14:00 thanks 2015-03-31 15:30:31 my s6 service scripts are here in the mean time, until i figure out what to package, and what do document on the wiki: https://github.com/stef/s6-services 2015-03-31 15:45:09 <_mjones> @ncopa, do you have the partial apkbuild for ceph that you and I were working on the other day? I'd like to rebase on yours. 2015-03-31 15:47:14 here's an initial draft for the wikipage: https://www.ctrlc.hu/~stef/s6.txt 2015-03-31 16:03:42 <_mjones> Has anyone seen anything weird or broken with alpine's openssl? I have conformance tests failing for io.js for no obvious reason and that's a bit concerning. 2015-03-31 16:05:45 <_mjones> also, looks like the existing openjdk (icedtea 2.5.4) fails with fortify on edge, now. :( 2015-03-31 16:05:57 <_mjones> s/fails/fails to compile/ 2015-03-31 19:01:38 some could be updated before AL v3.2 freeze 2015-03-31 19:01:40 freeswitch - 1.4.15 2015-03-31 19:01:40 re2c - 0.14.2 2015-03-31 19:01:40 samba - 4.2 2015-03-31 19:01:40 gmp - 6.0 2015-03-31 19:01:40 tcpdump 4.7.3 2015-03-31 19:02:03 libgd has new url, http://www.boutell.com/gd/ -> http://www.libgd.org/ 2015-03-31 21:33:09 ncopa, clandmeter i dont remeber the reason of why the forum is down 2015-03-31 22:55:24 alacerda: time 2015-03-31 22:55:37 will get it back 2015-03-31 22:59:11 algitbot: you are missing one commit 2015-03-31 22:59:33 algitbot: you suck 2015-03-31 23:02:25 algitbot: goodnite 2015-03-31 23:02:50 ncopa: teach algitbot some manners :)