2022-05-01 03:44:42 ikke: you broke the hook so bad nobody can merge anything :p 2022-05-01 03:46:35 remote: date: invalid date '%Y-%m-%d'; probably missing the + 2022-05-01 05:52:29 when someone has a chance please let me know what you think of ! 120524 2022-05-01 05:52:39 !120524 2022-05-01 05:52:58 one digit too many 2022-05-01 05:53:30 ahem... staring at the wrong window.. 2022-05-01 05:53:32 !33784 2022-05-01 06:05:40 smcavoy: it's still a draft? 2022-05-01 06:10:21 yes, just wanted to get feedback. specifically on the initd stuff 2022-05-01 06:10:31 ok 2022-05-01 06:40:58 any suggetions on how to use `setcap` with a python script? exabgp would need to listen on 179/tcp 2022-05-01 06:52:48 Using setcap is discouraged 2022-05-01 06:53:13 But the question is, what is the alternative 2022-05-01 06:54:47 can't do that, no 2022-05-01 07:40:18 oh well.. good enough for now then. 2022-05-01 08:32:21 I'm thinking about packaging https://github.com/pop-os/wallpapers. In the README, it states that some of the photos are CC-BY-SA-4.0, while others (from Unsplash) are in the public domain. What should be put in the license field (what to use to refer to public domain)? 2022-05-01 08:44:58 https://wiki.spdx.org/view/Legal_Team/Decisions/Dealing_with_Public_Domain_within_SPDX_Files 2022-05-01 09:01:55 ikke: I did find that page, although I can't see what the conclusion is. It seems like there isn't any single identifier, so what should be used? 2022-05-01 09:03:39 You'd have to look at what exactly is specified and see if there is something that matches 2022-05-01 09:04:01 There is no single public domain declaration 2022-05-01 09:29:54 ikke: any chance you can try to see if this: https://gitlab.alpinelinux.org/alpine/infra/docker/appstream-generator/-/merge_requests/4 fixes the permissions in the appstream container? 2022-05-01 11:36:51 pabloyoyoista: fyi, I also wanted to check if I can get rid of the umask that causes this in the first place 2022-05-01 12:43:56 psykose: fyi, your work-around will no longer work either :) 2022-05-01 12:44:08 aware :) 2022-05-01 12:46:10 It did check multiple commits, or at least, that was the intention 2022-05-01 12:46:31 a premature break prevented thatt 2022-05-01 13:18:58 smcavoy: start the program as root and let it bind and drop privileges 2022-05-01 13:19:06 (check that it actually does drop privileges) 2022-05-01 13:21:58 ktprogra1: i kinda don't understand the point of packaging a large set of wallpapers. why can't you just download it manually? i see two benefits of packaging: one, you get automatic updates. that doesn't seem to apply here; the last update to any file was 5 years ago, except for one new file added 10 months ago. two, you can easily install all of the wallpapers at the same 2022-05-01 13:21:59 time. i guess this could be good if you want to automatically rotate between images. is that common? 2022-05-01 13:22:24 and if so, why can't you just download them as a zip and extract to your home directory? 2022-05-01 13:23:14 Hello71: No, I'm only using them on one computer, I just thought that since there's already some wallpapers packaged (although those are part of DEs) it would be okay/good to package these. 2022-05-01 13:23:51 i think packaging one upstream default wallpaper is reasonable 2022-05-01 13:24:15 but packaging a whole set of them which you still have to manually configure just doesn't seem easier than manually downloading and configuring 2022-05-01 13:24:48 Hello71: Ok, makes sense. 2022-05-01 17:28:34 Hello71: that is what it does now. I will leave that as is. thanks 2022-05-01 19:48:24 If you have ideas to add, please do: https://wiki.alpinelinux.org/wiki/Release_Notes_for_Alpine_3.16.0 2022-05-01 21:24:15 ikke: Oh thanks, that would be very nice! :) 2022-05-01 21:27:23 There were some changes regarding sudo/doas 2022-05-01 21:30:19 only that the sudo was moved to community 2022-05-02 03:46:21 ikke: Should I mention the initramfs consoles switch fix? It's kinda major for getting OpenRC output on systems which use consoles that are compiled as modules (https://gitlab.alpinelinux.org/alpine/mkinitfs/-/merge_requests/99#note_229398), for example Virtualization.framework on macOS 2022-05-02 12:21:32 Is it possible to use a gpg key on a yubikey to sign packages with abuild? What would I put in the abuild.conf? 2022-05-02 12:22:24 abuild does not use GPG keys 2022-05-02 12:22:27 nor apk 2022-05-02 12:22:40 It uses rsa keys 2022-05-02 12:24:25 Thank you for the reply, forgive my ignorance. My understanding was that I might be able to use my RSA GPG keys on the yubikey for Alpine development, is this possible or am I getting confused here? 2022-05-02 12:26:15 ikke: how about ceph17 !33752 and postfix add redis support !33818 for 3.16 2022-05-02 12:27:59 not possible by default, you would need to modify abuild to sign it with yubi 2022-05-02 12:29:40 @panekj How might I go about doing that? 2022-05-02 12:30:33 the signatures are made with `openssl dgst -sign thekey.rsa` 2022-05-02 12:30:37 git clone https://gitlab.alpinelinux.org/alpine/abuild.git 2022-05-02 12:30:39 a gpg key is.. not that 2022-05-02 12:30:56 modify abuild-sign to use yubi 2022-05-02 12:34:22 also you need to have rsa key in piv 2022-05-02 12:35:27 I see, won't be possible with my setup then. Thanks for the responses 2022-05-02 23:18:29 #join alpine-linux 2022-05-03 10:42:46 I'm pushing to my gitlab repo and got "shallow update not allowed" message. Any idea ? 2022-05-03 10:43:13 Did you clone with --depth? 2022-05-03 10:44:23 looks like it - do I have to reclone or there is way out of it ? 2022-05-03 10:45:16 git fetch --unshallow 2022-05-03 10:45:51 thanks ! 2022-05-03 12:10:35 Shiz: I just pushed kernel 3.15.37 to v3.15 branch. with early microvm hvc0 console support for -virt kernel 2022-05-03 12:10:47 \o/ 2022-05-03 12:11:31 i also verified that it works 2022-05-03 12:12:03 $ tpaste < microvm.sh 2022-05-03 12:12:03 https://tpaste.us/QN04 2022-05-03 12:13:29 boots alpine in ~1 sec 2022-05-03 12:16:26 ncopa: Will there be a new ISO release with it? 2022-05-03 12:16:42 maybe 2022-05-03 12:17:05 v3.16 will have it as well 2022-05-03 12:17:31 6c9cbf27cbcbcc459c5453bbb3251a142884225d 2022-05-03 12:18:07 what I find somewhat challenging is that it works so different with arm 2022-05-03 12:22:24 ncopa: i'm not sure that's a sufficient test setup, as microvm by default still has an isa serial port 2022-05-03 12:23:01 afaik you want to at least pass ,isa-serial=off to -machine microvm 2022-05-03 12:23:46 but yay! 2022-05-03 13:23:59 too bad /etc/apk/world doesn't allow comments 2022-05-03 14:08:54 Piraty: It does allow comments, but they have to be on separate lines 2022-05-03 14:09:19 But does apk preserve them? 2022-05-03 14:10:33 apk fix choked on a comment line with # 2022-05-03 14:11:31 but https://git.alpinelinux.org/apk-tools/tree/src/database.c#n1288 2022-05-03 14:12:57 That sounds like it's for /etc/apk/protected_paths.d 2022-05-03 14:13:15 yeah i'm looking at it 2022-05-03 16:39:05 I now have a test matrix with 18 different combinations per iso. bios/uefi boot, ext4,xfs,btrfs root fs, sys, crypt/lvmsys 2022-05-03 16:39:21 wow, nice 2022-05-03 16:39:28 Where are you running it, your own machine? 2022-05-03 16:39:33 yes 2022-05-03 16:39:44 https://github.com/ncopa/alpine-installer-testsuite/blob/main/tests/test_sys_install.py 2022-05-03 16:39:45 qemu 2022-05-03 16:40:11 currently i only test with a single disk 2022-05-03 16:40:20 should probably add a dual disk setup too, 2022-05-03 16:40:49 and test nvme and maybe virtio 2022-05-03 16:41:45 shoudl work with both x86_64 and x86 isos 2022-05-03 16:42:07 shoudl work with either alpine-virt, alpine-extended and alpine-standard 2022-05-03 16:43:07 i plan to add tests for diskless and datadisks as well 2022-05-03 16:44:43 not sure which is most important: dualdisk/raid, nvme or diskless 2022-05-03 16:45:30 nvme is where we have most issues atm (that I'm aware of) 2022-05-03 16:45:38 ok 2022-05-03 16:45:58 i know there are bugs or missing features for crypt combinations also 2022-05-03 16:46:19 i'd like to be able to typ type: 'cryptsys', simlar to how we do 'lvmsys' 2022-05-03 16:46:26 to avoid the second question 2022-05-03 16:53:02 would be nice if we could run them from gitlab-ci when creating release tags (for release candidates) 2022-05-03 16:57:49 Yes, certainly 2022-05-03 17:07:41 I think time-wise it will be after 3.16, but I can look in what we can deploy in Azure to help with this 2022-05-03 18:26:33 not sure why this build is failing https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/28360 but it should be good to go ikke 2022-05-03 18:28:02 11233 commits behind the target branch 2022-05-03 18:28:05 please rebase :) 2022-05-03 18:29:29 ikke: re: this https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/28671#note_233215 2022-05-03 18:29:35 I used to use cura without building that 2022-05-03 18:29:51 can we just put it in as a recipe that users have to "abuild -r" if they want it? 2022-05-03 18:30:43 Id rather get that merged sooner rather than later because Ive fixed up that MR multiple times because it became unmergeable because of changes to master 2022-05-03 18:31:24 nice, my maim MR builds properly 2022-05-03 18:31:26 heh 2022-05-03 18:48:25 We used to have non-free, but we decided to drop that (though, this is not even non-free) 2022-05-03 18:53:57 ok lmk about your decision. Im ok with either or 2022-05-03 18:56:02 anjan: if you want it merged quickly, I would first leave out that package, and then we can think about what to do with that later 2022-05-03 18:56:26 ok sounds good. Ill do that tonight sometime 2022-05-03 18:56:33 Thanks for advising ikke 2022-05-03 20:17:19 should hare depend on binutils? 2022-05-03 20:17:33 binutils is part of alpine-sdk, but using hare alone requires binutils 2022-05-03 20:17:43 not sure if alpine-sdk stuff should be in dependencies or not 2022-05-03 20:19:37 ddevault: It doesn’t hurt to add it, so I‘d say just do it 2022-05-03 20:19:47 ack 2022-05-03 20:20:25 There are other packages that depend on binutils 2022-05-03 20:20:29 go for example 2022-05-04 07:39:00 do we have something in alpine that mimics `systemctl reboot --firmware-setup`? 2022-05-04 08:06:03 the pmOS guys may know something about that, since it's fairly common on android devices 2022-05-04 08:06:14 maybe there's some ACPI thing for x86 2022-05-04 08:06:31 I can recommend a good therapist for anyone who dives into ACPI to find out 2022-05-04 08:31:21 that's assuming they go back to the surface alive 2022-05-04 08:34:14 from what i understand this sets something in efivars 2022-05-04 08:35:59 IIRC it might just be a standard linux thing 2022-05-04 08:36:21 some syscall and then reboot using that syscall would do 2022-05-04 08:36:29 this is the same thing pmos guys also use 2022-05-04 09:04:35 https://github.com/systemd/systemd/blob/09541e49ebd17b41482e447dd8194942f39788c0/src/shared/efivars.c#L174 2022-05-04 09:05:39 Should I mention the initramfs consoles switch fix in the 3.16 release notes? It's kinda major for getting OpenRC output on systems which use consoles that are compiled as modules (https://gitlab.alpinelinux.org/alpine/mkinitfs/-/merge_requests/99#note_229398), for example Virtualization.framework on macOS. 2022-05-04 09:06:35 i guess most pmos devices do not use efi firmware, so it must do something else 2022-05-04 09:18:01 ktprogra1: feel free to add it. We will look later at what we will include in the release notes published on the website 2022-05-04 09:35:44 nmeum: I'd like to replace the patch for busybox udhcp retries/timeout 2022-05-04 09:36:20 we currently have 5 retries, with 3 seconds delay between (the helptext is currently wrong btw...) 2022-05-04 09:36:37 it means we wait totally 15 seconds 2022-05-04 09:36:43 which I think is good 2022-05-04 09:37:02 but I'd like to change it to 15 retries with 1 sec delay 2022-05-04 09:37:58 reason is that on ip link set dev eth0 up, it takes a second before the link get carrier, so the first try will always fail 2022-05-04 09:38:19 which means that it takes always minimum 3 seconds to get a dhcp lease 2022-05-04 09:39:52 the best thing would be to check for /sys/class/net/eth0/carrier to turn 1 before executing udhcpc (from ifupdown/ifupdown-ng) 2022-05-04 09:41:11 but that requires some sort of ugly hacks in busybox ifupdown *and* ifupdown-ng. 2022-05-04 09:42:16 trivial fix is to patch udhcpc to use default timeout to 1 sec instead of 3. It reduces the time to get a dhcp address from 3 to 1 sec 2022-05-04 09:55:38 ncopa: https://skarnet.org/software/bcnm/bcnm-waitif.html no ugly hacks needed 2022-05-04 09:55:50 notification is a good thing :) 2022-05-04 09:56:35 (I use this in conjunction to ifup when I have services that depend on the existence of an interface) 2022-05-04 09:58:22 ncopa: yes, I would ideally prefer something that could be upstreamed to busybox 2022-05-04 10:00:18 skarnet: yeah, that is the proper way to do it. not sure if its worth to add another dependency when we can tweak the udhcpc defaults 2022-05-04 10:01:21 the idea isn't to add another dependency but to use the same mechanism 2022-05-04 10:01:38 the code isn't huge or difficult, just annoying 2022-05-04 10:04:47 i could not find any way to do that using existing tools 2022-05-04 10:05:58 indeed there isn't, because if there were, I wouldn't have needed to write bcnm-waitif 2022-05-04 10:06:02 :thinking guy meme: 2022-05-04 10:06:57 but I guess you'd rather not try and upstream a 150-line patch to udhcpc 2022-05-04 10:07:07 so, yay more polling, I suppose 2022-05-04 10:08:08 we could add it to ifupdown-ng too 2022-05-04 10:08:44 ifmond will definitely have that mechanism 2022-05-04 11:51:21 ikke: Ok, thanks 2022-05-04 12:45:28 does networkmanager-openrc no longer get pulled in automatically for anyone else? 2022-05-04 12:55:17 Newbyte: Maybe open an issue and assign jirukta, they recently reworked NM 2022-05-04 13:03:39 done: #13778 2022-05-04 13:18:26 ncopa: I just noticed that setting VIRTIO_CONSOLE=y wasn't done for the lts kernels on edge, was that intended? 2022-05-04 13:19:59 ikke: Is this okay? I couldn't think of any other way to phrase it. https://wiki.alpinelinux.org/w/index.php?title=Release_Notes_for_Alpine_3.16.0#Fixed_not_showing_boot_output_using_consoles_with_drivers_compiled_as_modules 2022-05-04 14:05:31 ktprograms: yes. thats intended. I dont think we want that for lts kernels? 2022-05-04 14:23:55 ncopa: Isn't it enabled for linux-lts in 3.15? 2022-05-04 14:25:53 no. only for s390x 2022-05-04 14:42:19 ncopa: This seems to set it for all? https://gitlab.alpinelinux.org/alpine/aports/-/commit/a84cd27a2892cb9207a1fc507247fee053874a5e 2022-05-04 15:03:23 that's only virt 2022-05-04 15:24:34 yup, only for virt kernels 2022-05-04 15:24:57 my testsuite already found a couple of bugs in setup alpine 2022-05-04 15:25:16 mdadm on top of nvme woudl not pull in the nvme kernel module in initramfs 2022-05-04 15:26:23 nice1 2022-05-04 15:44:08 also luks on nvme 2022-05-04 15:44:26 https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10498 2022-05-04 15:45:13 it works as a special case for lvm 2022-05-04 15:47:06 right 2022-05-04 15:47:33 is there some way to find the underlying device under luks? 2022-05-04 15:47:54 i think these bugs are already mostly known. as i see it, the benefit of a comprehensive test suite is that these bugs can be fixed with less worry of breaking other cases 2022-05-04 15:48:30 yeah 2022-05-04 16:04:15 ncopa: 'cryptsetup status /dev/mapper/root' shows the device 2022-05-04 16:40:43 i vaguely recall this was discussed already in irc or issue tracker 2022-05-04 16:45:22 ah, right, setup-disk already gets it 2022-05-04 16:45:36 https://gitlab.alpinelinux.org/alpine/alpine-conf/-/blob/master/setup-disk.in#L631 2022-05-04 16:55:59 is lvm on crypt supposed to work? 2022-05-04 16:56:46 yes 2022-05-04 17:24:23 i now have a test matrix of 108 different install combinations. uefi/bios, 1 or 2 disks (raid1), virtio/ide/nvme, sys/lvmsys/cryptsys (not yet lvmcryptsys), ext4/xfs/btrfs 2022-05-04 17:26:44 and I run out of diskspace of my /tmp 2022-05-04 17:28:06 nice 2022-05-04 17:40:37 any chance openscad package will ever be updated? :-p 2022-05-04 17:42:56 what's missing? 2022-05-04 17:44:41 they seem to take their time between releases... 2022-05-04 17:44:51 there aren't releases just date tags 2022-05-04 17:45:31 i don't see a newer one than 2021.01 2022-05-04 17:45:35 ah, bleh 2022-05-04 17:45:43 indeed 2022-05-04 17:45:54 we have the same version arch does, but up to now it sure has had a few patches 2022-05-04 17:46:02 if you have some specific bug you know the fix for it's easy to backport thigns 2022-05-04 17:46:31 Their landing page mentions 2021.01 2022-05-04 17:46:52 as does https://github.com/openscad/openscad/releases 2022-05-04 17:47:02 yeah. checking with #openscad 2022-05-04 17:47:18 hopefully they can tag something soon because there's a lot of new functionality ppl are using that's not in that release 2022-05-04 17:47:26 hah, the classic 2022-05-04 17:47:54 perhaps they're working on the Perfect Release 2022-05-04 17:49:16 :) 2022-05-04 17:49:44 maybe i'll just try to learn how to use apk to build my own packagae 2022-05-04 18:46:22 ================================================ 90 passed, 18 skipped in 679.58s (0:11:19) ================================================= 2022-05-04 18:49:24 i-i think you can reduce the amount of === there 2022-05-04 18:49:51 also nice :) 2022-05-04 18:50:52 motion to make the number of = equals to the number of tests passed :D 2022-05-04 18:51:33 that would make it smaller :( 2022-05-04 18:51:44 you just have to write more tests 2022-05-04 18:51:59 what better motivation than equals signs 2022-05-04 18:52:38 that is less than 90 2022-05-04 18:54:46 I make it 96, but who's counting 2022-05-04 18:55:20 oh wow 2022-05-04 18:55:23 it's actually 97 2022-05-04 18:55:23 hah 2022-05-04 18:55:26 my eyes don't work 2022-05-04 20:25:47 i found this on my alpine system which is surely a packaging bug: 2022-05-04 20:25:47 -r-sr-x--- 1 root root 31496 Jan 17 2021 /usr/lib/pppd/2.4.9/pppoe.so 2022-05-04 20:25:56 there is no reason for a .so to have +s 2022-05-04 20:26:25 is it like libcap where it's directly invokable for some reason 2022-05-04 20:26:39 i vaguely recall that might mean something on glibc, but it doesn't on musl 2022-05-04 20:26:46 i don't think so 2022-05-04 20:27:14 anyway it may give privilege escalation from gid=root to uid=root since it just runs whatever bogus code happens to be at the start of .text 2022-05-04 20:29:54 actually i don't understand how this becomes +s 2022-05-04 20:30:10 some bogus install command i guess 2022-05-04 20:30:34 i'm confused because i thought abuild had some gating to make sure suids can't get created without explicitly listing them 2022-05-04 20:31:38 suid check is only for +x as well 2022-05-04 20:32:05 https://github.com/ppp-project/ppp/commit/c6869931feed37330a96715fecc0c2c75f3f5dc0 2022-05-04 20:32:07 (which is the one that can be bypassed with options=suid) 2022-05-04 20:32:12 and there is a pie check which can't 2022-05-04 20:32:12 i=$(find "$dir" -type f -perm /6000 \ 2022-05-04 20:32:12 # check so we dont have any suid root binaries that are not PIE 2022-05-04 20:32:12 | xargs scanelf --nobanner --etype ET_EXEC \ 2022-05-04 20:32:12 | sed "s|ET_EXEC $dir|\t|") 2022-05-04 20:32:14 error "Found non-PIE files that has SUID:" 2022-05-04 20:32:14 if [ -n "$i" ]; then 2022-05-04 20:32:16 echo "$i" 2022-05-04 20:32:16 e=1 2022-05-04 20:32:18 fi 2022-05-04 20:32:18 # test suid bit on executable 2022-05-04 20:32:20 i=$(find "$dir" \( -perm -u+s -o -perm -g+s \) -a -type f \ 2022-05-04 20:32:20 if ! options_has "suid"; then 2022-05-04 20:32:22 -a -perm -o+x) 2022-05-04 20:32:22 if [ -n "$i" ]; then 2022-05-04 20:32:24 error "Found executable files with SUID bit set:" 2022-05-04 20:32:26 echo "$i" 2022-05-04 20:32:28 e=1 2022-05-04 20:32:30 fi 2022-05-04 20:32:32 fi 2022-05-04 20:32:34 oops 2022-05-04 20:32:36 did not mean to paste that 2022-05-04 20:32:57 conveniently converted to dalias shell script format :p 2022-05-04 20:33:01 LMAO 2022-05-04 20:33:30 :-) 2022-05-04 20:34:57 i think the original reason for 4550 is lost to history 2022-05-04 20:36:16 is there any plausible way to get gid=0 without uid=0 anyways 2022-05-04 20:36:53 well, the commit it's from is.. 2004 2022-05-04 20:37:01 with just 'new plugin sent in by person' 2022-05-04 20:37:03 indeed lost.. 2022-05-04 20:38:48 :-p 2022-05-04 20:39:14 hello71, iirc glibc does something special with +s for shared libs 2022-05-04 20:39:31 maybe something awful like letting you LD_PRELOAD a +s .so into a +s main program if it's in a trusted path? 2022-05-04 20:50:24 i think that's right 2022-05-04 20:51:13 https://github.com/bminor/glibc/blob/master/elf/dl-load.c#L1946 2022-05-04 20:56:57 how did anyone think that was a good idea?? 2022-05-04 21:39:43 it seems reasonable to have some sort of flag for this, although reusing the setuid bit is... not great 2022-05-04 21:40:07 also probably you shouldn't be preloading anything into suid binaries anyways 2022-05-04 21:57:05 yeah, there is just utterly no way "arbitrary .so that's okay to preload into some suid binary" is "okay to preload into any suid binary" 2022-05-05 07:38:18 morning 2022-05-05 07:38:30 hello 2022-05-05 07:38:40 o/ 2022-05-05 07:39:24 i will try add tests for diskless today 2022-05-05 07:39:47 one question with sys installs though 2022-05-05 07:39:59 ncopa: would it make sense to create a repo on gitlab? 2022-05-05 07:40:19 currently, if you sepcify 2 disks in setup-disk, it will create an mdadm raid1 device 2022-05-05 07:40:35 if you add cryptsys, it will encrypt the raid device 2022-05-05 07:40:49 if you enable lvm it will add lvm on top of the raid device 2022-05-05 07:41:10 my question is: what should we do if you sepcify 3 or more disks? 2022-05-05 07:41:31 currently it will set up raid5, which I think might not be the best idea nowdays 2022-05-05 07:42:06 ikke: i gues it would. I can move it from github if you want 2022-05-05 08:13:16 morning ncopa 2022-05-05 09:03:05 3-4 disks raidz1, 5-7 disks raidz2, 8+ disks raidz3 ;) 2022-05-05 09:10:53 ncopa: any objection against !33801? 2022-05-05 12:19:28 ikke: it still forks 2022-05-05 12:20:14 ncopa: where? 2022-05-05 12:20:49 printf is a shell built-in, so it does not have to fork 2022-05-05 14:58:22 i assume the $() subshell still counts as a fork 2022-05-05 14:59:06 A subshell is not a fork 2022-05-05 14:59:55 Or does not _have_ to be a fork 2022-05-05 15:05:48 ok, ash does fork on $() 2022-05-05 15:06:15 I suppose what I was aiming at is no fork+exec 2022-05-05 15:21:18 fwiw printf is not always a builtin depending on the shell, see mksh for instance 2022-05-05 15:22:01 however posix strongly suggests using printf too... 2022-05-05 15:29:07 Yes, but we're mostly targetting ash here 2022-05-05 15:29:37 Or at least, something similar enough to ash 2022-05-05 18:12:10 hey, does anyone know how this person managed to get the json from this secfixes webpage? https://gitlab.alpinelinux.org/ariadne/secfixes-tracker/-/issues/6 2022-05-05 18:12:31 I tried curl https://security.alpinelinux.org/branch/3.11-main/maintainer-issues -H "Accept: application/json" 2022-05-05 18:13:19 https://paste.sr.ht/~anjan/f177bf3e108c3642800ff82f5c10dabdc0b24bf5 2022-05-05 18:13:20 application/json+ld 2022-05-05 18:13:41 Ariadne: 406 Not Acceptable 2022-05-05 18:14:16 Any URI on this tracker may be fetched with Accept: application/ld+json to get an equivalent JSON-LD document. 2022-05-05 18:14:19 sorry, ld+json 2022-05-05 18:14:19 :) 2022-05-05 18:15:08 Ariadne: I got 500 Internal Server Error 2022-05-05 18:15:13 cool 2022-05-05 18:15:35 unfortunately i don't have time to investigate at the moment 2022-05-05 18:16:04 TypeError: float() argument must be a string or a number, not 'NoneType' 2022-05-05 18:16:14 ok no worries Ariadne 2022-05-05 18:16:20 https://tpaste.us/QN0z 2022-05-05 18:16:25 stacktrace ^ 2022-05-05 18:18:17 interesting, Ill open an issue ikke 2022-05-05 18:20:08 i guess there are no secfixes then. and that means, no bugs 2022-05-05 18:20:21 or only bugs? you decide! 2022-05-05 18:20:46 ah, craftyguy found that cvss3_score might not be a float https://gitlab.alpinelinux.org/ariadne/secfixes-tracker/-/blob/master/secfixes_tracker/models.py#L25 2022-05-05 18:21:33 It's None 2022-05-05 18:21:43 Which the stracktrace shows 2022-05-05 18:21:48 It just needs to handle that 2022-05-05 18:21:53 There we go 2022-05-05 18:21:55 yeah just needs a guard for that 2022-05-05 18:21:59 before casting 2022-05-05 18:22:03 ok, Ill see if I can code and test something up 2022-05-05 18:22:14 Ill submit a bug report so this time isnt atleast wasted 2022-05-05 18:22:31 What should it return in that case? 2022-05-05 18:22:35 Just an empty string? 2022-05-05 18:23:20 Or null? 2022-05-05 18:24:14 or 0 2022-05-05 18:28:07 null seems appropriate if it's unknown, unless a score of 0 is valid / has meaning 2022-05-05 18:28:28 or rather, if 0 doens't have meaning in this context, then maybe 0 instead of null 2022-05-05 18:35:42 bugtracker here: https://gitlab.alpinelinux.org/ariadne/secfixes-tracker/-/issues/8 2022-05-05 18:36:05 sorry, I must be pinging the author a bunch even though they said they were too busy 2022-05-05 18:36:17 (cause it's in the url heh) 2022-05-05 18:38:51 I've created an MR 2022-05-05 18:39:54 thank you guys. I really appreciate it 2022-05-05 19:10:13 Hmm, somethings afoot with gitlab (after upgrade) 2022-05-05 19:12:06 still down 2022-05-05 19:12:51 yes 2022-05-05 19:13:08 what do you see? 2022-05-05 19:13:28 I see dead people 2022-05-05 19:13:40 (hopefully not because of gitlab) 2022-05-05 19:14:12 i wish i saw dead people 2022-05-05 19:14:16 teach me.. 2022-05-05 19:16:42 A migration that was expected to be finished but still active 2022-05-05 19:29:19 ok, progress 2022-05-05 19:30:30 ok, gitlab is starting again 2022-05-05 19:30:55 did it just take that long 2022-05-05 19:30:58 Nop 2022-05-05 19:31:03 I had to manually run that task 2022-05-05 19:31:07 rip 2022-05-05 19:31:39 It's up again 2022-05-06 00:36:47 Hello everyone 2022-05-06 00:37:31 hi! 2022-05-06 00:37:54 I'm struggling to make a diskless boot loads wifi driver 2022-05-06 00:38:10 I would appreciate any comments 2022-05-06 00:38:16 I'll describe what I have done so far 2022-05-06 00:41:02 I upgraded to the latest kernel in full system environment, installed linux-lts-dev, alpine-sdk etc..I built the wifi driver using make install and it's loaded successfully. 2022-05-06 00:41:20 I made setup-bootable usb drive then used update-kernel and I'm able to boot with new usb drive 2022-05-06 00:41:36 however the wifi driver is not loaded 2022-05-06 00:42:40 what driver is this anyways 2022-05-06 00:43:08 i have unsquashed modloop-lts and copied my full install files into it which are /lib/firmware, /lib/modules/KERNELVERSION, squashed it again and it's still not loaded 2022-05-06 00:43:14 88x2bu 2022-05-06 00:43:37 https://github.com/RinCat/RTL88x2BU-Linux-Driver 2022-05-06 00:44:46 the modloop-lts file has the files inside but when I boot as ramdisk the files don't exist in /lib/modules... 2022-05-06 00:45:36 sounds like it is mounting some other modloop 2022-05-06 00:46:22 seems like rtl88x2bu is packaged as -src in testing 2022-05-06 00:46:38 not sure how akms works with diskless though, my guess is not well 2022-05-06 00:47:31 I'm not using akms ^^' 2022-05-06 00:48:00 I want to load it with current kernel not as dynamic 2022-05-06 00:48:29 I'll check if somehow I messed up with modloop file maybe it loads backed up modloop or something 2022-05-06 01:09:32 I ended up with both installs not working somehow 2022-05-06 01:09:57 But the last thing I did was deleting apks folder, it has old apk kernel files and linux-firmware 2022-05-06 01:10:22 I think there is something old loaded and overwriting the files 2022-05-06 01:12:49 I put new linux-firmware,kernel apk files but i'm still unable to boot, I think there are some traces of old files or configurations, starting from scratch with more care 2022-05-06 05:53:08 How long till the secfixes with everything fixed is live at security.alpinelinux.org 2022-05-06 06:06:08 The job runs hourly 2022-05-06 06:24:00 ikke: I meant getting secfixes with the new commits with the fixes we found today 2022-05-06 06:24:39 https://gitlab.alpinelinux.org/ariadne/secfixes-tracker/-/commit/df97d10bd16169246337a2d3488e518311688084 2022-05-06 06:24:42 specifically that 2022-05-06 06:25:13 christ 2022-05-06 06:25:25 patience man 2022-05-06 06:27:32 oh. sorry. I thought it was automated and something was up. My bad. If it requires input on your end, take your time. 2022-05-06 06:29:22 also I didnt mean to tag you again. 2022-05-06 06:47:51 Yeah, there is no automatic deployment 2022-05-06 07:14:46 what's the latest approach to dealing with stupid python bullshit again 2022-05-06 07:21:14 python-installer 2022-05-06 07:31:52 thanks ikke 2022-05-06 11:09:28 clandmeter: any objection to merging !33891? 2022-05-06 12:53:10 seems like gitlab issue lists can no longer be viewed while logged out 2022-05-06 12:54:07 in a very annoying way too 2022-05-06 12:54:14 instead of just saying unauthorised it's a generic error 2022-05-06 12:54:23 and everything in network returns 200 :p 2022-05-06 12:54:29 :/ 2022-05-06 13:10:06 also the 'lint' stage always returns green in gitlab ci regardless of output 2022-05-06 13:10:49 F.U.N. 2022-05-06 13:12:18 well the error is "Cannot return null for non-nullable field Issue.author" x 20 2022-05-06 13:16:41 Where do you see that error? 2022-05-06 13:17:41 Ah, found it 2022-05-06 13:17:46 graphql 2022-05-06 13:42:40 ikke: could you merge !33962 2022-05-06 14:27:15 using update-kernel on full system install to a usb drive for diskless use doesn't copy the drivers, I'm struggling since yesterday to make the wifi driver work in diskless usb environment. 2022-05-06 14:27:57 Editing modloop-lts manually to add all full system /lib/modules and lib/firmware gives same result 2022-05-06 14:28:47 after booting diskless, /lib/modules doesn't reflect what's in modloop-lts file.. files aren't there not only not loaded 2022-05-06 14:32:48 I started from scartch, installed full alpine in a system 2022-05-06 14:32:48 added 3.16/edge repositories, upgraded packages, install linux-lts/linux-lts-dev , alpine-sdk and everything necessary to be able to 'make install' wifi driver 2022-05-06 14:32:49 setup-bootable from iso image to usb, then update-kernel on usb/boot 2022-05-06 14:32:52 Am I missing something? 2022-05-06 15:00:32 Its working now! 2022-05-06 15:00:45 megactl: What was the issue? 2022-05-06 15:02:20 Seems when I edited modloop-lts I compressed it wrongly, because the only way I found on how to do it is from the wiki 2022-05-06 15:02:21 https://wiki.alpinelinux.org/wiki/DIY_Fully_working_Alpine_Linux_for_Allwinner_and_Other_ARM_SOCs 2022-05-06 15:02:50 I was using this 2022-05-06 15:02:50 mksquashfs squashfs-temp/ modloop-sunxi -b 1048576 -comp xz -Xdict-size 100% 2022-05-06 15:03:02 but when I did it as written in update-kernel script 2022-05-06 15:04:23 https://github.com/alpinelinux/alpine-conf/blob/4b89713f216d0c07a84aaadfce4e503d115b3a34/update-kernel.in#L336 2022-05-06 15:04:23 mksquashfs $MODLOOP "$STAGING/$MODIMG" -comp xz -exit-on-error $mksfs 2022-05-06 15:04:24 where $mksfs = -Xbcj x86 2022-05-06 15:04:25 it worked 2022-05-06 15:05:40 also there is no way to add kernel drivers using update-kernel options, -e seems only for /lib/firmware files , not /lib/modules/$KERNELVERSION... drivers 2022-05-06 15:58:27 Will restart gitlab to update to a new patch release 2022-05-07 10:32:52 ncopa: the busybox dhcp patch you commited actualyl made the situation worse for me, i.e. I often don't get a dhcp lease within the new time frame because the link isn't up yet 2022-05-07 10:33:51 I would really suggest that we solve this properly in ifupdown-ng (i.e. analog to bcnm-waitif) instead of using a timeout mechanism in busybox directly 2022-05-07 11:58:19 I noticed that if you have multiple repos configured in /etc/apk/repositories and apk fail to fetch the index from one of them it still tries to fetch packages from it and doesn't seem to fall back on the others when it fail to do so 2022-05-07 11:59:00 omni: I suppose because it still has an index cached 2022-05-07 11:59:41 Note that /etc/apk/repositories is not really meant to specify fallback repositories 2022-05-07 12:00:14 aha 2022-05-07 12:03:05 As repositories don't really have an identity, it would be difficult to do that without resulting into gotchas 2022-05-07 12:51:39 anybody else that have notices zfs zvol related instabilies since this weeks kernel update in v3.15? 2022-05-07 12:53:38 more specifically nginx reverse proxy with /var/lib/nginx/tmp on a zvol seems broken 2022-05-07 13:11:58 ahh https://github.com/openzfs/zfs/issues/12971#issuecomment-1119169481 2022-05-07 13:20:40 Seems like linux stable has regressions quite regularly 2022-05-07 14:38:25 I just noticed x264enc is missing in gstreamer, looks like there's a flag required in gst-plugins-ugly to enable plugins that contain GPL code 2022-05-07 14:38:28 any reason that flag is missing? 2022-05-07 14:43:40 maybe oversight? 2022-05-07 14:43:50 what is the flag? 2022-05-07 14:45:19 -Dgpl=enabled 2022-05-07 14:45:52 looking at the history there were rebuilds for x264, so it worked once 2022-05-07 14:46:41 probably also an issue on -plugins-bad and -plugins-good 2022-05-07 14:56:30 i noticed that error sometimes but also sometimes it just worked, which was strange 2022-05-07 15:01:09 i enabled it anyway, with -Dgpl=enabled, since we're licencing it as gpl, so 2022-05-07 15:01:19 i did disabled the cd/dvd ones it also enables though 2022-05-07 15:01:29 should be there in a few minutes :) 2022-05-07 15:08:14 looks like the zfs problem was handled in the zfs 2.1.3 changes for Linux 5.16 compat, maybe we shall bump zfs to 2.1.4 in stable? 2022-05-07 15:08:16 https://github.com/openzfs/zfs/pull/12975 2022-05-07 15:08:21 https://github.com/openzfs/zfs/releases/tag/zfs-2.1.3 2022-05-07 15:09:09 ah, that might fix it 2022-05-07 15:22:42 nr 2022-05-07 15:23:24 ACTION is sorry for accidentally sending junk... 2022-05-07 15:24:30 gst-plugins-good doesn't have a gpl flag, and -bad has it but it's mostly junk again, the only thing in there is x265enc 2022-05-07 15:24:40 i guess if you really want it i can add in x265enc too :) 2022-05-07 16:40:32 I don't think a lot of software uses x265enc and gstreamer, but it couldn't hurt 2022-05-07 17:57:23 psykose: Why are CD/DVD disabled? 2022-05-07 17:59:14 if you know of literally anything that uses 'cdiocddasrc' to read audio cds via gstreamer i will happily enable it 2022-05-07 17:59:28 it's probably two pieces of software in the whole universe :p 2022-05-07 18:34:58 Ahh, gotcha. Thought it was generic cd/DVD support which I definitely appreciate having. 2022-05-07 18:34:58 No idea what sources would use that codec, haha 2022-05-07 21:28:42 Hey, I followed: https://wiki.alpinelinux.org/wiki/Creating_patches to create a MR: 2022-05-07 21:30:12 https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/33853 2022-05-07 21:30:49 That was 2 May 2022 2022-05-07 21:32:17 Last night I was assigned a MR from J0WI (https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34071) for the same thing. 2022-05-07 21:32:54 Both MR's are effectively the same thing, so one can get merged, and the other deleted. 2022-05-07 21:33:49 The problem is I am not a reviewer and (AFAIK) do not have rights to push to master (and if I DO have those rights, you need to revoke them ) 2022-05-07 21:34:19 psykose will probably merge it once she sees it 2022-05-07 21:34:32 Any pointers on how to make sure a reviewer gets notified? 2022-05-07 21:35:19 nangel: just wait, if it doesn't get merged or any comment, complain here 2022-05-07 21:35:58 panekj: thanks. Part of this is I am trying to work on the devloper's manual, and "going through the whole 'do 1,2, then 3'" steps will help all of us who don't drop MR's every week. 2022-05-07 21:38:27 i mean.. shouldn't you have the rights, as a tsc member? 2022-05-07 21:40:32 i haven't noticed that lol, you have dev role 2022-05-07 21:53:14 psykose: I have messed up too many git repos to ever be allowed to touch aports. It is for your own good. 2022-05-07 21:56:06 Thanks for the input! 2022-05-08 04:39:06 ikke: also for your amusement, stunnel is as stuck as it always is on s390x 2022-05-08 04:39:28 Let's give it a stern look 2022-05-08 04:39:56 :3 2022-05-08 04:40:45 and it unstuckened itself when you did the usual 2022-05-08 04:40:48 Yes 2022-05-08 04:40:53 :3 2022-05-08 04:41:06 Normally I kill the entire process tree, but it will just get stuck again 2022-05-08 04:41:15 just killing the test and it continues ¯\_(ツ)_/¯ 2022-05-08 04:49:49 oh, you are also working on rsyslog 2022-05-08 04:50:51 are you doing 3.14? 2022-05-08 04:52:06 apparently 2022-05-08 04:53:34 ah 2022-05-08 04:53:35 didnt see 2022-05-08 04:53:37 yeah i did them all 2022-05-08 04:53:44 3.12 does not have anything resembling the same code that i can find 2022-05-08 04:53:49 ok 2022-05-08 05:05:18 ah no, it was just missing one file 2022-05-08 05:07:40 right 2022-05-08 05:07:50 https://github.com/rsyslog/rsyslog/commit/01feb44bf1d091ec45ff425a096a0169f290da4b#diff-e666705dfe5af14795afb1d8d1b0a2089d34cdba4f7e601a251210f645866893 is in v8.2004.0 2022-05-08 05:09:02 mhm 2022-05-08 05:09:05 just the contrib thing wasnt 2022-05-08 06:22:50 Hmm, for some reason x86 fails because the linker cannot find slapd-common.o 2022-05-08 06:22:52 https://gitlab.alpinelinux.org/alpine/aports/-/jobs/714405 2022-05-08 06:23:22 /usr/lib/gcc/i586-alpine-linux-musl/11.2.1/../../../../i586-alpine-linux-musl/bin/ld: cannot find slapd-common.o: No such file or directory 2022-05-08 06:24:19 $10 it's fixed with -j1 2022-05-08 06:24:20 perhaps concurrency issue / race condition? 2022-05-08 06:24:23 yea 2022-05-08 06:24:27 almost always is 2022-05-08 06:24:47 which means there is a dependency for a target missing 2022-05-08 06:24:53 yeah 2022-05-08 06:25:00 it's quite common for this configure soup 2022-05-08 06:25:07 we have quite a few of these as -j1 2022-05-08 06:25:30 because they default to -j1 unless you pass it manually, and i guess nobody realised for a really long time 2022-05-08 06:26:46 Just interesting it consistently happens on a single arch 2022-05-08 06:41:12 consistently until it passes 2022-05-08 06:42:20 trying to figure out how this works 2022-05-08 06:42:22 it uses libtool 2022-05-08 06:42:58 ldif-filter: ldif-filter.o $(XLIBS) 2022-05-08 06:42:58 $(LTLINK) -o $@ ldif-filter.o $(OBJS) $(LIBS) 2022-05-08 06:43:06 just add $(OBJS) to the top line 2022-05-08 06:43:08 it will pass every time :) 2022-05-08 06:43:21 tests/progs/Makefile.in 2022-05-08 06:43:41 Like this? 2022-05-08 06:43:43 slapd-tester: slapd-tester.o $(OBJS) $(XLIBS) 2022-05-08 06:43:59 Which is already the case 2022-05-08 06:44:17 it's not the case on the ldif-filter line 2022-05-08 06:44:30 ooh 2022-05-08 06:45:04 (i mean i can also be wrong, but this is where it fails, since it tries to build that but links to OBJS in the line below without having it as a dep) 2022-05-08 06:45:17 and it's the same thing from the logs 2022-05-08 06:45:29 sadly make does not automatically lift deps 2022-05-08 06:45:43 only the x: y line is the ordering 2022-05-08 07:22:31 pipeline is fixed now 2022-05-08 07:23:48 :) 2022-05-08 10:01:51 do we want to add skalibs and utmps to our bootstrap path or do we want to build busybox and openssh without utmps support when boostrapping? 2022-05-08 10:32:21 you probably don't need user accounting when bootstrapping, so I suggest keeping them out 2022-05-08 10:32:36 hmmm, openssh? in the bootstrap? 2022-05-08 10:33:12 if you need to ssh into an account while bootstrapping that's a different thing, but... why? 2022-05-08 10:34:20 Only reason I can think of is to support a default builder setup 2022-05-08 10:37:38 https://gitlab.alpinelinux.org/alpine/aports/-/blob/master/scripts/bootstrap.sh#L125 2022-05-08 10:37:42 I don't know, I haven't added it :D 2022-05-08 10:38:04 skarnet: right, only downside is that it makes the APKBUILD it bit more complex if you have to add if [ -z "$BOOTSTRAP" ]; then … everywhere 2022-05-08 10:40:02 nmeum: I have no horse either way in this race, skalibs and utmps are easy and quick to build so they wouldn't significantly impact the bootstrap, but as a matter of principle, *everything* that can be kept out of the bootstrap should be 2022-05-08 10:40:39 I tend to agree, I manually patched it out of my local tree for now 2022-05-08 10:41:41 trying once again to bootstrap rust for riscv64, so far I only had to patch the vendored libc create. let's so how far my motivation takes me this time… 2022-05-08 12:01:00 psykose: ping 2022-05-08 13:58:30 ok it looks good 2022-05-08 14:02:15 \o/ https://tpaste.us/jP05 2022-05-08 14:02:50 ikke: what would you need me to do to get a boostrapped rustc installed on the riscv64 builders, do you want to boostrap yourself or should I give you an apk? 2022-05-08 14:21:21 nmeum: How hard is to do the bootstrapping from source? 2022-05-08 14:21:38 I gather that you managed to build rust? 2022-05-08 14:22:09 yes, I boostrapped it using scripts/boostrap.sh with a few patches on top 2022-05-08 14:22:31 should be doable to do it form source 2022-05-08 14:22:51 Ok, if you could give the steps, I could give it a try 2022-05-08 14:23:40 you need !34151 and you need to slightly modify scripts/bootstrap.sh since it is broken as usual: https://tpaste.us/DBxa 2022-05-08 14:24:06 nmeum: Can't we push the fixes to bootstrap.sh? 2022-05-08 14:24:17 those are not fixes, just workarounds :p 2022-05-08 14:24:22 aha, ok 2022-05-08 14:24:24 I pushed fixes for busybox, curl and openssh though 2022-05-08 14:24:37 for example, openrc doesn't boostrap currently because the buildsystem was switched to meson 2022-05-08 14:24:42 but you don't need it for rustc anyhow 2022-05-08 14:24:51 Right, I saw that issue comming by 2022-05-08 14:25:28 I don't see an issue with just removing it from boostrap.sh locally for the rust boostrap 2022-05-08 14:25:38 nod 2022-05-08 14:26:03 so the steps are: apply !34151, use https://tpaste.us/DBxa and run `CBUILDROOT=/var/tmp/riscv64 ./bootstrap.sh riscv64` that should be it 2022-05-08 14:26:13 ok 2022-05-08 14:26:42 ok and you also need !34149 2022-05-08 14:31:37 do you have the time to start the boostrap today so we have results by tomorrow? 2022-05-08 14:33:09 yes, working on it 2022-05-08 14:33:48 ty 2022-05-08 14:38:46 nmeum: so is the idea that I first apply the patches, and once bootstrapped, we push it to make aports consistent? 2022-05-08 14:39:00 I mean, the rust patches 2022-05-08 14:39:18 yes 2022-05-08 14:39:21 ok 2022-05-08 14:39:23 I can also push the rust patches right now 2022-05-08 14:39:32 it's just that it won't build unles we bootstrap it first 2022-05-08 14:40:33 the rust patches are also trivial, just a few missing libc constants and an issue with an unsupported assembler directive 2022-05-08 14:43:29 bootstrape ich rust jetzt auch noch für s390x? 🤔 2022-05-08 14:43:55 oops wrong channel 2022-05-08 14:44:00 haha 2022-05-08 14:44:53 Should the llvm13 patch have a pkgrel bump? 2022-05-08 14:45:19 I assume I should build llvm13 2022-05-08 14:46:33 I didn't bump pkgel because it shouldn't change anything for non-boostrap builds and scripts/boostrap.sh should build it without a pkgrel bump 2022-05-08 14:46:48 ah, makes sense 2022-05-08 14:48:17 ok, running the bootstrap script now 2022-05-08 14:49:02 great, thanks 2022-05-08 14:49:14 let me know if you run into any issues :) 2022-05-08 14:59:31 nmeum: https://tpaste.us/BkW9 2022-05-08 15:14:31 Complete output https://tpaste.us/70DO 2022-05-08 15:32:12 When should g(cc|++)-riscv64 be built? 2022-05-08 15:45:25 > >>> gcc: Package is up to date 2022-05-08 15:45:35 did it not build gcc at all? 2022-05-08 15:46:12 No 2022-05-08 15:47:21 strange. you are running the bootstrap on x86_64, yes? 2022-05-08 15:48:05 No, on the builder, thought that was the idea 2022-05-08 15:48:23 I misunderstood 2022-05-08 15:48:59 ah, yeah you have to run it on x86_64 to cross-compile rust for riscv64 from x86_64 2022-05-08 15:49:11 sorry, should have stated that more clearly 2022-05-08 16:00:51 nmeum: and then we should install that package as -bootstrap and build rust? 2022-05-08 16:00:58 (Not saying it's finished) 2022-05-08 16:12:50 yes, that's what i was thinking 2022-05-08 16:13:08 ok 2022-05-08 16:13:24 (that should be how it was bootstrapped for other arches as well, at least that's my understanding) 2022-05-08 16:14:54 q 2022-05-08 16:34:36 nmeum: Running into this: https://tpaste.us/enrb 2022-05-08 16:38:45 huh, that looks wild 2022-05-08 16:39:31 yea 2022-05-08 16:39:50 It's in a docker container, but abuild -r just works fine 2022-05-08 16:39:55 x86_64 2022-05-08 16:40:12 I also boostrapped in a docker container 2022-05-08 16:42:13 It's executing CTARGET=riscv64 BOOTSTRAP=nobase APKBUILD=/home/buildozer/aports/main/binutils/APKBUILD abuild -r 2022-05-08 16:42:18 the bootstrap script should actually run apk with initdb and initialize the apk database beforehand 2022-05-08 16:44:04 Oh, I see, it checks if the directory already exists or not 2022-05-08 16:44:42 Fixed 2022-05-08 16:45:13 sweet 2022-05-08 16:51:00 ERROR: /usr/lib/libzstd.so.1.5.2: Could not find owner package 2022-05-08 16:51:24 gcc-pass2-riscv64 2022-05-08 16:56:09 should zstd be in the package list? 2022-05-08 16:57:52 No, it already happens before it reaches that part 2022-05-08 17:01:18 It happens when tracing dependencies btw 2022-05-08 17:03:07 hm, I didn't get that locally not sure 2022-05-08 17:05:54 might also be easier to just use my bootstraped version instead of boostraping again 2022-05-08 17:06:14 also: I am currently testing this with a few more complex aport and looks like it segfaults on some crates currently so need to address that as well 2022-05-08 17:13:26 Do you get what this is about: https://gitlab.alpinelinux.org/Newbyte/aports/-/jobs/714894#L971 2022-05-08 17:14:18 error: failed to run custom build command for `gmp-mpfr-sys v1.4.8` 2022-05-08 17:15:35 yeah, but I don't really understand the error. I was wondering if someone could shed some light 2022-05-08 17:15:45 I was thinking maybe I could try disabling LTO and all that for x86 2022-05-08 17:15:57 nmeum: Where is the package that you bootstrapped? 2022-05-08 17:16:25 https://dev.alpinelinux.org/~nmeum/edge/riscv64/ 2022-05-08 17:16:54 but as I said, there does seem to be some work left to do since it can't compile rust itself yet 2022-05-08 17:17:04 ok 2022-05-08 17:17:24 I will ping you again once I am done with that (: 2022-05-08 17:19:45 :) 2022-05-08 17:23:36 Newbyte: i think it's due to we are running x86 on x64 but I might be wrong 2022-05-08 17:24:14 from quick look at that project, there might be something leaking about x64 and it assumes it builds for x64 2022-05-08 17:24:46 yeah, that's correct 2022-05-08 17:25:03 So maybe there is a switch to force it to use x86 2022-05-08 17:28:01 panekj: oh, that makes sense. thanks! 2022-05-08 17:28:13 `--host=` should work 2022-05-08 17:28:31 but you would need to pass that somehow to gmp 2022-05-08 17:28:32 panekj: would I give abuild-meson that? 2022-05-08 17:28:35 oh, hm 2022-05-08 17:29:36 oh, it's one of the gnome apps? 2022-05-08 17:30:26 mhm, but since its just build.rs it shouldn't be a problem to patch for vendored dep 2022-05-08 17:34:12 panekj: how would I patch a vendored dependency? 2022-05-08 17:36:47 abuild unpack, go to builddir, dep should be somewhere around vendored/gmp-src probably, cp build.rs build.rs.orig, modify build.rs and diff it 2022-05-08 17:37:09 I'm wondering though if it wouldn't be better to just link it so system libs 2022-05-08 17:37:15 `gmp-dev` 2022-05-08 17:42:06 panekj: yeah but would abuild be able to patch that considering that the dependency is downloaded in the build stage? 2022-05-08 17:43:21 yes, if you provide prepare() function, which should be done for that aport 2022-05-08 17:44:09 seems like it does support linking to a shared library if you add some option: https://gitlab.com/tspiteri/gmp-mpfr-sys/-/issues/21 2022-05-08 17:44:23 CARGO_FEATURE_USE_SYSTEM_LIBS 2022-05-08 17:46:23 I'm already building your branch with that 2022-05-08 17:48:34 Yeah, I'm doing that locally right now as well 2022-05-08 17:48:47 if gdm doesn't suspend my system 2022-05-08 17:49:04 great project btw, tag is `0.6.0`, Cargo has `0.1.` 2022-05-08 17:49:16 s/0.1./0.1.0 2022-05-08 17:49:16 panekj meant to say: great project btw, tag is `0.6.0`, Cargo has `0.1.0` 2022-05-08 17:50:38 the build setup is definitely awkward in some places, but it's a nice music player 2022-05-08 17:53:47 https://tpaste.us/Pjdx 2022-05-08 17:54:13 but I would generally advise to patch out whatever meson is doing so it conforms to our standard 2022-05-08 17:55:17 missed `mpfr-dev` in makedepends :') 2022-05-08 17:55:39 was about to ask about that 2022-05-08 17:58:06 I'm guessing from names to what it links, probably could look at source but I usually just let it build and errors will tell me 2022-05-08 17:58:50 sadly my nixos vm is taking very sweet time grabbing packages for past 2 hours so I can't get into my standard alpine install and suffice with very slow WSL2 :c 2022-05-08 17:59:46 I just delegated a spare PC to be my Alpine builder 2022-05-08 18:00:35 wish I could do that but all my other machines are <= 4t 2022-05-08 18:01:16 `these paths will be fetched (2654.13 MiB download, 12494 MiB unpacked):` this might be the reason why it's so slow 2022-05-08 18:03:05 seems like it doesn't like being built in offline mode 2022-05-08 18:41:59 probably because it takes `true`/`false` string argument, instead of `1` 2022-05-08 18:42:09 `error in environment variable `CARGO_NET_OFFLINE`: provided string was not `true` or `false`` 2022-05-08 18:50:06 nmeum: I removed alpine-sdk and installed just abuild and build-base, it seems to go further now 2022-05-08 18:50:56 Not certain though 2022-05-08 18:51:35 Oh no, still same error, it just continues and fails every package 2022-05-08 18:53:06 [18:55:16] missed `mpfr-dev` in makedepends :') 2022-05-08 18:53:07 this reminds me 2022-05-08 18:53:19 for reasons unknown to me, mpfr don't do patch releases 2022-05-08 18:53:25 instead, they dump a bunch of patches on their website 2022-05-08 18:53:43 both fedora and gentoo (after I realised fedora were, and then looked into it) now apply a bunch, you might to as well 2022-05-08 18:53:45 just like readline 2022-05-08 18:53:47 https://www.mpfr.org/mpfr-current/#bugs 2022-05-08 18:54:02 yeah, although at least readline and bash announce them! 2022-05-08 18:54:09 afaict these were just silently added to the site :( 2022-05-08 19:16:46 do we have any alpine packages which include firefox addons 2022-05-08 19:49:49 ddevault: https://pkgs.alpinelinux.org/contents?file=*.xpi&path=&name=&branch=edge 2022-05-08 19:50:12 maybe browserpass? 2022-05-08 19:50:26 thanks panekj! 2022-05-08 19:51:59 Newbyte: it doesn't build in offline mode because of CARGO_HOME in build-aux/cargo.sh 2022-05-08 19:52:19 I hate stupid scripts like that 2022-05-08 19:53:02 imo, that package should be patched anyway so it builds offline and respects our fetch in prepare() 2022-05-08 20:03:43 https://tpaste.us/6Ejl 2022-05-08 20:13:15 + apkbuild https://tpaste.us/Vvd1 2022-05-08 20:35:03 panekj: I doubt it's malice. from what I understand the author is new to Rust 2022-05-08 20:35:31 so he might be receptive to fixing this upstream 2022-05-08 21:05:27 I'm not saying it's malice, I'm just a little frustrated when people do that :P 2022-05-08 21:05:42 it's also not limited to rust 2022-05-08 21:20:27 nmeum: pong 2022-05-08 21:20:32 appears i broke all the bootstrapping eh :) 2022-05-09 07:11:27 psykose: breaking bootstrapping is fine, happens all the time :) just had some questions regarding the rust aport but I think I sorted those out 2022-05-09 07:11:50 yeah, sadly i don't know why the compiler fails :c 2022-05-09 07:26:47 the stage2 builds fine but stage3 seems a bit messy, I think I will try to do build stage2 again without our alpine-specific triplet (and patches) and check if it also crashes and if so report that upstream? 2022-05-09 07:27:24 seems to me that rustc on riscv64 musl is not well tested in general since the stage2 didn't even compile without libc crate patches 2022-05-09 07:32:22 mhm 2022-05-09 07:32:28 i bet almost nobody has it working 2022-05-09 07:32:45 so, would be nice if upstream looked at it 2022-05-09 07:36:13 right, but I bet rust upstream doesn't like bug reports with custom triplets 2022-05-09 07:38:00 the only thing our custom triple sets is non-static crt default and pie, and then there's some stray patches 2022-05-09 07:38:04 i don't think the triple really matters 2022-05-09 07:40:06 (and fwiw they've already accepted the non-static crt default for musl upstream, just haven't actually made it yet, so that part is okay anyway) 2022-05-09 07:47:30 so you are suggesting just reporting this upstream as is? 2022-05-09 07:48:05 do you have a link for the "non-static crt default for musl" discussion upstream? 2022-05-09 07:55:43 ttthis i think https://github.com/rust-lang/compiler-team/issues/422 2022-05-09 07:55:49 but really this discussion spans like 30 issue pages 2022-05-09 07:56:52 well, unless they explicitly close issues that are not unknown triples instantly, i think the people will understand that the triple is not the cause of the segfaults 2022-05-09 07:57:28 and.. you can probably reproduce the same issue by just dropping the triple patch anyway, even with the slight difference 2022-05-09 08:00:11 alright, I guess I will write a bug report later today then. I suppose I can still build it without the patches if they explicitly ask me to do so. thanks :) 2022-05-09 21:30:39 I upgraded edge today (have been to busy at work lately) and I just wonder if something have changed regarding kernel modules or kernel config or some other change that I have missed. I get some errors when booting on 5.15.36, things like unknown filesystem vfat when mounting /boot/efi. When reverting back to 5.15.24 it boots just fine. I also don't have any working wifi in 2022-05-09 21:30:41 5.15.36-lts but that also work in 5.15.24-lts. 2022-05-09 21:33:17 HRio: Re zfs issues with Linux 5.15.37+, I built zfs-lts 2.1.4 from aports master, and I can confirm that it fixes the nginx sendfile issues. 2022-05-09 21:34:36 I would vote to upgrade zfs-lts 2.1.2 -> 2.1.4 in stable. 2022-05-09 21:34:40 EvTheFuture: it kind of sounds like your upgrade wasn't successful, like /lib/modules/5.15.36 might be incomplete/missing on your system 2022-05-09 21:35:23 EvTheFuture: the changes today to upgrade to 5.15.37 were just a version bump and nothing to do with packaging really: https://git.alpinelinux.org/aports/commit/main/linux-lts?id=a79940306e76efb11382ab87dae551ed709cb0b1 2022-05-09 21:36:35 craftyguy_: I haven't upgraded for some time so it might not be a recent change it can be a week or two or maybe even three weeks old. 2022-05-09 21:40:06 Anyway, I got it working on the old 5.15.24 so I can work tomorrow. Will try to investigate tomorrow evening. 2022-05-09 21:41:39 EvTheFuture: ah. maybe something in here broke it for you: https://git.alpinelinux.org/aports/commit/main/linux-lts?id=f2d708907ba2fccc8b499c588f713b47d5cba2ed 2022-05-09 21:42:03 or something else in the git log since 5.15.24: https://git.alpinelinux.org/aports/log/main/linux-lts 2022-05-09 23:13:30 /join #s6 2022-05-09 23:13:40 malvo: i'm ok, thanks 2022-05-09 23:13:55 trying out a new irc client lmao 2022-05-09 23:13:57 EvTheFuture: you didn't reboot after upgrading 2022-05-10 00:04:14 malvo: excellent idea :D 2022-05-10 05:11:33 xordspar0: Thanks I created #13797 to track this 2022-05-10 05:12:59 HRio: Thanks! 2022-05-10 05:50:17 Hello71: I did and when booting again is when it fails. 2022-05-10 08:45:52 I'm working on #13797 2022-05-10 09:13:21 ncopa: thanks 2022-05-10 15:01:33 does abuild build in a (namespaced?) chroot by default? 2022-05-10 15:01:33 what's the abuild equivalent of `xbps-src pkg foo`? (in case anyone knows) i aim for the resulting apk file and https://wiki.alpinelinux.org/wiki/Abuild_and_Helpers#abuild didn't seem to produce one (or i don't know where to look 2022-05-10 15:10:47 found it. $HOME/packages . duh 2022-05-10 15:11:48 Piraty: you need rootbld to build in a chroot 2022-05-10 15:12:46 so all the packages it is installing (due to -r) , are installed into / ? 2022-05-10 15:13:10 yes 2022-05-10 15:13:16 >/ 2022-05-10 15:13:17 ;/ 2022-05-10 15:13:20 and afterwards removed again 2022-05-10 15:13:22 but yes 2022-05-10 15:13:29 odd 2022-05-10 15:22:48 113 passed, 19 skipped in 617.09s (0:10:17) 2022-05-10 15:23:04 i have added a few tests to test diskless installs 2022-05-10 15:23:16 i now only have data disk 2022-05-10 15:23:27 i now only have data disk left 2022-05-10 15:23:39 but I guess its enough to create to tag a rc? 2022-05-10 15:24:02 oh, I should also add a test for the xen iso 2022-05-10 15:24:36 i wonder if there are some easy way to create multiples qemu guest with a shared user network? 2022-05-10 15:25:41 Create a bridge and connect them all to that bridge? 2022-05-10 15:26:14 It does require root permissions though 2022-05-10 15:26:23 slirp4netns? 2022-05-10 15:31:13 easy way is to use -net bridge, but requires root as ikke said 2022-05-10 15:32:18 -net user can be ok if you only want some limited ports 2022-05-10 15:32:46 -net socket is the general mechanism 2022-05-10 15:33:14 or -net vde, if you want to use vde 2022-05-10 15:33:33 looks like -net socket is the way to go 2022-05-10 15:33:50 i wonder if we should add a test for nfs server/client 2022-05-10 15:34:07 https://gitlab.alpinelinux.org/alpine/aports/-/issues/13789 2022-05-10 15:34:39 i guess that adding a test for that is not a high prio (yet) 2022-05-10 15:36:00 we do 3.16_rc1 then? 2022-05-10 15:36:16 riscv builder is still not idle, but I guess it never is? 2022-05-10 16:33:16 ncopa: better to wait a bit for removal php7 from community 2022-05-10 16:34:04 the blocker is #13460 2022-05-10 16:40:58 andypost[m]: When I have time, I'll see if the latest release still has issues 2022-05-10 16:47:36 src/print.c:85:17: error: ignoring return value of ‘write’ declared with attribute ‘warn_unused_result’ [-Werror=unused-result] 2022-05-10 16:47:40 -Werror ... 2022-05-10 16:53:00 export CFLAGS="$CFLAGS -Wno-error" 2022-05-10 17:49:21 ikke: maybe we'll do switch and fix reports if they come in 3.16, I wanna remove php7 before tagging RC 2022-05-10 17:50:00 at what point do packaging changes merged into aports stop showing up in what will eventually become 3.16? 2022-05-10 17:50:43 once ncopa created the 3.16-stable branch 2022-05-10 17:50:49 like if I send a patch now to something in main or community in aports/master, will it show up in 3.16? or did I miss the cutoff (or is this mentioned somewhere and I missed it) 2022-05-10 17:50:51 and switchees the buidlers over to follow that 2022-05-10 17:50:58 No, that happes at release 2022-05-10 17:51:13 so when v3.16.0 is tagged 2022-05-10 17:51:23 ok I see. I just saw some mention of an RC earlier, *typically* in other projects that means there's already a branch, so I was confused :) 2022-05-10 17:51:41 ikke: got it, thanks for clarifying :) 2022-05-10 17:54:38 ikke: is it ok to move php81 to community at this point? I think it's good to ship both releases 2022-05-10 17:55:12 In June php82 coming to testing 2022-05-10 17:55:33 Yeah, I don't see why not 2022-05-10 17:56:28 Ubuntu'22 even set it default!? 2022-05-10 18:11:53 that reminds me to move ceph17 to community 2022-05-10 20:52:54 golang releases 2022-05-10 20:55:47 just syscall fixes 2022-05-10 20:55:51 security wise 2022-05-10 20:56:00 where was that script to find everything to rebuild again 2022-05-10 20:57:17 part of atools 2022-05-10 20:58:58 ah, i see 2022-05-10 21:00:56 i assume the golang.org/x/ packages are not the same as the ones from std? 2022-05-10 21:01:03 this doesn't seem to print anything from go base 2022-05-10 21:01:16 correct, they are external modules 2022-05-10 21:02:05 well, that seems completely useless to find out what to rebuild then 2022-05-10 21:02:08 unless i missed something 2022-05-10 21:02:39 i guess there's no easy way at all then 2022-05-10 21:07:48 and even if we used this for periodic scanning of external modules, it doesn't print versions of them, so it's not useful for that either 2022-05-10 21:08:11 i guess the information is not there at all 2022-05-10 21:08:15 unfortunate 2022-05-10 21:46:42 I built linux-lts 5.15.35 and that boots just fine on my laptop. However something seems to have happened witn NM. I have to connect to wifi manually with wpa_supplicant. I'll try to build later versions to try to figure out if/when it stopped working. 2022-05-10 21:47:49 EvTheFuture: networkmanager recently split off the wifi stuff into a separate subpackage, so make sure networkmanager-wifi is installed 2022-05-10 21:50:37 craftyguy: Ah, thanks that must be it... 2022-05-10 21:55:42 ya that one tripped up a few folks :P 2022-05-10 22:18:00 It seems like the kernel issue was a false alarm as well. upgraded again (after have booted ith manually built linux-lts v 5.15.36) witj apk upgrade -a which installed the latest one which booted just fine. Weird... but I'm happy it works. 2022-05-11 05:33:56 morning 2022-05-11 05:35:26 craftyguy: the idea is that we push everythign to git master til we tag a release and do the branching then. The idea is to encourage everyone to help getting the release out rather than start work on next new features 2022-05-11 05:47:55 ncopa: ah, that makes sense. thanks for explaining :) 2022-05-11 06:04:25 Anyone more than me that got problems with the new version of PipeWire when using bluetooth headset profile? For me the source is starting up in SUSPENDED state and it looks like i can't resume the source... I have to test mor though... 2022-05-11 06:06:43 hmm, not that I've noticed. how new is the pipewire update? 2022-05-11 06:07:05 ACTION hasn't used his bluetooth headset in about a day, but it worked the last time 2022-05-11 06:07:44 ah the last pipewire update is >1 week 2022-05-11 06:09:22 pipewire-0.3.51-r0 x86_64 {pipewire} (LGPL-2.1-or-later) [installed] 2022-05-11 06:11:12 At the moment I have to use the internal microphone in the computer since the headset source always end up in the state SUSPENDED... wonder if any config files have changed... I'll dig into it when I have more time after all meetings :) 2022-05-11 06:38:05 When looking at the pipewire log, it looks like it goes straight from creating to suspended when the headset is being connected: impl-node.c: 379 node_update_state()] (bluez_input.XX_XX_XX_XX_XX_XX.headset-head-unit-268) creating -> suspended 2022-05-11 06:50:41 hi all any plans for 3.16? 2022-05-11 06:51:06 Yes, it's in the works 2022-05-11 06:51:24 ncopa wants to tag the first rc soon 2022-05-11 06:51:27 ikke, is there any tracker on gitlab? 2022-05-11 06:52:02 i saw this - https://gitlab.alpinelinux.org/alpine/tsc/-/issues/40 2022-05-11 06:52:08 morning need help merging https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34275 into main 2022-05-11 06:52:58 ikke, ncopa but it looks like far from release 2022-05-11 06:59:22 https://gitlab.alpinelinux.org/alpine/aports/-/milestones/183#tab-issues 2022-05-11 07:00:07 But not all issues in that milestone will be fixed before release 2022-05-11 07:28:58 psykose, hi, i was looking for !33100 ci job but was not able to find it, tried pipelines, jobs (your message from 2022-04-11 15:58:30 ) 2022-05-11 07:29:19 ha 2022-05-11 07:29:33 it was merge request :) 2022-05-11 07:29:43 algitbot, thanks :) 2022-05-11 07:47:06 psykose, ignore it, it is already in edge, great 2022-05-11 14:17:46 Just a quick update of my headset problem. I reverted back to pipewire 0.3.50 and now the bluetooth headset source is in state RUNNING and it can be used. In version 0.3.51 it immedietly goes from CREATING to SUSPENDED... Have anyone heard if anyone else have seen this issue? 2022-05-11 17:17:17 EvTheFuture: the changelog for 0.3.51 mentions only 1 bluetooth change: https://github.com/PipeWire/pipewire/blob/master/NEWS#L78 2022-05-11 17:17:23 but maybe it is a regression? 2022-05-11 17:17:47 I'd probably report it upstream at this point, if reverting back to 0.3.50 "fixes" it 2022-05-11 17:19:22 oops here's the upstream link: https://gitlab.freedesktop.org/pipewire/pipewire/-/blob/master/NEWS#L78 2022-05-11 17:19:30 stupid github mirrors polluting search results 2022-05-11 17:19:39 if you know how to bisect you can try that :) 2022-05-11 17:20:56 ya that's honestly the best way to move forward on that problem :D 2022-05-11 17:22:17 craftyg: Thanks. I also saw that there where no obvious changes to this, but just wanted to know if anyone else stumbled upon this. I might have a bit of a different setup than most people since I use Alpine as a KVM host and run everything in separate VMs and push the audio via a separate network interface from the VMs. Might have been some config changes, bit I checked earlier 2022-05-11 17:22:19 and nothing obvious had changed. But I suspect it might have something to do with suspend when idle maybe. 2022-05-11 17:24:11 psykose: the biggest challenge I guess is finding time :D 2022-05-11 18:56:28 Ariadne: btw... I have been testing kyua a bit today. its pretty nice 2022-05-11 18:56:34 nicer than bats 2022-05-11 18:56:40 yeah kyua is great 2022-05-11 18:57:34 i did find it a bit higher entry bar to get started, but once passed the initial "hello world" its pretty straightforward 2022-05-11 18:57:56 i think I will use much more of it in the future 2022-05-11 20:34:05 ncopa: probably a silly question, but will the linux-lts kernel in edge be the same linux-lts kernel in 3.16 (i.e. kconfig enabled now in edge linux-lts will be also enabled in the 3.16 linux-lts) ? 2022-05-11 20:34:25 yes 2022-05-11 20:34:34 they are exactly the same atm 2022-05-11 20:34:52 asking because there are some differences between the two lts kernels in edge vs 3.15, specifically some options not enabled in the 3.15 kernel that are on edge 2022-05-11 20:35:17 ok, I kind of expected that answer :P 2022-05-11 20:35:18 Enabled options are not backported by default 2022-05-11 20:35:44 ok. I submitted a patch to 'backport' one option to 3.15's linux-lts 2022-05-12 11:38:35 im adding unit tests for alpine-conf. this is a huge improvement. now we test that `setup-* -h` is consistent. print to stdout, exit with success (instaed of error) etc 2022-05-12 11:39:06 and im finding bugs while at it 2022-05-12 11:39:22 for example `setup-proxy none` would not remove configured proxies 2022-05-12 11:40:33 that is a giga amount of fixes right there 2022-05-12 11:40:38 proud of you ncopa :) 2022-05-12 11:41:38 the best thing is that i will be much more confident when reviewing merge requests 2022-05-12 11:42:03 my biggest regret is that i didnt do this way sooner 2022-05-12 11:57:29 I want to believe my insistence played a part in motivating you to do it :) 2022-05-12 13:49:03 Great job guys! 2022-05-12 14:25:55 Is there a reason why a patch release for 3.15 hasn't been released along with the openssl patch? 2022-05-12 14:26:20 Also: why weren't other branches patched? 2022-05-12 14:27:19 which openssl patch? 2022-05-12 14:28:02 this? 9dac81a0eecfa45770f930d15a84793a31625a06 2022-05-12 14:29:14 looks like we were not affected by it 2022-05-12 14:30:33 !34181 no additional details though 2022-05-12 14:31:43 ncopa: where does it say that? 2022-05-12 14:32:02 https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34181/diffs#ec642615cc0564dfdbb98166ff142d5ef9e81ca7_51_51 2022-05-12 14:32:14 fixed in version '0' 2022-05-12 14:32:24 which means it never affected us 2022-05-12 14:32:32 Huh. 2022-05-12 14:32:49 also https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2022-1292 2022-05-12 14:33:26 "The c_rehash script does not properly sanitise shell metacharacters to prevent command injection. This script is distributed by some operating systems in a manner ..." 2022-05-12 14:33:29 Oh, we don't ship that script 2022-05-12 14:33:31 that;s why 2022-05-12 14:33:36 we dont use the c_rehash script 2022-05-12 14:33:50 but at some point we did, so i think the secixes info is technically wrong 2022-05-12 14:34:11 Yeah, it gets explicitly deleted in the package 2022-05-12 14:34:18 s/package/APKBUILD 2022-05-12 14:34:18 Nulo meant to say: Yeah, it gets explicitly deleted in the APKBUILD 2022-05-12 14:35:01 All other CVEs in https://www.openssl.org/news/secadv/20220503.txt are for OpenSSL 3.0. I guess Alpine dodged that bullet (?) 2022-05-12 14:35:46 or we have good devs :) 2022-05-12 14:36:11 Dodged the bullet _because_ of good technical decisions 2022-05-12 14:36:30 aa83340f268830cb09ffae6d2af3f0b87f8b56bd 2022-05-12 14:36:47 Ah, I thought OpenSSL 3 wasn't packaged at all 2022-05-12 14:36:54 but, yeah, no hurry to make new release 2022-05-12 14:41:26 As far as I can tell, only spdk in testing depends on openssl3 https://pkgs.alpinelinux.org/package/edge/testing/x86_64/spdk 2022-05-12 14:49:10 amusing bug: alpine hangs when booting in qemu with -vga virtio, but only when the window is maximized. 2022-05-12 14:50:40 it prints "Kernel panic - not syncing: System is deadlocked on memory", but only to serial 2022-05-12 14:51:34 i wonder if this is related to people reporting that alpine hangs on boot; something is going wrong and kernel panics but for some reason doesn't display the message 2022-05-12 14:52:05 i wonder if possibly mkinitfs is doing something that causes the errors to not be displayed 2022-05-12 23:50:48 ncopa: i tried to find the first version, actually 2022-05-12 23:50:55 but i found that script was being deleted like 6 years ago 2022-05-12 23:50:58 or.. more 2022-05-12 23:51:04 i think it was since 2.x days, or maybe 3.0 2022-05-12 23:51:23 so i thought 0: made more sense as otherwise it's some ancient version anyway, but i suppose it can be updated 2022-05-13 03:38:56 psykose: but the stable branches were never affected at least 2022-05-13 03:39:08 not the current ones, no 2022-05-13 03:39:31 nod 2022-05-13 05:37:48 psykose: thank you for the help with the boost1.77 patches :D 2022-05-13 05:37:53 :) 2022-05-13 05:37:58 i am also working in boost 1.78 right now 2022-05-13 05:38:19 seems to be unordinary so if the rebuilds work then i guess i will send it 2022-05-13 05:38:33 are those patches in 1.78? 2022-05-13 05:38:56 they are backported from 1.78 2022-05-13 05:39:04 ah nice 2022-05-13 05:39:09 (or boost-math-standalone-1.77, technically since they are there too) 2022-05-13 05:39:30 and the geometry one no idea, but it is in 1.78 as well 2022-05-13 05:39:38 ok 2022-05-13 05:53:55 is this an acceptable backport to 3.15? https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34264 2022-05-13 05:54:29 only if ncopa is fine with it 2022-05-13 05:54:44 note it will also need a pkgrel bump for all the -lts kernel module packages 2022-05-13 05:56:11 ah 2022-05-13 05:58:57 probably best to make that change then when the lts kernel gets another bump(?) 2022-05-13 05:59:09 (assuming ncopa is OK with it, heh) 2022-05-13 05:59:58 more or less 2022-05-13 06:48:46 morning 2022-05-13 06:49:47 craftyguy: yeah, i think that is acceptable. will have a look at it, but not sure if it will be today 2022-05-13 06:50:21 and yes, i will do it with the next kernel bump. 5.15.39 is out already, so it should be in too far future 2022-05-13 06:51:45 btw, you dont need to bother bump pkgrel for that. I cna manually cherry-pick it together with the version bump. 2022-05-13 06:51:58 the reason is that we also need to rebuild the 3dparty kernel modules 2022-05-13 14:59:21 Ariadne: did you ever have another go at rust on rv64 2022-05-13 15:09:34 ddevault: nmeum_ tried the other day 2022-05-13 15:09:38 got to stage2 2022-05-13 15:10:09 stage2 was not working properly though 2022-05-13 15:27:18 aight 2022-05-13 15:27:28 wl-clipboard-x11 and xclip have conflicting path /usr/bin/xclip 2022-05-13 15:28:51 APKBUILD doesn't have a mechanism to propagate that to apk ? 2022-05-13 15:29:30 propegate what? 2022-05-13 15:30:07 why would you have xclip installed together with wl-clipboard-x11 ? 2022-05-13 15:30:29 package conflict, so apk-add will not install one (or remove the other, depending on if it's an implict dependency or explicit one) 2022-05-13 15:30:45 it has `provides` and `replaces` 2022-05-13 15:30:45 You can specify these packages conflict 2022-05-13 15:35:51 ok it does have https://git.alpinelinux.org/aports/tree/testing/wl-clipboard-x11/APKBUILD#n15 yet apk-add allows to install both and one remains broken 2022-05-13 15:37:30 That just means you can run apk add cmd:xclip and it can select this package 2022-05-13 15:37:41 They should directly conflict 2022-05-13 15:38:01 by adding !xclip to the depends of wl-clipboard-x11 2022-05-13 15:40:24 i wasn't aware that !foo means confligs=foo 2022-05-13 15:40:40 conflicts=foo ofc 2022-05-13 16:56:21 I'd like to update testing/gprbuild... the package in aports is "2019" but upstream has switched to using semver and the new version is "22.0.0"... how should I handle this? 2022-05-13 16:56:51 to clarify, I'm referring to the pkgver field 2022-05-13 17:01:47 synack: One options is to just do it 2022-05-13 17:02:02 ie, just start using the new version scheme 2022-05-13 17:02:33 At least when upgrading to a new major version, users are advised to use --available, which means apk will 'downgrade' it 2022-05-13 17:02:48 ooh, interesting 2022-05-13 17:34:16 seems like the builders are busy 2022-05-13 17:34:42 I was about to tag 3.16.0_rc1 but maybe I should wait til Monday 2022-05-13 17:55:33 it's 19:55 on a Friday evening SHIP IT 2022-05-13 18:08:44 it's just an rc... ;) 2022-05-13 18:18:48 but if you say you'll tag it on monday, perhaps folks will scrable to get heir commits in before that 2022-05-13 18:34:55 It's just that ncopa wants the builders to be idle before making a release (to make sure there are no broken packages) 2022-05-13 18:40:14 ddevault: https://github.com/rust-lang/rust/issues/96876 2022-05-13 18:41:41 the stage2 packages are here https://dev.alpinelinux.org/~nmeum/edge/riscv64/ but as described in the bug report it segfaults on more complex inputs and I am simple not familiar enough with rust compiler internals to fix that on my own 2022-05-13 18:42:08 *simply 2022-05-13 18:45:05 thanks nmeum_! 2022-05-13 18:46:54 nmeum_: did you upstream the libc crate patches you had to make? 2022-05-13 18:49:11 yes, I was just minor change though the biggest issue was that rustc-1.6.0 vendors an older version of the libc crate, thus I had to backport a few patches which were already upstream by someone else 2022-05-13 18:49:16 *it was jist a minor change 2022-05-14 08:11:14 Can one move more than 1 packages between repos in a single mr? (!34395) 2022-05-14 08:20:12 Sure, if the packages are related, it makes sense to move them at the same time 2022-05-15 01:02:26 I made a (rushed) acf responsive skin poc and here (https://fars.ee/-Is-) are some screenshots 2022-05-15 01:02:56 great work 2022-05-15 01:02:57 if anyone have ideas or suggestions on the design please tell me ^^ 2022-05-15 08:17:19 hello, I'm using the aports repository to build my own customized alpinelinux livecd and that's working really lovely so far but what would be the right approach to modify packages and build and add my own packages for the livecd environment? 2022-05-15 08:44:02 Looks like openjdk18 is stuck on ppc64le 2022-05-15 08:52:05 it is 2022-05-15 10:20:27 Now not anymore 2022-05-15 18:41:21 test 2022-05-15 18:41:43 failed successfully 2022-05-15 18:59:31 failure successed failfully 2022-05-15 19:00:30 woho, it works 2022-05-15 19:00:33 thanks :D 2022-05-15 19:07:09 could someone take a look? 2022-05-15 19:07:09 https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34416 2022-05-15 19:07:23 maintainer seems to be inactive for a year 2022-05-15 19:48:47 heh 2022-05-15 19:48:49 https://github.com/fabaff "sporadically on.." 2022-05-15 19:51:21 it's not untrue 2022-05-15 19:55:06 Seems they are active in nix packages 2022-05-16 04:57:42 ncopa: good morning, whenever you arrive, and thanks for publishing alpine-conf-3.14.0_rc2! this makes it possible for me to send the big thing. 2022-05-16 04:57:49 So, folks, here we go: 2022-05-16 04:58:02 !34459 2022-05-16 05:00:55 Full description and rationale of the changes in the MR; this impacts busybox-initscripts, alpine-base, eudev, udev-init-scripts, and cloud-init, and creates new and exciting packages; not for the sheer fun of it, but because the end layout is much more rational and logical than it was before, so please don't be afraid. 2022-05-16 05:04:17 merging this will also allow minimal to push his changes to cloud-init in an easier way. :) 2022-05-16 06:11:11 I suppose ncopa will not being an rc today either 2022-05-16 06:20:20 not being an rc? 2022-05-16 06:20:55 doing* 2022-05-16 06:25:47 as long as someone merges my MR it doesn't matter to me if an rc goes out today or not :P 2022-05-16 06:26:08 fyi, this was not a reaction to your MR 2022-05-16 09:21:35 I'd like to do 3.16.0 rc1 asap. and the 3.16.0 next week 2022-05-16 09:22:02 what do we need for that to happen? 2022-05-16 09:23:22 can we do the 3.16.0 next Monday or Tuesday? 2022-05-16 09:37:33 note that !34459 needs to be in the same release as alpine-conf-3.14.0 else things will break, so it would be awesome if it could be merged for 3.16.0 :) 2022-05-16 09:43:34 what would break? 2022-05-16 09:45:32 setup-devd wouldn't be properly supported and setup-udev would be inconsistent with it 2022-05-16 09:46:51 (setup-devd references services that are only provided by my MR, the change to alpine-conf was always meant to be done at the same time, that was just made more difficult by the fact that alpine-conf has a separate numbered release from aports) 2022-05-16 09:51:59 so current alpine-conf is brokne? 2022-05-16 09:52:03 broken* 2022-05-16 09:54:47 ... like I've always told you 2022-05-16 09:54:56 the change to alpine-conf and the MR need to go together 2022-05-16 09:55:53 which is why I needed a release of alpine-conf *before* the alpine-conf aports was changed, so I could integrate the alpine-conf aport bump into the MR 2022-05-16 09:55:57 but you chose to do differently 2022-05-16 09:56:03 it doesn't matter as long as the MR gets merged 2022-05-16 09:58:01 ok. i get it. its broken and its my fault 2022-05-16 09:58:10 that's not what I'm saying... 2022-05-16 09:58:56 I'm saying "you did the right thing by releasing alpine-conf, now keep doing the right thing and ensure the MR is merged before the release" 2022-05-16 09:59:22 skarnet meant to say: I'm saying "you did the right thing by bumping alpine-conf, now keep doing the right thing and ensure the MR is merged before the release" 2022-05-16 09:59:22 s/releasing/bumping 2022-05-16 09:59:33 I think I'd like to have some tests in place first 2022-05-16 10:00:10 I don't trust myself to not break things anymore 2022-05-16 10:00:12 ... wasn't that why you delayed the alpine-conf bump in the first place? I thought you had done them and that was why you had bumped it in aports 2022-05-16 10:00:43 well, I didn't add tests for everything yet 2022-05-16 10:01:01 i know for sure that mdev works with different install combinations 2022-05-16 10:01:04 I saw you (or someone else) added a ptpdev test to busybox-scripts (and it broke my initial pipeline, I've fixed the MR since) 2022-05-16 10:01:22 but I dont have test for udev or mdevd yet 2022-05-16 10:01:28 mdev is still the default 2022-05-16 10:01:42 so nothing that doesn't break now will break afterwards 2022-05-16 10:03:06 the change has been ready for 3 months and a half. 2022-05-16 10:07:42 the MR has gotten a couple of reviews and comments already 2022-05-16 10:10:46 looks like we might be able to do the rc1 today after all. the builders are mostly done. just rpi kernel that is in the queue now 2022-05-16 13:24:18 hum.. I dunno if we can ship v3.16 with the current state of setup-devd 2022-05-16 13:30:02 the fix is simple 2022-05-16 13:30:23 let's get to something we can agree on wrt 34459 2022-05-16 13:30:26 merge it 2022-05-16 13:30:29 problem solved 2022-05-16 13:30:40 (and if setup-devd needs to be changed, that can be done too) 2022-05-16 14:04:16 i'd like to at least move mdevd to main 2022-05-16 14:04:18 https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34459/diffs?commit_id=b871e5aaca73f4463737e578cd7e8d4067ee69e6 2022-05-16 14:09:03 i think I'm gonna cherry-pick that for now 2022-05-16 14:12:51 please read my latest comment before doing anything. 2022-05-16 14:14:05 moving mdevd to main without reworking the device manager framework is exactly what I did not want to do because it would increase tech debt in the framework of device managers. 2022-05-16 14:14:17 can't we just 2022-05-16 14:14:21 wait until 3.17 release 2022-05-16 14:14:22 :D 2022-05-16 14:14:25 (and if I had done that it could have been perceived as me wanting to just promote mdevd, which is not the case) 2022-05-16 14:14:53 (also, coordinating these sort of changes is what the TSC exists to do, this probably should have gone through it) 2022-05-16 14:14:59 given the pushback this is getting, I think this will de facto have to wait until 3.16 is out anyway 2022-05-16 14:15:35 in part, it's getting pushback because it's (a) late, and (b) did not go through the TSC system change proposal process 2022-05-16 14:15:50 I don't get how it's a system change 2022-05-16 14:16:09 it's a lower level change than most aports changes, yes 2022-05-16 14:16:23 but the default doesn't change, the system doesn't change, etc. 2022-05-16 14:16:32 because of the `setup-devd` thing 2022-05-16 14:17:09 anything that changes several packages and introduces new infrastructure would be a system change even if the end result is nothing changes for a user 2022-05-16 14:17:11 device managers are already de facto pluggable and setup-udev is already a thing 2022-05-16 14:17:19 and an ugly hack to boot 2022-05-16 14:17:23 sure 2022-05-16 14:17:24 setup-devd cleans that up 2022-05-16 14:17:31 but cleaning that up is still a change 2022-05-16 14:18:15 well if it needs a TSC discussion to pass I'm all for it 2022-05-16 14:18:34 at this point, a couple more months of bureaucracy won't change much 2022-05-16 14:20:20 also, if setup-devd is at a level where TSC must be involved, then expect a lot of fun in the incoming months, I'm gonna give the TSC a workout 2022-05-16 14:20:22 and may make for less friction =) 2022-05-16 14:20:26 i am skeptical of your assumption that it would only take a couple of months 2022-05-16 14:22:36 skarnet: the TSC moves quickly for system changes 2022-05-16 14:22:40 you're saying "it's late" but again the thing has been ready for 3.5 months and it's only now that I could push it because it had to be done at the same time as the alpine-conf bump 2022-05-16 14:22:56 you think I *like* dumping big changes right before a release? 2022-05-16 14:22:58 it's not your fault that it's late 2022-05-16 14:23:03 you think I like making things harder for everyone? 2022-05-16 14:23:25 and I don't really want to blame anyone else either 2022-05-16 14:23:49 ncopa: BTW I added the mdev /dev/disk/by-* tests to busybox-initscripts that you asked for previously, !32330 2022-05-16 14:23:50 i'm not blaming you, but if you had gone through the TSC, then there would have been more eyeballs on the alpine-conf thing because the "big picture" would have been understood 2022-05-16 14:24:30 you've been going "i need alpine-conf changed" without showing us the big picture 2022-05-16 14:25:12 I didn't know alpine-conf was so special 2022-05-16 14:25:19 it's not like mkinitfs or the kernel or... 2022-05-16 14:25:39 anyway I gtg, bbl. 2022-05-16 14:39:56 mdevd, mdev, udev, eudev, systemd-udev 2022-05-16 14:40:13 Linux is synonym of mess 2022-05-16 14:43:11 software that link against libudev can still work without udev running? (e.g. Xorg, retroarch and all other input based applications) 2022-05-16 14:44:42 does adding new script to alpine-conf also need TSC coordination? 2022-05-16 14:48:35 panekj: depends, if it is a system change then it probably doesn't hurt 2022-05-16 14:56:41 skarnet has probably brought up this thing 15 times over the past 3 months (probably more), and nobody seems to have known to suggest that a tsc issue was the real place it should have been put until now 2022-05-16 14:57:32 is it really the fault of it being put in the wrong place, and not everyone just ignoring him over and over and then claiming it's too late? :p 2022-05-16 15:16:34 "software that link against..." <- Yes, via libudev-zero, I guess. 2022-05-16 15:21:02 eudev-libs are also split from eudev, so implictly being linked to libudev doesn't pull it in or run it. whether the program actually works is another story, but i would assume most do 2022-05-16 15:29:46 also libudev-zero provides udev-libs? 2022-05-16 15:30:47 i guess you could make a libudev-really-zero that pretends that no devices are attached to save like 40k but i don't see the point 2022-05-16 15:32:18 before rc can we get abuild; previously discussed many times; MRs merged :) 2022-05-16 15:33:20 Ariadne: setup-devd is sort of ok as is, and I think it is fixable without involving TSC 2022-05-16 15:35:14 we can do it without breaking current stuff 2022-05-16 15:35:26 (i thnk) 2022-05-16 15:51:42 panekj: which abuild MR? 2022-05-16 15:58:04 skarnet: i took 4 of the commits, so we at least move forward. Can you please rebase https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34459 locally? 2022-05-16 16:07:38 ncopa: sorry, but nak 2022-05-16 16:08:01 ok? why? 2022-05-16 16:08:45 so I basically went out of my way to write a framework for device managers to do things *cleanly* in the long run and not just push my shit 2022-05-16 16:08:58 and you're saying "nah we'll just move mdevd to main and add the necessary hacks" 2022-05-16 16:09:05 no, I'm not adding the necessary hacks 2022-05-16 16:09:15 I'm not going to be the guy who adds tech debt to Alpine because it's easier 2022-05-16 16:09:31 I'm the one you call to make the deep necessary changes 2022-05-16 16:09:37 if it's not the right time, fine 2022-05-16 16:09:41 wait until 3.16 is out 2022-05-16 16:09:53 if it needs TSC involvement, then we'll involve the TSC 2022-05-16 16:10:09 but I'm not adding another hack to the pile of hacks 2022-05-16 16:10:15 the mdevd move to main does not need TSC involvement 2022-05-16 16:10:17 my job is to destroy those piles 2022-05-16 16:10:33 but ok. you are the maintainer there, so I'll revert it 2022-05-16 16:10:56 ncopa: !135 !143 !142 !140 !134 and probably plenty of others which have stagnated in the repo 2022-05-16 16:11:07 what about the busybox-initscripts permission fix. do you mind if I keep it? 2022-05-16 16:11:08 algitbot nooooo 2022-05-16 16:11:50 I mean, if you want to move the mdevd package to main, with mdevd-openrc and everything, why not indeed, and any fixes that don't involve anything are fine 2022-05-16 16:12:08 but the real work is the mdev split and setup-devd 2022-05-16 16:12:20 and I don't want the MR to become "we're just moving mdevd to main" 2022-05-16 16:12:31 because I could have done that 4 months ago and spared us a lot of energy 2022-05-16 16:12:42 and that's *not enough* 2022-05-16 16:14:06 ok. it was the best I could offer before the 3.16 release. I thought that having mdevd in main was better than nothing, specially since current alpine-conf/setup-devd currently needs it 2022-05-16 16:14:14 the thing is 2022-05-16 16:14:40 if you don't include the framework changes then setup-devd needs to be entirely rewritten, in a hacky way just like setup-udev 2022-05-16 16:14:46 ikke: please consider to get rid of php7 as !34281 the only blocker, not sure next release will be ready for 3.16 2022-05-16 16:15:12 I'm not objecting to having mdevd in main, I'm objecting to cherry picking that part of the MR and shelving the rest forever 2022-05-16 16:16:28 so sure, please pick these commits, but then setup-devd needs to be adapted to the weird chimaera that is the current 3.16 device manager landscape 2022-05-16 16:17:01 and 1. I don't want to be the one who does that, 2. it absolutely needs to be revisited after 3.16 is out 2022-05-16 16:17:35 setup-devd needs to be adapted regardless if we move mdevd or not, and I will do it 2022-05-16 16:18:25 i can keep mdevd in main but the depends_openrc needs to go in that case 2022-05-16 16:18:40 depends_openrc="mdevd mdev-conf" 2022-05-16 16:18:53 because there is no mdev-conf (yet) 2022-05-16 16:18:57 you're not splitting busybox-initscripts? 2022-05-16 16:19:06 nope, not yet 2022-05-16 16:19:21 then mdevd has a dependency on busybox-initscripts, nope 2022-05-16 16:19:21 can I test anything or help out in any other way here? 2022-05-16 16:20:17 skarnet: and if I split out mdev-conf as a separate package? 2022-05-16 16:20:22 that works 2022-05-16 16:21:14 so what do you need me to rebase exactly? and how can I do it while keeping the rest of the commits in a safe place? 2022-05-16 16:21:53 (prob won't do it tonight, I have to go in 30 minutes and then I'll be sleeping, but I can work on it early in the morning so it's ready when you arrive) 2022-05-16 16:22:23 should be doable. the other commits are in git master currently and I need to decide to revert or split out mdev-conf 2022-05-16 16:22:44 i'm aiming for splitting out mdev-conf but that will introduce a conflict for your rebase 2022-05-16 16:23:16 skarnet: can you deal with a git rebase conflict or should I revert? 2022-05-16 16:24:45 I'm not comfortable enough with git, if you're making changes I should probably let you work, push and merge, and on my side manage to stash my changes then pull your version and work on the rest for later 2022-05-16 16:25:27 so whatever you think is best for 3.16, I'll manage what I have locally to put it in shape for after 3.16 2022-05-16 16:25:37 unless you need me to do stuff locally and push again? 2022-05-16 16:25:44 (in which case it will have to wait until tomorrow) 2022-05-16 16:25:50 no, that should not be neccessary 2022-05-16 16:27:04 ok, I'll clean up the current git master and setup-devd for the 3.16 release and we deal with the rest later 2022-05-16 16:27:10 okay 2022-05-16 16:27:15 I'm sure we can help you with any git issues 2022-05-16 16:27:32 omni: I'm sure of it, but not in the next 20 minutes 2022-05-16 16:27:58 skarnet: no, I meant for later, when you get back =) 2022-05-16 16:28:33 but ah, you meant not comfortable enough to do git things in 20 minutes now? 2022-05-16 16:29:07 I like to think I'm smart enough to figure out anything, even git-related, given infinite time 2022-05-16 16:30:08 sure, but we can never be as great as we know we can be 2022-05-16 16:30:26 because we're mortal 2022-05-16 16:34:53 ok. i think we are making progress then. 2022-05-16 16:35:11 \o/ 2022-05-16 16:35:33 im gonna take a walk and then I'll trey clean up setup-devd for release. if the builders are ready by then I might even do rc1 2022-05-16 16:35:38 git rebase is easy, fixing conflicts is .... 2022-05-16 16:35:55 ncopa: can we please get abuild fixes first 2022-05-16 16:36:15 fixes+improvements 2022-05-16 16:37:26 well, the abuild train left in april when we bootstrapped the 3.16 builders 2022-05-16 16:38:15 well, I was told before all that we will get to merging and then nothing happened 2022-05-16 16:38:42 https://gitlab.alpinelinux.org/alpine/abuild/-/merge_requests/135 will require all rustpackages to be rebuilt 2022-05-16 16:38:54 and potential bugs needs to be fixed 2022-05-16 16:38:59 I'm aware 2022-05-16 16:39:03 i'd guess that is a couple of days of work 2022-05-16 16:39:05 I've rebuilt community with that change 2022-05-16 16:39:10 nothing broke from that 2022-05-16 16:39:30 we need get 3.16 out early next week 2022-05-16 16:39:40 as in, "it did build", so there can be potential bugs in runtime 2022-05-16 16:40:09 can we at least get the stuff merged so it can be bumped in edge? 2022-05-16 16:40:12 and almost all of today was spent on this mdevd stuff 2022-05-16 16:40:29 and **ready** for after 3.16 2022-05-16 16:41:13 edge == 3.16 2022-05-16 16:41:21 right now 2022-05-16 16:41:40 will require all go packages to be rebuilt https://gitlab.alpinelinux.org/alpine/abuild/-/merge_requests/143 2022-05-16 16:41:52 man... this will delay the rc1 another week 2022-05-16 16:42:00 whats the consequence of not include those in 3.16? 2022-05-16 16:42:07 it can be after or before, I don't really care about that, I just want to have proper working package in edge 2022-05-16 16:42:43 it has been massive PITA for months to maintain/build packages 2022-05-16 16:43:20 why? 2022-05-16 16:43:28 so if we dont do this it will become a PITA to do security fixes for rust and go in the future? 2022-05-16 16:44:01 because whenever I'm making some package or doing anything with packages I'm being told in MR that I have to re-do parts of it because of open MR in abuild 2022-05-16 16:44:18 we are already doing lots of custom things in aports based on open MRs in abuild 2022-05-16 16:44:22 yet abuild is still outdated 2022-05-16 16:44:25 https://gitlab.alpinelinux.org/alpine/abuild/-/merge_requests/140 is acceptable, but I dont understand why its important for v3.16 2022-05-16 16:45:15 ncopa: it's not really important for 3.16, it's important to get things merged 2022-05-16 16:45:25 although few of those changes should be in 3.16 2022-05-16 16:45:38 *since we already do them in aports* 2022-05-16 16:45:44 i think none of them should be in 3.16 2022-05-16 16:45:55 currently we are working on getting 3.16 out 2022-05-16 16:46:29 i had a look at them all 2022-05-16 16:46:45 https://gitlab.alpinelinux.org/alpine/abuild/-/merge_requests/60 2022-05-16 16:47:15 3 months ago: "ncopa will take a look soon" 2022-05-16 16:47:51 same time that valve uses? 2022-05-16 16:48:09 sigh 2022-05-16 16:48:14 i really need that walk now 2022-05-16 16:48:29 re https://gitlab.alpinelinux.org/alpine/abuild/-/merge_requests/60 2022-05-16 16:48:31 i dont like it 2022-05-16 16:48:53 because it reminds me of gentoo's magic eclasses and helpers 2022-05-16 16:49:00 we already do that in aports 2022-05-16 16:49:33 which are very convenient if you are familiar with them, but painful if you are not or are new 2022-05-16 16:49:41 and that's my main problem, we are doing things in aports which haven't been merged in abuild 2022-05-16 16:49:54 i know we do, and I'm kinda ok with it, but im not sure we want encourage its use 2022-05-16 16:50:08 now im losing the sunset 2022-05-16 16:50:10 right, but people use newapkbuild 2022-05-16 16:50:20 and newapkbuild does old stuff and then devs tell people to do new stuff 2022-05-16 16:50:26 sorry that i cannot help you more today 2022-05-16 16:50:33 anyway, I don't want to hold you 2022-05-16 16:50:41 3.16 wil still be the best alpinelinux release so far! 2022-05-16 16:50:51 enjoy ur walk (if your still going) 2022-05-16 16:51:00 today has been exhausting. i will probably need take tomorrow off 2022-05-16 16:51:04 good evening 2022-05-16 16:51:08 and don't take it as an attack on you or devs, but it's been pretty frustrating to do anything 2022-05-16 17:37:43 ncopa: with the move of mdev.conf to mdev-conf shouldn't ptpdev, ptpdev.bats, persistent-storage, (and persistent-storage.bats when my MR eventually gets merged) also move to there? 2022-05-16 18:07:54 minimal. pptdev and persistenet-storage, yes. *.bats. no. see 161eb826caaab56473211ea6b8202f358fb91477 2022-05-16 18:10:14 ncopa: ok but that only shows mdev.conf moving, not ptpdev and persistent-storage 2022-05-16 18:11:12 ncopa: also can you confirm whether or not I'm on the right track with my persistent-storage.bats tests? 2022-05-16 18:22:24 amove etc/mdev.conf lib/mdev 2022-05-16 18:22:51 the ptpdev and persistent-storage scripts are all under lib/mdev 2022-05-16 18:22:54 so they are moved 2022-05-16 18:23:01 ah, sorry, was being slow :-) 2022-05-16 18:23:21 minimal: i will have a closer look at the bats tests later 2022-05-16 18:23:30 ncopa: ok, thanks 2022-05-16 18:24:08 i am trying to clean up the setup-devd a bit. I think it uncovered a few unexpected things 2022-05-16 18:24:48 https://build.alpinelinux.org/ looks like builders are idle soon too, so I should probably prioritize tagging the rc1 2022-05-16 18:30:17 Riscv builder needs kick and restart 2022-05-16 19:02:06 ok! here comes rc1 2022-05-16 19:04:21 i had a look at setup-devd. there are a few surprises in there when switching between dev managers 2022-05-16 19:05:19 it is fixable, but I think we need a better testsuite for it 2022-05-16 19:07:06 but 3.16.0_rc1 is out at least! \o/ 2022-05-16 19:26:24 \o/ 2022-05-16 20:00:48 andypost[m]: I'll look at Zabbix tomorrow 2022-05-16 20:00:54 Have been a bit busy 2022-05-17 18:10:18 psykose: interesting, why did you remove that activate-linux thing? 2022-05-17 18:10:54 It would be great if commit messages could contain detail on why things happen, especially when packages are being removed 2022-05-17 18:12:53 I agree 2022-05-17 18:12:56 #13818 2022-05-17 18:15:07 was trying to upgrade gnome-feeds, turns out we need webkit2gtk-5.0 :p 2022-05-17 18:15:25 should i name the package webkit2gtk-5.0 (like arch) or webkit2gtk4 (like at least centos)? 2022-05-17 18:15:56 (my 5950x is already hard at works building it, so that i can make sure i got it right...) 2022-05-17 18:16:17 sadly, as i understand it, you can't build the gtk4 version alongisde the gtk3 version, so we need a separate package for it :( 2022-05-17 18:17:53 "WEBKIT_OPTION_DEFINE(USE_GTK4 "Whether to enable usage of GTK4 instead of GTK3." PUBLIC OFF)" 2022-05-17 18:24:38 at least the wayland port has proper licence 2022-05-17 18:36:53 ikke: ah thanks the reasoning for removing it make sense at least 2022-05-17 18:38:40 how do I Activate Linux®? 2022-05-17 20:05:51 I wonder how can one mint nfts with this software. m( 2022-05-17 20:06:56 Ermine: Make NFTs from screenshots I suppose 2022-05-17 20:07:29 "This includes the source code, binaries, screenshots, artifacts produced, websites, the name of the software, or anything else from or of this software." 2022-05-17 20:08:28 i don't think it is a good use of Alpine's limited master mirror space to distribute shitposts 2022-05-17 20:09:08 Ah, screenshots. 2022-05-18 14:25:21 im finding interesting bugs with the alpine-installer-testsuite 2022-05-18 14:25:47 btrfs on lvmsys with 2 disks fails 2022-05-18 14:27:41 but only with alpine-standard iso. with alpine-virt it passes 2022-05-18 14:30:02 ncopa: is disk via virtio_blk or virtio_scsi with alpine-virt? 2022-05-18 14:31:16 good question. I dont know really. It is with -drive if=virtio,file=disk.img 2022-05-18 14:36:58 I think that is virtio_blk then. You can specify "-device virtio-scsi-device" for virtio_scsi - might be worth checking if that works/fails. AFAIK virto_scsi uses a lot of the general SCSI/SATA code whereas virtio_blk has its own. 2022-05-18 14:41:30 im trying to reproduce it 2022-05-18 14:47:31 i was able to reproduce it but only had greaphical console in qemu. And only black screen due to driver not loaded in initramfs 2022-05-18 14:47:35 will try with serial now 2022-05-18 14:56:16 No space left on device 2022-05-18 14:56:24 i guess 512M ram is not enough 2022-05-18 15:07:46 overlaytmpfs? 2022-05-18 15:07:55 no 2022-05-18 15:08:18 i think update-kernel needs space for the entire kernel and things in /tmp 2022-05-18 15:08:47 aha 2022-05-18 15:09:14 but weird... it also fails with 2G ram 2022-05-18 15:09:19 in the test suite 2022-05-18 15:09:27 but when installing i manually, it works 2022-05-18 15:11:28 gzip: write error: No space left on device 2022-05-18 15:12:28 ncopa: you said this was for a lvmsys (i.e. Sys-mode) install, not run-from-ram so why is update-kernel being run? 2022-05-18 15:13:31 true 2022-05-18 15:13:37 maybe its the disk image? 2022-05-18 15:13:54 ==> initramfs: creating /boot/initramfs-lts 2022-05-18 15:13:54 gzip: write error: No space left on device 2022-05-18 15:14:06 that is when it fails 2022-05-18 15:17:34 is that actually /boot/initramfs (i.e. in memory) or in reality /mnt/boot/initramfs (on the disk via a chroot)? 2022-05-18 15:17:59 the disk via chroot. so it must be the disk size 2022-05-18 15:18:28 its only 1G disk 2022-05-18 15:18:48 right - which firmware packages are installed? it its the full set then that'll be large, plus some or all of the firmware will also be placed into the generated initramfs-lts file 2022-05-18 15:19:13 i think firmware is excluded smart 2022-05-18 15:19:43 it only includes the needed/used firmware and in qemu its linux-firmware-none 2022-05-18 15:19:59 ok, it passes when disk size is set to 2G 2022-05-18 15:21:07 try a "mkinitfs -l | grep firmware" to get an idea of what it put in there 2022-05-18 15:22:30 ha 2022-05-18 15:22:39 no, its uefi paritioning that is a bit stupid 2022-05-18 15:22:54 it uses 500M for efi partition 2022-05-18 15:23:01 but only use 360k 2022-05-18 15:23:16 then with swap partition 2022-05-18 15:23:25 its only 230MB left for root 2022-05-18 15:23:29 ncopa: yeah for my own custom-built images I currently have a 1MiB ESP partition (using FAT12) 2022-05-18 15:23:59 Grub doesn't put the kernel and intramfs in ESP, unlike Syslinux and so there's basically just a single grub EFI there 2022-05-18 15:24:32 ok! this is good 2022-05-18 15:25:26 ncopa: from memory the "issue" would be regarding what various UEFI implementations actually support 2022-05-18 15:25:47 I "think" Intel/Microsoft talk about that 500MB/MiB figure 2022-05-18 15:26:24 so there's a risk if you use FAT12 (or perhaps even FAT16) that some strange UEFI implementation out there might not boot off it 2022-05-18 15:26:55 5.3M /usr/lib/grub/x86_64-efi 2022-05-18 15:27:42 i think we could use 2MB by default 2022-05-18 15:28:05 or maybe 4 2022-05-18 15:28:56 so it seems that the smallest a FAT32 filesystem can be is 32MB (with 512 byte sectors) and 256MB (with 4K sectors) 2022-05-18 15:29:10 according to https://en.wikipedia.org/wiki/File_Allocation_Table#FAT32 2022-05-18 15:29:11 [WIKIPEDIA] File Allocation Table#FAT32 | "File Allocation Table (FAT) is a file system developed for personal computers. Originally developed in 1977 for use on floppy disks, it was adapted for use on hard disks and other devices. It is often supported for compatibility reasons by current operating systems for personal computers and many mobile..." 2022-05-18 15:30:42 the type of FAT filesystem (FAT12 / FAT16 / FAT32) is *governed* by its size - you cannot have a 2MB FAT32 filesystem for example 2022-05-18 15:35:49 so a 4MB sized device, mkfs.vfat will create FAT12? 2022-05-18 15:38:22 ncopa: for my own script I'm passing relevant params to mkfs.fat: https://github.com/dermotbradley/create-alpine-disk-image/blob/main/lib/disk-image-functions#L1712 2022-05-18 15:41:14 i'd like to keep the setup-disk simpler. not care about FAT12/FAT16. just let mkfs.vfat figure out the details 2022-05-18 15:42:18 not sure about mkfs.vfat as I haven't tried it for this 2022-05-18 15:43:41 " In reality most firmware only support FAT32" https://wiki.osdev.org/EFI_System_Partition#Format 2022-05-18 15:44:09 it only seems to handle FAT32 and so you won' be able to create a FS less than 32MB/256MB in size (depending on the sector size of the device) if Wikipedia is accurate about those figures 2022-05-18 15:46:54 a.img: DOS/MBR boot sector, code offset 0x3c+2, OEM-ID "mkfs.fat", sectors/cluster 4, root entries 512, sectors 8192 (volumes <=32 MB), Media descriptor 0xf8, sectors/FAT 6, sectors/track 32, serial number 0x4ac93f73, unlabeled, FAT (12 bit) 2022-05-18 15:47:13 its a 4MB disk image 2022-05-18 15:47:39 and it mentions "mkfs.fat", not "mkfs.vfat" 2022-05-18 15:59:07 4MB was not enought for cryptsys 2022-05-18 15:59:32 im gonna go for 36MB and enforce fat32 2022-05-18 16:00:38 cryptsys on a UEFI system? the ESP partition is always unencrypted 2022-05-18 16:00:58 root file system 2022-05-18 16:01:29 ah ok. I thought the 4MB was a reference to the size of the ESP partition's FAT fs 2022-05-18 16:02:07 the grub efi image becomes bigger than 4MB when rootfs is encrypted 2022-05-18 16:02:22 this testsuite is very useful 2022-05-18 16:03:33 FAILED tests/test_sys_install.py::test_sys_install[virtio-2-cryptsys-UEFI-xfs] 2022-05-18 16:07:24 ncopa: strange, looking at the logs from one of my own Grub/UEFI/LUKS builds the GRUB EFI file is 212992 in size, that's 213K in size 2022-05-18 16:08:39 and that's with a 1MB ESP partition 2022-05-18 16:10:23 though that's with a ext4 rootfs rather than xfs 2022-05-18 16:12:55 i think my local /tmp got exchausted due to I run 10 in parallel 2022-05-18 16:13:03 so 10 disks * 1G 2022-05-18 16:13:20 actually, some tests have dualdisk 2022-05-18 16:13:33 and I only have 32GB /tmp 2022-05-18 16:13:48 sorry, 16G 2022-05-18 16:24:01 ok good! its time for rc2 then 2022-05-18 16:34:37 ncopa: um, I think you missed an important point I made earlier regarding min FAT32 size related to sector size: "For Advanced Format 4K Native drives (4-KB-per-sector) drives, the minimum size is 260 MB, due to a limitation of the FAT32 file format. The minimum partition size of FAT32 drives is calculated as sector size (4KB) x 65527 = 256 MB." 2022-05-18 16:35:14 NVME drives may well have 4K sectors and some AF HDDs 2022-05-18 16:36:42 so the install would either need to have a common ESP partition size for all situations, or else to check the value of "/sys/block/nvme0n1/queue/physical_block_size" and size accordingly 2022-05-18 16:37:23 oops, "logical_block_size" value rather than "physical_block_size" 2022-05-18 16:43:23 ncopa: you could use these qemu options to test with a 4K drive: "-device virtio-blk,logical_block_size=4096,physical_block_size=4096" 2022-05-18 20:00:07 hum. 2022-05-18 20:00:35 so it might be safer to just set it to 262 or something 2022-05-19 06:48:29 good morning 2022-05-19 06:48:34 I'd like to do rc2 today 2022-05-19 07:02:17 ncopa, hi, is rc available as docker image? 2022-05-19 07:04:04 ncopa: Is there/will there be an ISO for the rc? I'd like to test the Virtualization.framework console fix. 2022-05-19 07:13:36 ktprograms: you can find them on the mirrors 2022-05-19 07:19:51 Ah, thanks! :) 2022-05-19 08:26:48 indy: not yet, but I should do that as well 2022-05-19 08:34:13 hey ncopa ! 2022-05-19 08:35:04 good morning, i kinda want to do things but don't know what to do 2022-05-19 08:36:38 maybe you could look over the issuesand see if there are any lowhaning fruits you could help solve? 2022-05-19 08:37:29 just responding to issues is helpful. it is demotivating to report issues if nobody responds 2022-05-19 08:39:21 oh we have a handful merge requests for mkinitfs 2022-05-19 14:14:32 I'm gonna add a setup-desktop script for v3.16 release: https://gitlab.alpinelinux.org/alpine/docs/user-handbook/-/issues/2#note_237093 2022-05-19 14:43:57 seems like xfce uses consolekit2 instead of elogind 2022-05-19 14:52:37 Still? consolekit2 is unmaintained upstream so that really shouldn't happen anymore 🤔 2022-05-19 14:52:50 yeah 2022-05-19 14:53:03 i don't know how to fix it though 2022-05-19 14:55:21 do we need to install an email client? does people still use email clients or only browser? 2022-05-19 14:57:02 We could let users install it themselves if they require one? 2022-05-19 14:57:33 yeah. thats what im thinking 2022-05-19 14:57:40 but i think we should install firefox by default 2022-05-19 14:58:08 and things like consolekit2 (or elogind) and gvfs 2022-05-19 14:58:12 so basic thigns works 2022-05-19 14:58:22 gvfs for trashcan for example 2022-05-19 14:59:05 is there not a meta package for xfce that can install such things? 2022-05-19 14:59:09 We do that with plasma and gnome 2022-05-19 14:59:14 what other "must have" apps should be installed? 2022-05-19 14:59:21 not really 2022-05-19 14:59:38 so if users want a bare bones install they can just install "plasma-desktop", but if they want some basic stuff installed they install "plasma" 2022-05-19 14:59:43 I think just a browser + the desktop should be enough 2022-05-19 14:59:44 Might be good to start doing that then? 2022-05-19 15:00:05 I hope that "firefox by default" will be "firefox-esr" ? 2022-05-19 15:00:10 well, we used to have a meta package, but i stopped with that because then you cannot remove those deps 2022-05-19 15:00:41 im doing `setup-desktop xfce` now, which pulls in the default packages 2022-05-19 15:00:45 That's why either the deps should be minimal, so just firefox, gvfs, etc, or something like optional deps should be added to apk (I'd like the latter very much personally) 2022-05-19 15:00:51 but then you can uninstall whatever you want afterwards 2022-05-19 15:01:21 right, i think we have an xfce4 package with absolutely minimal 2022-05-19 15:01:42 but things like xfce4-terminal etc is kinda good to have 2022-05-19 15:01:58 the idea is that setup-desktop give you a usable desktop with sane defaults 2022-05-19 15:01:59 even that should be removable imo 2022-05-19 15:02:23 thats why it is not pulled in as dependency, but installed with setup-desktop 2022-05-19 15:02:27 in that issue you linked you mentioned only GNOME and XFCE are supported. Can we not add KDE Plasma to it? 2022-05-19 15:02:33 makes sense 2022-05-19 15:02:56 ofc we shoudl do kde plasam 2022-05-19 15:03:08 but I had to start somewhere and i dont have time to do them all 2022-05-19 15:03:14 I can help there 😉 2022-05-19 15:03:22 (got a link to the source?) 2022-05-19 15:03:24 thats what I hoped for :) 2022-05-19 15:03:32 in a sec. will push 2022-05-19 15:03:59 Maybe Cogitri and/or Newbyte can help with the GNOME part 2022-05-19 15:04:14 @team/gnome :) 2022-05-19 15:06:49 https://gitlab.alpinelinux.org/alpine/alpine-conf/-/blob/master/setup-desktop.in 2022-05-19 15:18:07 PureTryOut: make sure to add a test for kde also in tests/setup_desktop_test 2022-05-19 15:19:12 oh good call, I was just about to submit a MR without one lol 2022-05-19 15:19:37 there seems to be no setup_desktop_test yet? 2022-05-19 15:19:38 git pull --rebase 2022-05-19 15:19:45 there is now :) 2022-05-19 15:19:52 oh I forked too early lol 2022-05-19 15:19:56 apk add kyua && make test 2022-05-19 15:21:40 protip: run `make iso` to create an iso that you attach to qemu (-drive file=alpine-conf.iso) 2022-05-19 15:22:36 qemu-system-x86_64 -m 4096 -enable-kvm -cdrom ~/Downloads/alpine-virt-3.16.0_rc1-x86_64.iso -drive file=alpine-conf.iso -serial stdio 2022-05-19 15:23:08 after login you paste in your serial console: mount LABEL=ALPINECONF /media/floppy/ && cp -rv /media/floppy/* / && setup-alpine -q 2022-05-19 15:24:53 and then you can run test your modifed setup-desktop script 2022-05-19 15:29:11 wow. gnome actually works... 2022-05-19 15:30:50 ncopa: any further thoughts on the /dev/disk/by-* stuff? 2022-05-19 15:31:24 just havent had time to take a closer look at it yet sorry 2022-05-19 15:35:02 gnome uses 1.5G 2022-05-19 15:35:06 its massive 2022-05-19 15:36:34 ah... it also installs evolution and stuff 2022-05-19 15:45:19 -rc2 tagged 2022-05-19 15:47:28 ncopa: still time to get package updates in before 3.16? 2022-05-19 15:48:43 new upstream cloud-init release came out yesterday and I'm just finishing up final local testing before raising a MR (in the new couple of hours) for the version pudate 2022-05-19 15:55:21 minimal: as long as it does not break anything 2022-05-19 15:55:35 im gonna do the release on monday 2022-05-19 16:00:50 nice 2022-05-19 16:01:07 my pipeline finished succesfully, if that means anything :) 2022-05-19 16:15:10 ncopa: when I was starting with Alpine, I would have appreciated a setup-desktop that got my DM, DE, audio, and user stuff settled. Put me in the right groups, get me udisks2, gvfs, get me the right login manager, password/secrets manager, etc. 2022-05-19 16:17:07 I'm still not sure I did any of that all correct, but my setup Mostly Works ™️ 2022-05-19 16:18:36 Like, what *Dev? Mdev? Udev? I dunno if I made the "right" choices, and having a canonical setup is a good reference point for users, if nothing else 2022-05-19 16:18:39 when i first touched alpine, my wish was it would never include setup-desktop :p 2022-05-19 16:20:10 i still think alpine is not the ideal OS for desktop usage, but you can make most of it work (except the proprietary stuff). 2022-05-19 16:20:40 Ah... Well, I am a desktop-oriented user, so different expectations, I suppose. I appreciate the Alpineness of Alpine, which is why I built my desktop on it, but it is super intimidating to start with different enough that researching how to roll your own fully functional desktop environment is tough 2022-05-19 16:20:53 and hope nothing breaks, like chrome, firefox and similar. 2022-05-19 16:22:03 Going on two years of it on desktop, and noting ever broke that wasn't my fault (or a temporary issue with deps on Edge). Alpine is stellar and bulletproof as a desktop OS 2022-05-19 16:22:30 same, works awesome on my laptop (i dont game on it toh so not missing muchZ) 2022-05-19 16:24:22 I do game a fair bit, but having solid flatpak support and discoverability via GNOME software for flatpak and APK made that transition much less painful than I thought it would be 2022-05-19 16:24:59 flatpak is cheating 2022-05-19 16:25:01 it somewhat works, but I don't care about alpine, I only really like apk 2022-05-19 18:18:45 panekj: $dayjob is working on a gnu/linux distribution that uses apk (and melange of course) 2022-05-19 18:28:17 Ariadne: I know, we spoke about it, still waiting for public release :P I also don't mind musl much since I'm mostly an AMD househould, I just need a little bit more than what Alpine is which is what I'm working on 2022-05-19 18:29:08 also still waiting for that IRC/zulip/discord/whatever you'll decide on 2022-05-19 18:32:55 Ariadne: may I ask why you choose apk? 2022-05-19 19:14:44 Newbyte: why wouldn't i choose apk? 2022-05-19 19:17:22 Ariadne: it's not the most robust package manager 2022-05-19 19:17:52 I disagree 2022-05-19 19:17:57 same tbh 2022-05-19 19:18:12 robust in what sense 2022-05-19 19:18:16 Depends on what you're doing of course 2022-05-19 19:18:24 sounds like a you problem 2022-05-19 19:18:27 But the replaces/provides feature does not seem very battle-tested 2022-05-19 19:18:29 Oh, really? 2022-05-19 19:18:59 the replaces/provides feature works exactly as it is designed to 2022-05-19 19:19:04 https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/32615 2022-05-19 19:19:08 so it's designed to be bad? :p 2022-05-19 19:19:17 it is not the apk-tools maintainers' problem that people do not use it as designed 2022-05-19 19:19:31 Can you explain how depending on this now removed package resulted in wlroots being installed? 2022-05-19 19:20:02 i am not tech support 2022-05-19 19:20:10 Phoc explicitly depended on wlroots-phosh, and yet apk would install the wlroots package. But only on new installations. Previous users would get wlrootsphosh 2022-05-19 19:20:17 i am not tech support 2022-05-19 19:20:20 * Phoc explicitly depended on wlroots-phosh, and yet apk would install the wlroots package. But only on new installations. Previous users who upgrade would get wlroots-phosh 2022-05-19 19:20:26 i am not tech support 2022-05-19 19:20:28 Yes, I know you are not tech support. But you claim that it's working as it should 2022-05-19 19:20:33 And it's clearly not 2022-05-19 19:20:45 then open a bug and attach a patch 2022-05-19 19:21:31 otherwise i suggest you take your commentary elsewhere 2022-05-19 19:21:43 Please 2022-05-19 19:21:46 You claimed it's a "me problem" 2022-05-19 19:21:54 You could have not done that and I would happily not have brought this up 2022-05-19 19:22:05 a little harsh for someone frustrated with an actual apk bug, no? 2022-05-19 19:22:14 it is not 2022-05-19 19:22:18 it is 2022-05-19 19:22:25 though yes, do open an actual issue as well 2022-05-19 19:22:33 that way someone could actually fix it 2022-05-19 19:22:41 or give more info on why it is the case 2022-05-19 19:22:56 irc message bug reports are a little less useful to fix bugs 2022-05-19 19:23:20 `wlroots-phosh` is not a package shipped in alpine 2022-05-19 19:23:35 Ariadne: yes, it was removed due to this bug and we solved it in a different way 2022-05-19 19:23:37 it.. literally was 2022-05-19 19:23:56 i was one of the people that ran into this same bug 2022-05-19 19:24:43 psykose: yeah, I agree that IRC bug reports suck. the issue is that I don't know how to reproduce it exactly so I don't feel confident I can write a good bug report about this 2022-05-19 19:24:54 "bug" 2022-05-19 19:25:06 Newbyte: it would still be worth a report 2022-05-19 19:25:33 you seem confident enough to say "apk is not a robust package manager" 2022-05-19 19:25:42 Ariadne: if apk gives people wlroots-phosh on upgrades but wlroots on new installations, is that predictable behaviour to you? 2022-05-19 19:25:50 maybe if you are not confident enough to file a bug, you should be not confident enough to talk shit 2022-05-19 19:26:59 Newbyte: no. if the dependency is explicitly `wlroots-phosh` then it should choose that always. but this is a trivial solver bug, and *all* package managers have them. 2022-05-19 19:27:20 Newbyte: i look forward to your apology for defaming apk 2022-05-19 19:28:05 why would someone have to apologise for saying something about some piece of software 2022-05-19 19:28:16 Newbyte: looking at it again phoc didn't have a depends= on it, and the makedepends doesn't count 2022-05-19 19:28:27 this would be solved with a soprefix= on the extra wlroots 2022-05-19 19:28:31 wow, so it was not an apk bug at all 2022-05-19 19:28:32 amazing 2022-05-19 19:28:45 oh, that's interesting. but I wonder why people upgrading got the right wlroots-phosh then? 2022-05-19 19:28:57 i forget the consistency rules 2022-05-19 19:29:06 like, this was only a problem for people doing new installations 2022-05-19 19:29:08 because when two equivalent dependency nodes are available, it picks whichever one is sees first 2022-05-19 19:29:09 but tl;dr both wlroots had the same so: provide, so which one you got was based on other things 2022-05-19 19:29:29 also, if wlroots-phosh is modified, it should use a different SONAME 2022-05-19 19:29:33 with an soprefix= they become not the same so:, and so things built against one cannot pull the other 2022-05-19 19:29:42 because it describes a different, and possibly incompatible, ABI 2022-05-19 19:30:03 this really applies to any multi-dependency thing and should be used liberally, if it's not guaranteed the soversions are unique in the different aport 2022-05-19 19:30:16 no, if you are taking a library and modifying it 2022-05-19 19:30:24 then you should change the SONAME 2022-05-19 19:30:34 unless it is the same ABI, in which case, why is it modified 2022-05-19 19:30:35 yes, and soprefix= is a really easy way to change that 2022-05-19 19:30:44 soprefix= does not change the SONAME. 2022-05-19 19:31:03 what i am saying is that wlroots-phosh should be libwlroots-phosh.so.0 or whatever 2022-05-19 19:31:50 so, in summary: 2022-05-19 19:32:28 - wlroots-phosh was improperly patched upstream (upstream should not collide SONAMEs with the upstream its forking) 2022-05-19 19:32:40 - wlroots-phosh was improperly packaged in alpine 2022-05-19 19:32:58 - the explicit dependency on wlroots-phosh was not actually an explicit dependency on wlroots-phosh 2022-05-19 19:33:16 - the combination of the three things above mean "apk is not a robust package manager" 2022-05-19 19:33:17 yes? 2022-05-19 19:33:23 is this the correct summary, newbyte? 2022-05-19 19:33:40 it seems that wlroots and wlroots-phosh come from same upstream which is wlroots 2022-05-19 19:34:07 then if they are the same, why two packages? 2022-05-19 19:34:15 pkgdesc="Modular Wayland compositor library patched to work with Phosh" 2022-05-19 19:34:21 so many things are wrong with this, but it's cool that we are blaming apk 2022-05-19 19:34:38 so if it is patched, then the SONAME should be patched too 2022-05-19 19:34:44 this really is basic shit 2022-05-19 19:35:07 yeah, it makes sense now. thank you for the explanation. I agree with the first three statements, although I do think that these intricacies could be better documented. I'm not sure where I would have found this. As for the fourth statement, I have many other issues with apk but I don't really want to continue this argument here as I suppose it's a subjective thing. 2022-05-19 19:35:26 if you hate apk why do you use alpine/pmOS 2022-05-19 19:35:34 I never said I hate apk 2022-05-19 19:35:34 why not go use a distribution with a package manager you like 2022-05-19 19:35:36 and leave us alone 2022-05-19 19:35:54 why are you so hostile 2022-05-19 19:35:58 and the reason it was patched is because phoc (the compositor used by phosh) relies on some undefined behaviour in wlroots that was changed, and so a commit has to be reverted 2022-05-19 19:36:14 Ariadne: I have a feeling you are taking this personally 2022-05-19 19:36:16 here's an idea: have you considered fixing phoc 2022-05-19 19:36:20 so the ABI and API is the same, but a small detail was changed 2022-05-19 19:36:42 yeah, but I am also busy with maintaining pmOS in general 2022-05-19 19:36:48 and maintaining GNOME here in Alpine 2022-05-19 19:37:08 and then I also have to balance that with my studies and so on. I can't do everything 2022-05-19 19:37:37 i'm not asking you to do everything, i am asking you to not imply that the apk solver was designed by idiots 2022-05-19 19:42:56 What I said is that I think there are package managers that are more robust than apk and I stand by that. apk could never be as robust as e.g. rpm since it doesn't check transactions before committing them, so the upgrade runs and if some package conflicts with another this results in the upgrade failing midway through and not before it starts. That's not to say rpm handles upgrades flawlessly (it doesn't), and it's also not to say that I 2022-05-19 19:42:56 think apk's solver was "designed by idiots". I do not think it is the most robust package manager in the world of computing, and that is all I said. If you want to continue discussing this then send me a direct message or something like that, but I don't want to keep talking about it here. 2022-05-19 19:43:15 wow, more incorrect assertions 2022-05-19 19:45:23 apk and rpm have the same transactional consistency, actually 2022-05-19 19:45:31 Newbyte, without any interest in whatever this conversation is, it might be helpful to know that "not the most X ever" is rarely interpreted literally - most people will interpret it as "bottom tier", not "anything other than the number 1" 2022-05-19 19:45:43 i'm sorry that we don't printf("Checking transactional consistency\n") like dnf does 2022-05-19 19:45:59 would it help if we add a printf 2022-05-19 19:45:59 Habbie: that is a very good point. I could have expressed myself better 2022-05-19 19:46:18 because so far this is just wrong point after wrong point after wrong point 2022-05-19 19:46:22 Ariadne: it would be helpful if apk told me that two packages conflict before the upgrade happens instead of midway through 2022-05-19 19:46:24 apk is by far *safer* than rpm 2022-05-19 19:46:29 IT DOES 2022-05-19 19:46:43 Newbyte, ack - just take that home for future interactions :) 2022-05-19 19:46:59 if you do not provide the solver with the necessary data to make the conclusion that the transaction introduces breaks 2022-05-19 19:47:02 i cannot help you 2022-05-19 19:48:25 Habbie: appreciate it :) 2022-05-19 19:49:18 Newbyte, np :) 2022-05-19 19:49:26 every time you do *anything* with apk, it solves for /etc/apk/world before committing anything 2022-05-19 19:50:06 this is a stronger guarantee than anything rpm does (nothing, actually), and stronger than the dnf/yum consistency check 2022-05-19 19:50:09 https://ariadne.space/2021/04/25/why-apk-tools-is-different-than-other-package-managers/ 2022-05-19 19:51:18 of course, if packagers do not properly annotate the dependency requirements of their packages, this will result in problems, but that has nothing to do with apk. if you do not provide correct inputs for dnf's solver, for example, it will fail to solve correctly too. 2022-05-19 19:52:46 this is all painstakingly documented in the apk manual pages 2022-05-19 20:22:23 at any rate, if anyone wants to know what apk *actually* does when you do upgrades or adds or deletes, https://ariadne.space/2021/10/31/spelunking-through-the-apk-tools-dependency-solver/ 2022-05-19 20:22:55 Newbyte: hopefully this article will provide some insight on how apk is actually sufficiently robust 2022-05-19 20:23:16 apk should read minds 2022-05-19 20:23:43 sam_: it should pick a random solution based on quantum entanglements 2022-05-19 20:23:59 if we install a quantum DPU in every pmOS phone 2022-05-19 20:24:08 i think we could have a good solution 2022-05-19 20:26:34 this sounds like a true enterprise feature 2022-05-19 20:26:36 much more like RPM 2022-05-19 20:27:00 RPM or RPM5 2022-05-19 20:27:02 My problem with apk is that APKINDEX downloading phase slows down over time 2022-05-19 20:27:16 over... time? 2022-05-19 20:27:27 do you mean as APKINDEX grows larger? 2022-05-19 20:27:37 because APKv3 indices help a lot there :D 2022-05-19 20:28:34 Idk, apk update works fast on fresh installations, but after some time, it begins to take seconds to load APKINDICES on the same system 2022-05-19 20:29:34 ... you sure that it's really as fast on a fresh install, and it's not to do with "I installed alpine 5 years ago and it was fast, but now it ain't"? 2022-05-19 20:30:57 Ermine: what kind of storage are you using? 2022-05-19 20:31:21 HDD 2022-05-19 20:31:32 have you tried e2defrag 2022-05-19 20:32:06 No, but I will, thank you 2022-05-19 20:33:42 you might make sure to defrag /lib/apk/db/installed also 2022-05-19 20:34:19 by necessity, it can get large 2022-05-19 20:35:02 What is excessive for apk update? On my Celeron craptop, it is 4s or so on < 100MBps WiFi. That... seems negligible to me. Certainly faster than APT on the same hardware 2022-05-19 20:50:52 Newbyte: the GNOME software backend for apk is dangerously broken, but that isn't apk's fault, libapk is not being used correctly and does not consider the upgrade transaction as a single transaction, but rather each package is upgraded individually 2022-05-19 20:52:30 Saijin_Naib: disk operations? apkv2 db is ascii text file which might be a bit slow in certain cases 2022-05-19 20:54:38 Ariadne: yeah, cogitri mentioned something like that before 2022-05-19 20:57:17 Yay, defrag apparently worked! 2022-05-19 21:00:38 Anyway, apk is indeed really fast (after zypper, it feels superluminal) and clever 2022-05-19 21:03:59 panekj: yeah thats why i suggested defrag, the "read one line at a time" behavior when parsing indices and db is likely horrible if the file is fragment 2022-05-19 21:04:18 ikke: well, this is highly dangerous and no wonder Newbyte concluded apk was so broken 2022-05-19 21:04:23 at any rate https://gitlab.alpinelinux.org/alpine/aports/-/issues/13825 2022-05-19 21:09:55 Saijin_Naib: thank you for useful feedback. Do you mind test setup-desktop in alpine 3.16.0_rc2? 2022-05-19 21:10:06 I think udisks2 is a good suggestion 2022-05-19 21:10:33 I'm working on setup-user now, which will set up the groups etc 2022-05-19 21:13:23 Does pam also belong, or does that already come along with something else? 2022-05-19 21:13:48 I'll try to setup a VM to test, yes 2022-05-19 21:21:48 i hope `setup-desktop` does not install this gnome-software-plugin-apk 2022-05-19 21:21:52 it is highly dangerous 2022-05-19 21:22:48 Ariadne, ikke: Yes, g-s expects that packages can be upgraded individually 2022-05-19 21:23:10 Cogitri: yes, it is fine if that is what is requested, but "upgrade all" must upgrade all 2022-05-19 21:23:29 upgrading each package individually in alphabetical order is not safe 2022-05-19 21:24:30 upgrading individual packages may also result in other packages being upgraded as side effect 2022-05-19 21:24:42 +1 2022-05-19 21:25:39 ACTION misses the day when "setup-alpine" set up networking, an ssh server, and an ntp client. 2022-05-19 21:26:21 nangel: it still does 2022-05-19 21:26:42 setup-alpine does not (yet) ask you for desktop env 2022-05-19 21:27:14 :) yes, I know... I just remember the old days. :) 2022-05-19 21:27:56 cute trick to test it (unless you have very slow internet): 2022-05-19 21:28:05 qemu-system-x86_64 -m 2048 -enable-kvm -serial stdio -cdrom http://dl-master.alpinelinux.org/alpine/v3.16/releases/x86_64/alpine-virt-3.16.0_rc2-x86_64.iso 2022-05-19 21:28:35 setup-alpine -q && setup-desktop xfce && adduser someuser && openrc 2022-05-19 21:29:56 Unknown protocol 'http' 2022-05-19 21:30:09 qemu must be built with curl backend. 2022-05-19 21:30:35 apk add qemu-block-curl 2022-05-19 21:31:33 btw, i dont think it makes any sense to ask for keyboard layout when logged in via serial console 2022-05-19 21:31:55 i need to figure out how to detect if /dev/console is serial 2022-05-19 21:32:18 but not tonight. good night! 2022-05-19 21:32:31 o/ 2022-05-19 21:33:08 REALLY dumb question, but does it matter if it is running under qemu? 2022-05-19 21:48:07 Oh shit I merged something I shouldn't have... Reverting 2022-05-19 21:50:18 all the builders were busy so it should pick up no diff 2022-05-19 21:50:19 lucky 2022-05-19 21:50:29 otherwise, you would need to bump rel on it and most of it would be built anyway 2022-05-19 22:03:29 yeah :/ 2022-05-20 00:40:22 nals 2022-05-20 00:40:24 er 2022-05-20 02:00:14 ncopa: that trick is freaking awesome. setup-desktop worked nicely, as far as I can tell. Simple and no-fuss, just like the other setup-* scripts I've used from Alpine 2022-05-20 05:17:45 nangel: no. it doesnt matter 2022-05-20 05:18:09 good mornin 2022-05-20 06:26:51 ncopa: using that example qemu command you gave, how would I make the storage persistent? after a reboot it just seems to boot into the live image again but I rather have it automatically start the display manager it just setup in the setup-desktop script 2022-05-20 06:34:23 oh I guess I can just run `openrc` 2022-05-20 06:34:36 there is no window popping up though, but the right services run 2022-05-20 06:53:39 ah derp installing a qemu-ui-* package would help ofc 2022-05-20 06:57:26 Hello ppl, I've been struggling with some Rockchip SoCs (3328a armv7) TV-boxes recently 2022-05-20 06:58:01 I was disappointed with Armbian (i,e, systemd and all the bloat) 2022-05-20 06:58:34 so I've made an Alpine 3.15 "frankenstein" hack that runs on them 2022-05-20 06:59:46 I have used the 4.4.194 kernel provided by rockchip folks, packed a .modloop with the right modules 2022-05-20 07:00:46 it works pretty well, now my only real purpose with those lame SoCs is to make mpv run on them with full video hardware acceleration 2022-05-20 07:01:12 for h.264 streams via ffmpeg 2022-05-20 07:02:00 I have no idea where to start ...any pointers ? 2022-05-20 07:03:26 mainline kernel would help 🙈 2022-05-20 07:04:16 I know, but it seems mainline is lacking support for the hardware required 2022-05-20 07:04:57 i.e. if I use a different kernel I lose the hooks for the "rockchip media package" I need to get video hardware accel working 2022-05-20 07:10:37 those €9 tv-boxes are a total pain, but I have like, 6 of them running armbian and it totally sucks. Wasn't on Debian for some time, it's a bloatware nightmare 2022-05-20 07:12:38 ncopa: the current gnome setup-desktop entry installs the `gnome` meta package, but wasn't the point of this script to not need such a meta package so the user can uninstall things not strictly necessary should they want too? 2022-05-20 07:17:25 PureTryOut: correct. I just based it on https://gitlab.alpinelinux.org/alpine/alpine-conf/-/merge_requests/62 2022-05-20 07:17:47 hmm ok, that should probably be changed then 2022-05-20 07:17:58 yeah. i woudl appreciate help with that 2022-05-20 07:18:15 I just added the deps of the plasma meta package as a list in this script, as those apps are not strictly necessary but help making a functional out of the box experience 2022-05-20 07:19:02 PureTryOut: to make it peristant, add a -drive file=diskimage.img option and run `setup-disk` 2022-05-20 07:19:29 (no need for persistency anymore, I just didn't knew I could run `openrc`) 2022-05-20 07:19:47 👍 2022-05-20 07:20:23 so how minimal should the minimum set of applications be? Like I can just install a terminal, or install kde-applications-base which pulls some more basic applications in like a file manager and a PDF reader 2022-05-20 07:27:58 I guess I'll let it install the bare minimum most people will expect 2022-05-20 07:38:20 ncopa: uh tests are failing while they succeeded yesterday (I just rebased), but I see no error explaining what went wrong 2022-05-20 07:38:39 https://gitlab.alpinelinux.org/PureTryOut/alpine-conf/-/jobs/725693 2022-05-20 07:48:20 oh derp fixed, I was missing a \ 2022-05-20 08:43:15 PureTryOut: i'd say its up to the desktop env maintainer to decide how minimal it should be 2022-05-20 08:43:51 for xfce i just install xfce4-terminal and firefox and thats it. should probably do udisks2 too if its not already pulled in 2022-05-20 08:43:54 Makes sense, ok. Made a MR for KDE Plasma support 2022-05-20 08:47:32 what do you think about sorting of th elist? https://gitlab.alpinelinux.org/alpine/alpine-conf/-/merge_requests/68/diffs#note_237206 2022-05-20 08:47:46 its non-critical though 2022-05-20 08:48:37 minimal: cloud-init makes alpien slow to start due to 2022-05-20 08:34:42,526 - url_helper.py[WARNING]: Calling 'http://10.0.2.2/latest/meta-data/instance-id' failed [21/120s]: request error [HTTPConnectionPool(host='10.0.2.2', port=80): Max retries exceeded with url: /latest/meta-data/instance-id (Caused by 2022-05-20 08:48:37 NewConnectionError(': Failed to establish a new connection: [Errno 111] Connection refused'))] 2022-05-20 08:48:53 how can i make it only use NoCloud provider? 2022-05-20 08:50:28 ncopa: that sorting is fine by me, updated the MR with it 2022-05-20 08:51:30 awesome! thank you! <3 2022-05-20 08:51:56 Np! Thank you for making this easier for the user 😄 2022-05-20 08:55:20 i know its a bit silly but i'm actually a bit excited about this setup-desktop thingy :) 2022-05-20 08:58:00 its kinda nice 2022-05-20 08:58:30 easy desktops while keeping base small since its optional 2022-05-20 09:02:54 No space left on device 2022-05-20 09:02:58 thank you plasma... 2022-05-20 09:07:04 wow. plasma is actually pretty nice. first time i test it in years 2022-05-20 09:07:13 except that the mouse pointer is gone 2022-05-20 10:02:45 ncope, PureTryOut: on the topic of the setup-gnome-desktop. Duplicating the dependencies and the maintenance between the gnome meta-package and setup-desktop feels unnecessary. Could we either trim the gnome package to get only completely necessary things in (possibly by creating more subpackages) or directly remove it and keep track of basic dependencies in setup-desktop? 2022-05-20 10:02:59 s/ncope/ncopa/ 2022-05-20 10:02:59 pabloyoyoista_ meant to say: ncopa, PureTryOut: on the topic of the setup-gnome-desktop. Duplicating the dependencies and the maintenance between the gnome meta-package and setup-desktop feels unnecessary. Could we either trim the gnome package to get only completely necessary things in (possibly by creating more subpackages) or directly remove it and keep track of basic dependencies in setup-desktop? 2022-05-20 10:03:57 for KDE Plasma I'm going towards the latter 2022-05-20 10:04:36 ncopa: mouse pointer gone? Can not reproduce 🤔 And yeah it's pretty nice 😄 We have some weird Alpine/musl only bugs though, some of them have workarounds in our packages applied 2022-05-20 10:05:29 most annoying thing for me personally is akonadi_imap crashing and taking the rest of akonadi with it (thus also taking down calendar and addressbook and what not). The crash seems to be Alpine/musl specific sadly but I haven't been able to get a good backtrace yet 2022-05-20 10:06:03 pabloyoyoista_: Would calling something like `apk info -R gnome` in setup-desktop to get the deps be a good idea? 2022-05-20 10:06:56 Sorry, didn't notice I was on my alt nick (ktprogra1) 2022-05-20 10:08:08 ktprograms: That could also work, yes. Would have to confirm with Cogitri and Newbyte though. Since I have the feeling that the gnome package mostly existed because the lack of a setup-desktop 2022-05-20 10:08:43 Keeping the gnome metapackage could still be useful for users who prefer to install stuff without using setup-desktop. 2022-05-20 10:11:46 Well, those people could just install their dependencies manually if setup-desktop is the default :P Rethinking about it, the main problem with the gnome metapackage I have is that users will find it annoying when we change dependencies. For example, gedit got lately superseeded by gnome-text-editor. Once we package the former, we should replace the gedit dependency in the metapackage. That means suddenly people will 2022-05-20 10:11:46 get their text editor replaced without asking for it 2022-05-20 10:11:58 Instead of us just updating the default 2022-05-20 10:12:18 Makes sense. 2022-05-20 10:17:06 That’s the default behavior on other distros too though 2022-05-20 10:17:26 You mean getting your editor replaced? 2022-05-20 10:22:34 Yup 2022-05-20 10:22:50 Or rather packages that aren’t explicitly requested might be replaced 2022-05-20 10:23:02 Do apt recommended packages get changed when the meta package changes? (For example looking at ubuntu-desktop) 2022-05-20 10:23:45 So it's "expected behaviour" to replace apps when the metapackage changes? 2022-05-20 10:24:57 I guess that also makes sense in some way. But then I'd like to review the gnome metapackage to trim it a bit more 2022-05-20 10:26:02 I'll put that in my TODO-list :D 2022-05-20 11:11:37 ncopa: it doesn't appear if you move it around for a bit? 2022-05-20 11:14:17 wow, 'dune --help' looks uggly, not sure I've used that before and it's a regression or has always been like that... 2022-05-20 11:16:43 npm from main to community seems like a good move 2022-05-20 11:41:19 could be qemu something. no it does not appear when I move it around 2022-05-20 11:41:30 it works outsize qemu window but not inside 2022-05-20 11:44:40 Is it up to distro maintainers to add other DEs to setup-desktop? MATE and LXQT are striking me as missing, and both are certainly fully-usable (and nice, IMO) on Alpine 2022-05-20 12:27:32 wyk72: i suspect nobody here knows that kind of stuff. i think #postmarketos may know, or may be able to point you to someone who knows 2022-05-20 13:36:48 any suggestions on where to go with https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/34488? the runner at least needs over 1h to build aarch64 kernel even if I force-push fix for current conflicts, and I'd also like to know why it got put on hold (as no explicit reason was given) 2022-05-20 15:17:01 minimal: hi 2022-05-20 15:17:14 minimal: cloud-init makes alpine slow to start due to 2022-05-20 08:34:42,526 - url_helper.py[WARNING]: Calling 'http://10.0.2.2/latest/meta-data/instance-id' failed [21/120s]: request error [HTTPConnectionPool(host='10.0.2.2', port=80): Max retries exceeded with url: /latest/meta-data/instance-id (Caused by 2022-05-20 15:17:14 NewConnectionError(': Failed to establish a new connection: [Errno 111] Connection refused'))] 2022-05-20 15:17:39 minimal: any idea how to make cloud-init use nocloude provider only? 2022-05-20 15:24:56 ncopa: modify the /etc/cloud/cloud.cfg file and place the following in there: datasource_list: ['NoCloud'] 2022-05-20 15:27:56 ncopa: which DataSource was the above WARNING referring to? 2022-05-20 15:28:28 not sure. the default from setup-cloud-init 2022-05-20 15:34:12 ncopa: I'm guessing you didn't give it a NoCloud/ConfigDrive iso/fs for metadata so it then started going through the full set of DataSources (which expect a Metadata Server to exist) 2022-05-20 15:34:49 exactly 2022-05-20 15:35:02 I just booted it in qemu for debugging 2022-05-20 15:37:21 yeah cloud-init is typically used to create images tailored for specific DataSources (i.e. Cloud Providers) 2022-05-20 15:38:33 admin user. should be in 'admin' or 'wheel' group? 2022-05-20 15:42:53 ncopa: for cloud-init use? that depends on sudo/doas config 2022-05-20 15:43:41 no, this is not about cloud-init 2022-05-20 15:54:25 ncopa: re how to detect if console is serial, how about "cat /sys/dev/char/5:1/active", that returns the name of the tty device that /dev/console is mapped to 2022-05-20 15:54:44 ah! nice! 2022-05-20 15:54:58 very userful 2022-05-20 15:55:03 will tweak that when I get time 2022-05-20 15:55:06 now its weekend 2022-05-20 15:55:11 so you'd get "ttyS?" for a serial and "tty?" for a typical console and I guess "hvc?" for a virtio console 2022-05-20 15:55:22 thats perfect 2022-05-20 15:57:25 I added that to my notes 2022-05-20 15:57:30 have a nice weekend everyone! 2022-05-20 15:57:40 on monday we do 3.16.0 release 2022-05-20 15:58:15 nice weekend :) 2022-05-20 15:58:52 ncopa: the EFI sizing change, you went with physical_block_size, should be logical_block_size AFAIK 2022-05-20 16:00:34 ugh... have you tested it? i wasnt able to get the virtual thing working within reasonable time. the -device nvme,physical_block_size didnt make any diff 2022-05-20 16:01:56 ncopa: basically physical_block_size is the real on-disk layout, logical_block_size is what I tells the OS (I have a HDD here which is AF and has 512/4K logical/physical) 2022-05-20 16:02:39 so I believe (correct me if I'm wrong) that all mkfs.fat will care about is the logical sector size 2022-05-20 16:03:31 ncopa: I'll do some testing over the weekend with qemu 2022-05-20 16:20:21 Ariadne: what do you think of !34494 2022-05-20 16:20:26 just some binutils s390x backports 2022-05-20 16:21:40 yes seems good 2022-05-20 16:21:49 i am glad IBM is stepping up to fix rust :) 2022-05-20 16:22:11 me too :) 2022-05-20 17:07:10 nice, now the ppc binutils failures are on rust 1.61.0 as well :) https://gitlab.alpinelinux.org/alpine/aports/-/jobs/726283 2022-05-20 18:24:47 andypost[m]: trying to see what the current state of zabbix on php8 isd 2022-05-20 18:30:25 ikke: thank you, I used to try apply patches without success even strict type's needs manual edits. I bet it easy to get what is broken atm and in early June next release should work even with 8.1 2022-05-20 18:31:20 But we have to move php7 before release 2022-05-20 18:31:24 yes 2022-05-20 18:32:57 And today next release managers are selected) https://wiki.php.net/todo/php82 2022-05-20 18:38:51 So I now get the issue that I reported earlier, trying to upgrade to the latest zabbix version to see if it changed 2022-05-20 18:39:08 It happens for me when I visit /zabbix.php?action=service.list 2022-05-20 18:46:47 andypost[m]: I think this is what is causing the actual issue: "PHP message: PHP Fatal error: Uncaught TypeError: array_slice(): Argument #3 ($length) must be of type ?int, string given" 2022-05-20 18:59:33 andypost[m]: ok, that was a simple fix 2022-05-20 18:59:57 Going to add a patch and then merge your MR 2022-05-20 19:04:01 Oh great 👍 I thought it will need to upgrade to RC version but there's no tarball 2022-05-20 19:04:30 There might still be issues, but no obvious issues 2022-05-20 19:07:03 ah nice, there is now a copy branch name button on gitlab 2022-05-20 19:35:55 andypost[m]: it has been merged 2022-05-20 22:52:04 Ikke: thanks! Where to update release notes for php https://gitlab.alpinelinux.org/alpine/tsc/-/issues/40#note_237348 2022-05-21 05:53:25 andypost[m]: https://wiki.alpinelinux.org/wiki/Release_Notes_for_Alpine_3.16.0 2022-05-21 06:37:31 ncopa: we need to start on the releasea notes 2022-05-21 06:38:51 ncopa: Typo in https://gitlab.alpinelinux.org/alpine/alpine-conf/-/blob/master/setup-user.in#L14 (s/Cerate/Create/) 2022-05-21 07:14:32 andypost[m]: ah, apparently Zabbix fixed the issue by disabling strict_types 2022-05-21 08:34:01 ktprograms: thanks. will fix 2022-05-21 08:34:43 im experimenting with magic keywords in ssh setup, to expand to https gitlab.a.o/user.keys 2022-05-21 08:35:47 so if I type 'al ncopa' it will make https://gitlab.alpinelinux.org/ncopa.keys the default in the question 2022-05-21 08:36:23 this makes it possible to just type 'al ncopa' to get the keys from alpine gitlab 2022-05-21 08:36:28 example: 2022-05-21 08:36:38 Allow root ssh login? ('?' for help) [prohibit-password] al ncopa 2022-05-21 08:36:38 Allow root ssh login? ('?' for help) [https://gitlab.alpinelinux.org/ncopa.keys] 2022-05-21 08:48:44 ncopa: Sounds like a good idea. Are you going to include keywords for gitlab.org and github.com? Maybe also codeberg.org as it's the most popular Gitea instance (I think) 2022-05-21 08:59:27 ikke: I find it weird that zabbix relaxing language instead of fixing bugs 2022-05-21 08:59:57 I suppose they wanted a quick-fix for 6.0 / 6.2 and then fix it more robustly later 2022-05-21 08:59:59 but no idea 2022-05-21 10:18:46 ktprograms wouldnt be a url argument be easier in that case? 2022-05-21 10:20:59 Misthios: The point is that there would be a short keyword. ncopa's idea, not mine. Just suggesting additional keywords. 2022-05-21 10:21:09 Ah 2022-05-21 10:43:50 Hi, I want to test rc iso http://dl-master.alpinelinux.org/alpine/v3.16/releases/x86_64/alpine-virt-3.16.0_rc2-x86_64.iso but I get a cert error if I try to switch to https 2022-05-21 12:16:19 use dl-cdn 2022-05-21 12:18:33 ohh, great, thanks! 2022-05-21 12:26:21 _rc3 tagged 2022-05-21 12:27:32 I did some improvements for ssh config. if you set up a user account you will not get the question about root logins, as it is assumed that you use the non-root user with doas 2022-05-21 13:14:27 ncopa: got physical_block_size/logical_block_size working with QEMU via virtio-blk - seems like QEMU doesn't support it for NVME device. Doing further testing 2022-05-21 13:18:11 so, it seems like xelatex wants to have icu-data-full installed, should I add this as a dep? 2022-05-21 13:19:20 c7s: You could create an MR, then malibu can determine whether it's required 2022-05-21 13:23:19 ikke: right, thanks 2022-05-21 16:01:12 I wonder if setup_crypt() should fail if there is a passphrase missmatch or it would be better to ask for it again, it's pretty annoying that you need to redo the whole setup process 2022-05-21 16:02:57 i think it used to ask three times and fail if you typed it wrong a single time 2022-05-21 16:03:19 I mean in setup-disk script 2022-05-21 16:04:19 https://gitlab.alpinelinux.org/alpine/alpine-conf/-/blob/master/setup-disk.in#L1211 2022-05-21 17:27:44 i know 2022-05-21 17:27:55 oh, it does ask three times 2022-05-21 17:27:59 i thought that was fixed 2022-05-21 18:33:50 ncopa: Re sizing of ESP vfat partition with 4K device sectors, the following works for QEMU testing: "-device virtio-blk,drive=data,logical_block_size=4096,physical_block_size=4096 -drive file=extra-disk.qcow2,if=none,format=qcow2,id=data" 2022-05-21 18:34:14 "fdisk -l" will then correctly show 4096/4096 sector sizes 2022-05-21 18:35:43 however there is an "issue" with Busybox's mkfs.vfat, I don't think in general it correctly handles disk size vs FAT type. It is certainly behaving differentl to dosfstools' mkfs.vfat for the same partition 2022-05-21 18:38:09 on a 4096/4096 "disk" I created a GPT partition label with a 50MB partition and then formatted it using both busybox and dosfstools - dosfstool decided to make it FAT16 whereas busybox makes it FAT32, when I force dosfstool to create a FAT32 (via "-F32") then output includes: "WARNING: Number of clusters for 32 bit FAT is less than suggested minimum." 2022-05-21 18:39:51 I think using busybox's mkfs.vfat there's a risk of creating a FAT32 partition (especially on devices with 4K logical sector size) that might not be compatible with some machines' UEFI implementations 2022-05-21 18:41:48 you can use "file -s" (e.g. "file -s /dev/vda1") to see the type of FAT filesystem created 2022-05-21 18:53:14 minimal: i've had issues with busybox mkfs.vfat in general, using mkfs.fat (dosfstools, i think?) instead for when alpine-cloud-images builds UEFI images 2022-05-21 18:55:55 tomalok: yes, a couple of days ago ncopa wanted to shrink the size of the UEFI ESP partition and I pointed out it needed to be FAT32 for cross-platform compatibility, and also pointed out that the min FAT32 size differs for devices with 512 byte and 4K sectors sizes. He added some code into setup-disk for this - the above is the feedback of my testing to illustrate actual issues 2022-05-21 21:16:41 donoban: file an issue on that. Sounds relatively easy to fix 2022-05-21 21:16:55 I have a problem with Carla, but I think it might be in pipewire because I have the same problem with calfjackhost. The problem is that it most of the time crashes at startup when it loads all audio processing plugins into pipewire (jack interface). If I pin Carla to one specific CPU core it works. Anyone have an idea on how to debug this issue or have seen this problem? 2022-05-21 21:17:31 I've registered an issue upstream for Carla at https://github.com/falkTX/Carla/issues/1616 2022-05-21 21:18:06 ncopa: you saw my earlier messages about the UEFI sizing? 2022-05-21 21:25:12 minimal: I saw. Very helpful, thank you. I’m not by the computer rn but we should add a test for it in alpine-installer-testsuite. Could you create an issue for it so I don’t forget? 2022-05-21 21:26:55 ncopa: alpine-conf issue? or aports issue? The physical->logical thing certainly should be fixed before release 2022-05-21 21:33:59 alpine-conf. What happens currently it you try install it? Will it create too small ESP? 2022-05-21 21:49:51 ncopa: I have not tested setup-disk directly but one concern is that BB's mkfs.vfat in general might create an "invalid" FAT32 fs which then potentially may not work (i.e. boot) on some (physical) machines. The phy/log issue is more that setup-disk may create a larger ESP partition than required (there are 512 log/4096 phy drives out there and AFAIK only the logical size is relevant) 2022-05-21 23:01:32 re: the draft Release Notes for 3.16, the utmp section, who decided to put that section is? As, in summary, utmp is *not* working correctly yet in Alpine, I've spent time off-and-on over the past months (including talking with skarnet) to try and sort it out 2022-05-22 10:17:15 ncopa: ok 2022-05-22 11:59:06 dotnet6-build keeps hanging on aarch64 2022-05-22 12:01:47 Or maybe it just takes long? 2022-05-22 12:09:41 Ok, it does still continue 2022-05-22 13:54:35 should we also mention in the release notes that python2 was removed? 2022-05-22 13:54:42 would probably be good to add that 2022-05-22 13:55:08 probably 2022-05-22 14:28:46 also don't think we have anything that wants py2 even to build 2022-05-22 14:36:29 can't we just say python is pyhton3 by default now? 2022-05-22 14:36:45 instead of having to add a package as a dependency for everything that needs it? 2022-05-22 14:38:00 sure, that also works 2022-05-22 14:38:41 i'd rather do that right after 3.16 though 2022-05-22 14:39:10 probably more than one build out there that sees a 'python' and then assumes some 2.7 version 2022-05-22 15:28:35 archlinux has been running with python == python3 for a very long time now, but they also patch a lot most likely 2022-05-22 15:42:57 i mean it would be fine 2022-05-22 15:43:04 we would just have to fix a few stray builds 2022-05-22 15:43:11 nothing needs it now 2022-05-22 15:43:16 nod 2022-05-22 15:43:23 and yes, makes sense to wait until after 3.16 2022-05-22 15:44:21 the last things were.. anki, which just needed 'python' 2022-05-22 15:44:31 qt6-qtweb (6.3.0 now has new chromium and needs 3) 2022-05-22 15:44:37 some random thing that didn't build and i disabled 2022-05-22 15:44:49 and.. that's it. the last things is py-setuptools which is just py2 itself 2022-05-22 15:45:14 i am up for already deleting it 2022-05-22 15:45:20 since it's testing/ anyway 2022-05-22 17:54:39 psykose: qt6-qtwebengine still has python2 in makedepends but that should be fix with the next release 2022-05-22 17:54:49 it already is 2022-05-22 17:55:41 ah, you changed it yesterday 2022-05-22 17:55:43 nice 2022-05-22 17:57:21 well, now i have to also actually fix the build, since i forgot to change the chromium patches 2022-05-22 17:57:22 but yes 2022-05-23 08:25:16 good morning 2022-05-23 08:25:19 lets do release today 2022-05-23 08:26:26 i will fix https://gitlab.alpinelinux.org/alpine/alpine-conf/-/issues/10514 first though 2022-05-23 08:31:27 rnp is failing a test on ppc64le (timeout) 2022-05-23 09:06:41 psykose: gtk4.0 can be merged now :) 2022-05-23 09:55:52 On release day? 😶 2022-05-23 09:58:11 yolo 2022-05-23 10:19:13 all builders seem idle now 2022-05-23 10:21:02 ok. im cleaning up the setup-* scripts a bit. need to resist fix everything 2022-05-23 10:23:51 and adjusting the tests also 2022-05-23 10:23:56 so they still pass 2022-05-23 10:28:37 I don't have a lot of time today myself to help 2022-05-23 10:30:49 ok. will do my best doing it myself then 2022-05-23 10:31:00 what is the status of release notes? 2022-05-23 10:31:17 argh! we should have sent question for release notes earlier 2022-05-23 10:32:00 There is some collected here: https://wiki.alpinelinux.org/wiki/Release_Notes_for_Alpine_3.16.0 2022-05-23 10:39:59 !34684 2022-05-23 10:45:35 wrt release notes, what makes for a significant update? 2022-05-23 10:46:52 Something affecting a lot of users for example 2022-05-23 11:44:50 finding last minute bugs 2022-05-23 11:45:00 and i think we should include this: https://gitlab.alpinelinux.org/alpine/mkinitfs/-/merge_requests/104 2022-05-23 11:47:10 could this !34730 enter on 3.16? 2022-05-23 12:52:30 ncopa: regarding Release Notes, the utmp section makes no sense as utmp support is not working 2022-05-23 12:54:10 not sure we want to mention it in the summary anyway. doubt many people care about utmp 2022-05-23 13:12:09 why is sway 1.7 a significant update? https://wiki.alpinelinux.org/wiki/Release_Notes_for_Alpine_3.16.0 2022-05-23 13:13:25 i dont use sway myself, but is user requried to do something when upgrading sway? will new version break backwards compatiblity in some way? 2022-05-23 13:14:14 looks like user don't need to do anything on upgrading it: https://github.com/swaywm/sway/releases/tag/1.7 2022-05-23 13:14:50 it is the least painful way to use Wayland, so I guess it is "significant" because people using Wayland would like to know it's up to date in the new Alpine release? 2022-05-23 13:15:53 I guess the same thing could be said of hundreds of packages in alpine 2022-05-23 13:17:34 hundreds, maybe even thousands! 2022-05-23 13:17:56 and i dont think we should try list them all 2022-05-23 13:18:08 because then will significant changes drown in the list 2022-05-23 13:18:33 significant changes is when user may need to do something when upgrading 2022-05-23 13:18:46 like rebuild all your locally built python/ruby modules 2022-05-23 13:18:56 backup/restore your postgresql database 2022-05-23 13:18:57 etc 2022-05-23 13:18:58 I don't think it's a significant update, the only "breaking" change is the default config, and basically everyone copies that config to modify it, and it's not like the package depended on alacritty anyway 2022-05-23 13:19:38 i think maybe even qemu can be removed from significant updates 2022-05-23 13:19:54 the upgrade was pretty smooth, for me at least 2022-05-23 13:21:36 ncopa: I'd like utmp support to be mentioned in (some version's) Release Notes *once* it is working - there have been at least occasional issues raised in the past about utmp as often people using Alpine for a server OS do expect things like who/w/last/uptime to indicate logins 2022-05-23 13:22:43 minimal: I'll get to busybox testing this week 2022-05-23 13:22:53 Are the installer changes on that page? (Setup-user,desktop) 2022-05-23 13:23:23 i am adding a few words on the installer scripts yes 2022-05-23 13:23:41 Oki 2022-05-23 13:24:43 I'm gonna tag rc5 while preparing the release notes with the latest fixes 2022-05-23 13:24:54 and then I'm gonna do 3.16 i think 2022-05-23 13:32:17 release notes proposal: https://gitlab.alpinelinux.org/alpine/infra/alpine-mksite/-/merge_requests/45 2022-05-23 14:10:21 we should mention this in release notes: https://gitlab.alpinelinux.org/alpine/aports/-/issues/13420 2022-05-23 14:10:22 ncopa: should we mention that by default xen does not include qemu in 3.16, but if needed xen-qemu can be installed? 2022-05-23 14:10:45 does it affect users who upgrade? 2022-05-23 14:13:12 I have not tested upgrade just initial install, but as xen-qemu is now optional, users of any features provided by qemu in alpine 3.15 problably should install xen-qemu 2022-05-23 14:49:56 i think it might be enough to just mention xen 4.16 2022-05-23 15:11:37 yes the release notices for Xen 4.16 gives a hint on this 2022-05-23 15:41:09 does this look ok? https://wwwtest.alpinelinux.org/posts/Alpine-3.16.0-released.html 2022-05-23 15:42:01 maybe we should note that pyton2 has been removed (moved to testing)? 2022-05-23 15:42:35 oh yes.. 2022-05-23 15:43:13 go 1.18? 2022-05-23 15:43:43 👍 2022-05-23 15:44:47 the notice from sudo was left from the previous notes 2022-05-23 15:44:57 'significant changes' 2022-05-23 15:47:22 i think the move was done after 3.15 release. Previous release ntoes said that it will move in the future 2022-05-23 15:48:13 There's a small typo in the Various improvement section, Adminitrator should be Administrator 2022-05-23 15:48:21 ncopa: ah, right 2022-05-23 15:48:32 first deprecated, now removed 2022-05-23 15:48:57 durrendal: thanks 2022-05-23 15:49:36 NP, excited for the 3.16 release :) 2022-05-23 16:12:07 On the release notes, the split of subpackages in networkmanager has been taken out from the version in the wiki. Might that be important? It could leave some users without internet after the upgrade, couldn't it? 2022-05-23 16:14:18 hm, yes it could 2022-05-23 16:17:17 Should the shadow-login change be in "significant changes" or "upgrade notes"? 2022-05-23 16:21:52 networkmanager comment added 2022-05-23 16:22:24 i think upgrade notes is good for shadow-login 2022-05-23 16:23:26 "significant changes" are also for people who often install alpine. things that now work differently 2022-05-23 16:23:49 for new installs the shadow-login does not matter 2022-05-23 16:24:05 but its good that users upgrading are aware of it 2022-05-23 16:24:17 "Possibility to add ssh keys" => maybe add a link to some documentation? (sounds interesting to me, but I have no idea how it works) 2022-05-23 16:24:57 heh.. 2022-05-23 16:25:56 and I was so content with myself for adding tests for all those new features. now i get reminded that docs are missing... :) 2022-05-23 16:26:27 :) 2022-05-23 16:27:36 and the coolest features are a bit hidden 2022-05-23 16:28:03 when prompted for ssh key (as normal user) you can type 'al', 'gh', or 'gl' 2022-05-23 16:28:40 and it will give you a new default with gitlab.a.o/user.keys, gihub.com/user.keys or gitlab.com/user.keys 2022-05-23 16:29:03 so if your username matches your github account you do: gh 2022-05-23 16:29:34 and for the root account you do: gh mygithubuser 2022-05-23 16:29:40 Ok, so re: shadow-login there is no user action require during upgrade, but users should be aware that they can switch to util-linux-login if they want. 2022-05-23 16:30:17 xordspar0: there is a link to an 'alarming' issue there, which says 'it may break things so users cannot login' 2022-05-23 16:30:34 but there is also a comment sayin that it should all work 2022-05-23 16:30:48 so the idea to mention it is to 2022-05-23 16:31:17 make users aware that the change is there, and if things goes wrong they know where to look 2022-05-23 16:31:25 but i dont think it will go wrong for anyone 2022-05-23 16:32:20 the thing is that your currently working shadow-login may get replaced by util-linux-login 2022-05-23 16:32:50 Ok, thanks for explaining. 2022-05-23 16:44:28 ncopa: thank you for the NetworkManager fix :) 2022-05-23 16:48:00 than you 2022-05-23 16:50:20 ok the time has come to do the tag... 2022-05-23 16:52:36 3.16.0 is in the oven 2022-05-23 16:56:11 podman should be added to the release notes. In 3 -> 4 there is user action required regarding networks. 2022-05-23 16:57:22 https://podman.io/releases/2022/02/22/podman-release-v4.0.0.html 2022-05-23 17:03:20 " To ensure that we don’t break existing users, the old CNI stack will remain the default on existing installations, while new installs will use Netavark." 2022-05-23 17:09:22 re: release notes: we should mention that python2 was removed 2022-05-23 17:09:43 I would also tone down the utmps section a bit as it not fully working afaik 2022-05-23 17:09:58 https://wwwtest.alpinelinux.org/posts/Alpine-3.16.0-released.html 2022-05-23 17:10:07 python2 is mentioned in "deprecation" section 2022-05-23 17:10:16 oh, sorry. I was looking at the wiki page 2022-05-23 17:10:20 i dont mention utmps at all, since it not working 2022-05-23 17:10:34 yea 2022-05-23 17:10:58 ncopa: I did see that in the blog post, but nevertheless my network did break. 2022-05-23 17:11:15 time to push all the stuff that I had on hold for edge :D 2022-05-23 17:11:30 yeah, 3.16-stable is branched now 2022-05-23 17:11:33 \o/ 2022-05-23 17:11:38 \o/ 2022-05-23 17:11:55 \o/ 2022-05-23 17:12:06 \oo/ 2022-05-23 17:12:32 \ooo/ 2022-05-23 17:13:52 \:D/ 2022-05-23 17:22:49 I'm trying to figure out some nifty oneliner to see what packages were upgraded to what based on 'git diff 3.15-stable 3.16-stable main community' 2022-05-23 17:25:13 omni: you prbably want to do that per package 2022-05-23 17:29:33 in the early days i produced a pkgdiff with that info 2022-05-23 17:29:42 but that was only for the packages that was included on the iso 2022-05-23 17:31:09 ok to push to prod? https://wwwtest.alpinelinux.org/posts/Alpine-3.16.0-released.html 2022-05-23 17:37:09 mm, the linked wiki page still has "TODO(jirutka)" comments 2022-05-23 17:37:53 todo: todo 2022-05-23 17:38:49 todos los todos 2022-05-23 17:38:54 the changelog is in general a bit shorter compared to what we had on the wiki but it looks good to me 2022-05-23 17:39:51 todos los todos van ser todificados este noche 2022-05-23 17:40:30 jirutka said he woudl fix those TODOs this evening 2022-05-23 17:40:57 ok. im merging it 2022-05-23 17:42:36 nmeum: that's usually the case, we try to keep the published release notes brief 2022-05-23 17:42:54 *nod* 2022-05-23 17:44:06 ncopa: on sudo note: "sudo was moved" or "sudo has been moved" 2022-05-23 17:45:33 On doas: you can write it this way: 'If you can use doas-sudo-shim so you can type sudo as before" 2022-05-23 17:48:12 im not native english, but "sudo was moved" sounds better to me? 2022-05-23 17:48:50 I'm not native english speaker too :] . But here you need passive voice anyway 2022-05-23 17:51:06 ok 2022-05-23 17:51:36 I think "Suggested replacement is `doas` or `doas-sudo-shim`." is enough 2022-05-23 17:52:14 "If you can use doas-sudo-shim so you can type sudo as before" that sounds terrible imo 2022-05-23 17:53:09 Also, you may want to change 1) nvme to NVMe and ssh to SSH; 2) llvm to LLVM, nodejs-current to Node.js-current, ruby to Ruby and rust to Rust 2022-05-23 17:53:37 lowercase gang 2022-05-23 17:54:10 second point is for consistent capitalisation of package names on the page 2022-05-23 17:54:36 well, there is a difference between package names and project names 2022-05-23 17:54:47 there is no 'python' package. its 'python3' 2022-05-23 17:55:00 nodejs-current is a package name 2022-05-23 17:55:23 Ok 2022-05-23 17:55:31 but I 'm ok either way 2022-05-23 17:56:21 im gonna take a walk and do the rest in a bit 2022-05-23 17:56:30 https://gitlab.alpinelinux.org/alpine/aports/-/issues/13838 2022-05-23 17:58:01 I'll work on invalidating latest-stable and updating pkgs.a.o in a bit 2022-05-23 18:03:19 panekj: I proposed this alternative because current one sounds like doas and doas-sudo-shim are alternatives to each other, but doas-sudo-shim is meant to be used with doas 2022-05-23 18:03:38 Btw added that php81 also shipped via community 2022-05-23 18:03:48 https://wiki.alpinelinux.org/wiki/Release_Notes_for_Alpine_3.16.0#php7_moved_to_testing 2022-05-23 18:04:35 Btw looks I missed to update priority but it doable later 2022-05-23 18:07:44 In 'Credits' section: "Thanks to everyone *sending patches*, bug reports, new and updated aports, and to everyone helping with writing documentation, maintaining the infrastructure, or *contributing* in any other way!" 2022-05-23 18:08:54 am I the only one having issues (b"bwrap: Can't make symlinks at...) with the new bubblejail? 2022-05-23 18:15:38 b"bwrap: Can't make symlink at /home/omni: File exists\n" 2022-05-23 18:17:24 I've tried removing everything related to the instance (manually, since there is no command for that, right?) and re-create it 2022-05-23 18:23:50 latest-stable has been purged on fastly 2022-05-23 18:40:05 Importing 3.16 on pkgs.a.o now 2022-05-23 18:44:05 secdb too 2022-05-23 18:44:53 It should look at releases.json 2022-05-23 18:45:07 But let me check to be sure 2022-05-23 18:46:54 But also secfixes-tracker 2022-05-23 18:47:38 Ermine: thank you for usefule feedback! I have updated https://wwwtest.alpinelinux.org/posts/Alpine-3.16.0-released.html 2022-05-23 18:47:46 andypost[m]: I added php 8.1 to highlights 2022-05-23 18:49:36 ncopa: thank you! Probably I misread earlier, sorry 2022-05-23 18:49:59 no. i just added it after your reminder 2022-05-23 18:50:42 ncopa: lgtm, thank you! 2022-05-23 18:52:30 When I'm updating provider-priority for 3 packages, what should be in commit message? 2022-05-23 18:55:02 community/{pkg1,pkg2,pkg3}: update provider priority 2022-05-23 18:55:14 and short explanation on why it was updated 2022-05-23 18:55:19 unless it is evident 2022-05-23 18:57:02 Ariadne: secfixes-tracker has been updated 2022-05-23 18:57:09 Also deployed 0.3.3 2022-05-23 18:57:18 kk 2022-05-23 19:00:50 Congrats on the 3.16 release guys! 2022-05-23 19:18:27 thanks! \o/ 2022-05-23 19:18:47 andypost[m]: thank you for the php 8.1 ping: https://twitter.com/synth_ruiner/status/1528817365617725442 2022-05-23 19:20:07 it just hit me how many ppl have contributed to aports... 296 says git shortlog 2022-05-23 19:20:24 ACTION throws new confetti 2022-05-23 19:20:36 s/new/more/ 2022-05-23 19:20:36 Ermine meant to say: throws more confetti 2022-05-23 19:20:40 \o/ 2022-05-23 19:20:45 so... what happened to ovmf in 3.16.0? 2022-05-23 19:21:08 RuntimeError: Unable to find ovmf APK via https://dl-cdn.alpinelinux.org/alpine/v3.16/community/aarch64 2022-05-23 19:21:26 what 2022-05-23 19:22:51 looks like it's only x86_64 now? 2022-05-23 19:23:03 ncopa: 😊 congratulations 🎉 even bombs and rickets can't stop alpinists) 2022-05-23 19:23:09 it didn't build 2022-05-23 19:23:45 tomalok: looks like it got renamed to aavmf 2022-05-23 19:24:20 5f9ea92a3bc5c730daaebd393206558bfdc8256c 2022-05-23 19:25:35 looks like aavmf is now (also) what i'm looking for :) will tweak the alpine-cloud-images builder to pull that in for aarch64 UEFI 2022-05-23 19:25:58 i wonder if we should have a provides=ovmf for backwards compat 2022-05-23 19:26:33 install path is different so it'd still not be 1:1 2022-05-23 19:28:11 ok 2022-05-23 19:37:22 okay, cloud images are on their way ... 2022-05-23 19:38:19 tomalok: how ~long will it take? 2022-05-23 19:40:36 So, Alpine starts journey towards 3.17? 2022-05-23 19:40:49 more or less 2022-05-23 19:41:23 abuild *pretty please* 2022-05-23 19:42:37 ikke: would normally have the images built locally by now -- but i seem to have some issues with Packer being able to SSH into QEMU - maybe due to recent setup changes, ncopa? i don't have time to dig deep right now -- lunchtime's ending, will take a look after DayJob 2022-05-23 19:43:17 ok, then I won't wait for the MR 2022-05-23 19:43:41 (uploading and publishing images to regions is what takes the most time, usually) 2022-05-23 19:48:15 speaking of images, is there anything preventing a riscv iso or rootfs from being added? 2022-05-23 19:48:22 (not trying to add more work, just curious) 2022-05-23 19:50:01 https://dl-cdn.alpinelinux.org/alpine/edge/releases/riscv64/ 2022-05-23 19:50:08 edge snapshots 2022-05-23 19:58:11 congratulations all on 3.16.0 2022-05-23 19:58:18 take a nice moment to breathe now :) 2022-05-23 20:01:12 I think ncopa is well underway doing that now :) 2022-05-23 20:01:50 ikke: <3 thank you I missed that 2022-05-23 21:01:01 now that 3.16 is branched, i guess it's time to rebase our gcc on 11.3 2022-05-23 21:01:06 or 12.1 if we want more adventure :) 2022-05-23 21:09:25 12.1 is quite calm so far 2022-05-23 21:09:27 nothing like 10 or 11 2022-05-23 21:09:29 just missing includes as usual from libstdc++ changes 2022-05-23 21:09:50 https://bugs.gentoo.org/show_bug.cgi?id=gcc-12 2022-05-23 21:09:53 most stuff just works 2022-05-23 21:10:30 fun 2022-05-23 21:11:24 much needed after 10 -fcommon and 11 std c++ change 2022-05-23 21:36:40 Llvm changes also looks challenging 2022-05-23 21:38:29 which ones 2022-05-23 21:43:37 At least "ptr" changes will affect all backends 2022-05-23 21:49:26 I mean https://llvm.org/docs/OpaquePointers.html 2022-05-23 22:32:13 ok, I think the bubblejail 6.0 issue may only affect a very small number of people 2022-05-23 22:51:02 *0.6.0 2022-05-23 22:59:57 what if the real home is '/home/user'? https://github.com/igo95862/bubblejail/commit/ff93af02ec5ee50b838d6d64ecc102d6699983a8#diff-1fc8cef1334a240cd8b5f26f51a9fa0089f13a34f66b39a1b9f70a1787e5e64bR312 2022-05-23 23:00:36 and I get this https://github.com/igo95862/bubblejail/commit/f3149ed4f4ff16b96eb3fc978cf99bd0ba37748c 2022-05-23 23:00:58 but I quite liked that bubblejail tried to hide those 2022-05-23 23:01:40 did you actually have 'user' as your home dir 2022-05-23 23:01:41 haha 2022-05-23 23:02:52 yes, and I'm sorry, for some reason I masked that fact by modifying my paste earlier 2022-05-23 23:04:16 that is a top class name :) 2022-05-23 23:04:22 sounds like you should make an upstream issue 2022-05-23 23:04:30 yeah 2022-05-24 01:26:57 andypost[m]: aiui opaque pointers are not actually enabled yet? 2022-05-24 01:27:16 although it sounds like probably they will be for 15 2022-05-24 01:32:09 yes 2022-05-24 09:14:06 How does 'report abuse' button on gitlab profile page work? 2022-05-24 09:15:28 admins get a list of reports 2022-05-24 09:16:20 This button says 'Already reported for abuse' on one of the pages, but I didn't click this button 2022-05-24 09:17:15 either you did or something is broken somewhere or only one person can report someone or.. who knows 2022-05-24 09:18:37 maybe someone else reported the user already? 2022-05-24 09:18:57 i don't see why it would lock 2022-05-24 09:19:23 sounds silly if you think about it 2022-05-24 09:19:39 it does sound silly 2022-05-24 12:02:47 hi all, 3.16 apk has problems with jfrog 'file format error', i remember it was discussed here, but can't find issue for apk-tools 2022-05-24 12:03:15 ah, this again 2022-05-24 12:03:36 I'm got it right - setlocale() has no effect on ctype functions if licale is missing? Even install of icu-libs-full does not help... For #13845 2022-05-24 12:04:03 icu-* and locale() are completely unrelated 2022-05-24 12:04:42 https://www.jfrog.com/jira/browse/RTFACT-26821?filter=-4&jql=text%20~%20%22apk%22%20order%20by%20created%20DESC containes merge request which is probably already in 2022-05-24 12:05:38 indy: https://gitlab.alpinelinux.org/alpine/apk-tools/-/merge_requests/100 2022-05-24 12:05:48 pretty sure it's this one, if i remember correctly 2022-05-24 12:06:17 oh, it's linked in the jira ticket you posted 2022-05-24 12:06:43 andypost[m]: yeah, seems it's not affected by locale 2022-05-24 12:07:25 musl locale support has.. some issues in my experience, wouldn't be surprised this is just entirely gone and i'm not missing anything 2022-05-24 12:09:09 psykose: thanks, I forgot about musl-locales... But then it points a bug - setlocale should return error for missing 2022-05-24 12:09:33 it's not related to it being missing or not, the code you linked just does not check the 'locale' 2022-05-24 12:10:47 I thought it should skip when locale is missing https://github.com/php/php-src/blob/master/ext/ctype/tests/lc_ctype_inheritance.phpt#L5 2022-05-24 12:12:47 it's not missing probably 2022-05-24 12:38:09 ncopa, https://gitlab.com/libosinfo/osinfo-db/-/merge_requests/455 2022-05-24 12:38:30 Added standard flavor to armv7 2022-05-24 13:03:31 psykose: so better to skip the test for musl) https://github.com/php/php-src/issues/8616#issuecomment-1135832578 2022-05-24 13:42:16 ncopa: alpine-cloud-images build issue turned out to be the old '-c' flag on setup-sshd... removed it and the Packer qemu boot_command worked again... didn't get VNC attached to qemu to see what was really going on, but perhaps it's not quite setting non-interactive with '-c'? 2022-05-24 13:45:19 hum... should still have compat with the -c option 2022-05-24 13:45:46 ah 2022-05-24 13:45:58 yeah. i think it got into interactive mode 2022-05-24 14:05:38 has anyone had a go at firefox-developer-edition 2022-05-24 15:23:06 hey, is there a way for random kde packages like kdenlive to not pull in the whole qtwebengine just for "accounts integration"? 2022-05-24 16:36:12 ptrc: We‘d need package variants/options for that and I don’t think we will have that in the foreseeable future 2022-05-24 16:47:26 aw, i was wondering if this could be somehow split and loaded dynamically 2022-05-24 17:17:30 ddevault: is there any need for it? 2022-05-24 18:39:39 https://wiki.alpinelinux.org/wiki/Release_Notes_for_Alpine_3.16.0 → there is a remaining `TODO(jirutka)` :) 2022-05-24 20:59:49 Since btree is not avaible for postfix on alpine, what should I use for smtpd_tls_session_cache_database? 2022-05-24 21:00:01 Is lmdb suitable for that? 2022-05-24 21:01:10 Apparently not: http://postfix.cs.utah.edu/LMDB_README.html 2022-05-24 21:43:44 Ikke: I used to https://www.postfix.org/SQLITE_README.html 2022-05-24 21:44:25 But storage always has issues 2022-05-24 23:57:51 I use lmdb and a grep through the logs and lines like this are the onl ones with full `cache lmdb:/var/lib/postfix/postscreen_cache full cleanup: retained=328 dropped=5 entries` 2022-05-25 00:14:25 i'm confused why it needs a fancy database for a tls session cache at all 2022-05-25 00:16:20 ah but yeah, lmdb seems fine 2022-05-25 00:18:14 doesn't support tlsmgr ofc, but not sure what that gives you 2022-05-25 00:19:59 wow, spam: https://gitlab.alpinelinux.org/explore/snippets 2022-05-25 00:23:58 thats alot of spam! 2022-05-25 00:35:07 not spam, but a list of close to all aports upgraded 3.15-stable -> 3.16-stable: https://gitlab.alpinelinux.org/-/snippets/196 2022-05-25 01:20:42 ncopa: could you merge !24236 ? there's nothing wrong with the actual change, and ondemand makes a better default than powersave, and there is also a bug currently where you can be stuck on powersave forever due to the governors being missing after boot 2022-05-25 02:06:26 Setting default CPU scaling governor to ondemand is a good choice. I can confirm, a few month ago as total newbie, that took me a while to realize that the rpi4 stuck on 600MHz. Actually I thought of it as a bug. And it took me a further while to find out how to unleash the real power of the rpi4 with setting the CPU scaling governor to ondemand. 2022-05-25 02:12:18 iirc normally the cpu has a range of frequencies and a boost (in a generic sense), and powersave is supposed to at least go a little above the lowest but never into boost 2022-05-25 02:12:37 at least that is what it seemed to do on my laptop, which has <1ghz base, 2.something max and 4+ boost 2022-05-25 02:12:46 on the pi for some reason this is completely broken though and it's just at 600 forever 2022-05-25 02:13:07 which is double annoying with https://github.com/raspberrypi/linux/issues/4875 2022-05-25 02:13:20 without manually upgrading to that kernel there is no way at all to go past 600, hah 2022-05-25 02:15:13 ah, no 2022-05-25 02:15:19 that behaviour was conservative 2022-05-25 02:15:25 powersave is indeed always-lowest 2022-05-25 02:15:39 conservative as in the 'conservative' governor 2022-05-25 02:16:30 and perhaps schedutil is a better choice than ondemand 2022-05-25 02:23:44 For my rpi4 with alpine 3.15 (5.15.32-0-rpi4) and Firmware Version Jan 20 2022 13:56:48 I can not confirm the bug https://github.com/raspberrypi/linux/issues/4875. CPU scaling governors are available and work fine. 2022-05-25 02:29:05 now try the latest .41 kernel from 3.16/edge 2022-05-25 02:37:03 For the default a setting conservative or ondemand or schedutil is from my point of view more user friendly as powersave. 2022-05-25 02:37:06 According https://www.kernel.org/doc/html/latest/admin-guide/pm/cpufreq.html scheutil - "This governor generally is regarded as a replacement for the older ondemand and conservative governors, as it is simpler and more tightly integrated with the CPU scheduler, its overhead in terms of CPU context switches and similar is less significant, and it 2022-05-25 02:37:06 uses the scheduler’s own CPU utilization metric, so in principle its decisions should not contradict the decisions made by the other parts of the scheduler." 2022-05-25 02:37:13 I never used schedutil so far but I justed changed it now to check it out. 2022-05-25 02:42:29 it is generally better, yeah 2022-05-25 02:42:36 and i agree, which is why i want to get that merged now 2022-05-25 07:11:08 panekj: yeah, I'm working on it 2022-05-25 08:01:06 if you just want the dev version of it i could do it, since i maintain firefox anyway 2022-05-25 10:05:14 ptrc, any plans for apk-tools update to 2.12.10? 2022-05-25 10:14:14 indy: dunno, i'm just a contributor 2022-05-25 12:42:10 ddevault: !34780 2022-05-25 12:42:21 psykose: nice, thanks 2022-05-25 12:42:51 works for me; but do test the ci artifacts first :) 2022-05-25 12:43:48 probably not today, but I will try to test it tomorrow 2022-05-25 12:44:06 sure thing 2022-05-25 13:50:08 got someone else to test it too so i merged it, but do give feedback when you get to it :) should work just like normal firefox, and they do work side-by-side 2022-05-25 15:09:07 indy: i'll probably push for that in the next week or so, there's a couple of things i want to backport from apk3 2022-05-25 15:33:32 Ariadne, i tried build apk-tools from git but unsuccessfully 2022-05-25 15:33:47 built it on what 2022-05-25 15:34:02 and with which build system 2022-05-25 15:34:04 i think i tried 3.15 2022-05-25 15:35:45 abuild with midified apkbuild 2022-05-25 15:36:51 the way you build apk-tools from git is to build it by hand 2022-05-25 15:37:01 you're gonna have to be a little more specific than that, like actually posting what you did and the logs 2022-05-25 15:55:28 none the less, gitlab CI shows it is presently building 2022-05-25 15:55:32 ACTION shrugs 2022-05-25 16:55:18 fyi i plan to push gcc 12 this weekend. 2022-05-25 17:00:23 \o/ 2022-05-25 17:01:54 ( https://www.gnu.org/software/gcc/gcc-12/changes.html tl;dr vectorization, and shadowstack on aarch64) 2022-05-25 17:17:19 awesome \o/ 2022-05-25 17:19:18 oh, gcc has CFI support via -fcf-protection 2022-05-25 17:19:26 since ages ago 2022-05-25 17:19:39 well, since it's based on CET, not that long ago I guess 2022-05-25 17:19:53 i already suggested enabling it at https://gitlab.alpinelinux.org/alpine/aports/-/issues/13246 2022-05-25 17:20:19 > Note that -fcf-protection is incompatible with -mindirect-branch (which is used to implement retpoline). In such cases it is recommended to disable -fcf-protection. 2022-05-25 17:23:19 i don't trust the ubuntu wiki 2022-05-25 18:33:42 if I would develop an aports package like main/linux-legacy, would there be interest in having that upstream? I have an original (i586) Pentium with less ram than the system requirement for Alpine. I am trying to develop this with the tinyconfig option. 2022-05-26 02:45:38 eloy_: such a package would be welcome in community, but probably not main 2022-05-26 10:26:41 psykose: firefox-developer-edition works great, thanks! 2022-05-26 10:50:09 :) 2022-05-26 10:51:44 unfortunately, it does not make developing against firefox any more pleasant 2022-05-26 11:09:51 i think the problem with this "linux-legacy" is that different people will have different ideas of what is "optional" 2022-05-26 11:10:13 alpine's existing kernel packages already have this issue 2022-05-26 11:56:08 Ariadne: I see, thanks 2022-05-26 12:42:49 Ariadne, wjat 2022-05-26 12:43:03 ops..sorry. what's the difference between apko and witchery ? 2022-05-26 13:17:50 "The ordinary user account created during setup couldn't log in from a graphical session, because the setup program hadn't created a home directory." https://www.theregister.com/2022/05/26/alpine_linux_316_released/ 2022-05-26 13:17:59 anybody know what that is about? 2022-05-26 13:18:08 I could log in when I tested it in vm 2022-05-26 13:26:24 ncopa: no idea, suggest you click on the "Corrections" link at the bottom of the article to message the author for more info 2022-05-26 13:29:04 i suspect they did a diskless install, rebooted, and then did setup-desktop 2022-05-26 13:29:14 in that scenario the /home is wiped 2022-05-26 13:39:30 ncopa: so should the setup script do a lbu save at diskless install time to prevent that? 2022-05-26 13:49:00 minimal: exactly my thought as well 2022-05-26 13:49:34 the thing is that /home is not included. the setup script should add $HOME/.ssh at least 2022-05-26 13:49:43 not sure we want add entire /home 2022-05-26 14:14:57 ncopa: doing "lbu add /home/*/.ssh" does expand to put individual entries for each user into /etc/apk/protected_paths.d/lbu.list 2022-05-26 15:04:54 fcolista: apko is more refined, does not require docker, etc 2022-05-26 15:05:26 but i have no objection to somebody taking over witchery if they want to 2022-05-26 15:06:48 I've watched your video about apko and melange, but after having read your blog post 'bout witchery 2022-05-26 15:07:10 So I was trying to figure how that was linked to apko+melange 2022-05-26 15:07:49 Probably I'll ping you since I'm still wondering how to definetly setup a build environment with ci and multiarch 2022-05-26 15:38:03 ncopa: why adding .ssh to lbu save but other config files not? I would prefer that home directories (for non system users) are recreated if booting in diskless mode 2022-05-26 19:10:17 liske: yeah. good point 2022-05-26 19:12:01 But any change would need to be explicitly committed with lbu 2022-05-26 20:18:53 Ariadne: will you change the gcc x86 baseline to i686 now, given the sse2 update? 2022-05-26 21:00:44 shouldn't it be pentium4 2022-05-26 21:01:01 if sse2 is truly required 2022-05-26 21:07:05 now that i think about it more, i suspect other distros also do this sort of "i686... sort of" business 2022-05-27 00:09:34 psykose: what update are you referring to? 2022-05-27 00:09:54 https://gitlab.alpinelinux.org/alpine/tsc/-/issues/35 2022-05-27 00:12:03 oh :/ 2022-05-27 00:12:15 so this one is no longer valid then https://gitlab.alpinelinux.org/alpine/tsc/-/issues/20 2022-05-27 00:15:06 apparently so 2022-05-27 00:15:19 i think we were one of the last ones with i586 as the baseline 2022-05-27 00:16:15 even debian is 'i686' (sse2) 2022-05-27 00:16:18 since debian9 2022-05-27 00:16:22 ah 2022-05-27 00:16:25 no sse in debian 2022-05-27 00:16:26 hm 2022-05-27 00:16:39 no idea what the numbers mean, really, everyone seems to have wildly different ideas 2022-05-27 00:16:42 yeah, I specifically choose Alpine for my retro 586 laptop. Wanted something more modern than Slackware and avoid Gentoo 2022-05-27 00:36:31 https://github.com/mirh/opemu-linux this seems to be a thing, can't find it for SSE2 and earlier though 2022-05-27 00:39:08 that is cursed beyond belief 2022-05-27 00:39:11 :D 2022-05-27 06:32:22 What do you guys think about including a chroot script to mount /dev and such on the boot iso 2022-05-27 06:33:10 not a terrible idea 2022-05-27 06:33:32 i have the shitty commands burned into my head but it's nice to save 15 seconds 2022-05-27 06:41:52 +1 to that 2022-05-27 11:46:17 are packages that don't need SSE2 still going to be compiled for 586? then a custom legacy kernel package might still make sense, to use it as an irc client for example. Don't need to have large toolchains 2022-05-27 13:14:51 it's not the kernel that needs sse2 2022-05-27 13:21:22 Hello71: I know, but if all the packages are going to be compiled for SSE2 in a future release, that kernel is useless. 2022-05-27 13:21:34 i guess 2022-05-27 13:31:56 psykose: -march=pentium4 yes 2022-05-27 13:32:11 it will be done as part of gcc 12 update 2022-05-27 13:32:28 thanks 2022-05-27 13:33:10 if somebody wishes to maintain an unofficial i586 alpine port they are free to do so, and i will be glad to help them with the infrastructure / setup / whatever but it’s time for us to move on 2022-05-27 13:36:07 that's fair enough :) 2022-05-27 13:39:48 So alpine is going to run only on pentium 4 or later cpu? 2022-05-27 13:53:53 Ermine: yes, it was deemed an inefficient use of our limited resources to try to get upstreams to stop requiring SSE2, and we want a consistent experience 2022-05-27 16:06:19 our go compiled apps already requires sse2 if I understand things correctly 2022-05-27 16:06:22 and rust too 2022-05-27 16:10:32 i'm pretty sure rust doesn't 2022-05-27 16:10:44 someone that was testing things on a non-sse cpu said it generated non-sse code 2022-05-27 16:11:38 ncopa: also do you know of anyone that can fix #13702 ? it's now blocking upgrading to new rust 2022-05-27 16:19:30 afaik rust does but not all code is vectorizable 2022-05-27 16:19:42 wait, no, we're talking about floats 2022-05-27 16:21:10 is sse not vectors 2022-05-27 16:22:41 yesn't 2022-05-27 16:23:25 :3 2022-05-27 16:23:45 are you saying if you headpat the vector it floats up into the air 2022-05-27 16:24:35 yesn't 2022-05-27 16:26:28 ah, rust added i586, which is actually i586, whereas i686 is actually pentium4 2022-05-27 16:27:18 how confusing 2022-05-27 16:29:37 x86 simply has too many random extensions all over the place 2022-05-27 16:30:10 (and it's definitely much less bad on other architectures that also have a lot of random extensions all over the place, i swear) 2022-05-27 16:31:34 I hope RISC-V will avoid becoming too messy... But I'm scared 2022-05-27 16:32:19 i would bet riscv will never get anything near desktop class hardware at all 2022-05-27 16:32:31 i think x87 is the only weird incompatible extension, and it's because it predates ieee 754 2022-05-27 16:32:52 i think x86 is the oldest isa still in common use 2022-05-27 16:33:39 psykose: not tomorrow no, but I wonder what will happen in 20 years when more patents have expired 2022-05-27 16:34:01 probably the same thing that happens now with the x86 patents that have expired 2022-05-27 16:34:03 (nothing) 2022-05-27 16:35:02 it's certainly a 'watch this space' though, i'll give it that 2022-05-27 16:38:19 psykose: fair point 2022-05-27 16:38:37 sadly cpus are expensive beasts.. 2022-05-27 16:40:03 either it will be super slow development and you won't see it anywhere even in 20 years or it will become a mess 2022-05-27 16:40:11 does anyone have some hints for debugging a non-booting kernel (which I compiled myself)? I have removed the quiet flag from extlinux.conf, now it prints something about initramfs and then the screen flickers for a second and goes to black. 2022-05-27 16:40:41 it's running in qemu btw 2022-05-27 16:42:32 means it loaded initramfs, then failed for some reason 2022-05-27 16:42:39 what did you change 2022-05-27 16:43:16 but riscv already has extensions, so good luck with getting "not messy" 2022-05-27 16:44:46 eloy_: out of ram? 2022-05-27 16:44:53 check serial output 2022-05-27 17:49:29 psykose: i had a look at the ppc64le binutils issue. its a bit complicate due to the test case is so big and take time to reproduce. 2022-05-27 17:49:47 i was not able to find anything in the upstream git so far, so maybe this should be reported upstream? 2022-05-27 17:50:15 I will be AFK next week so someone else needs to follow up 2022-05-27 17:52:15 btw. feel free to backport the alpine-conf-3.14.2 to 3.16-stable. I figured it only needs to get in before alpine 3.16.1, which we probably should do soonish 2022-05-28 03:25:57 yeah, need to do both; i actually forgot we need to release it too to fix the isos 2022-05-28 03:26:17 and yeah sadly the only way to reproduce anything with binutils is building a random thing.. 2022-05-28 08:02:31 psykose: xpinstall.signatures.required does not seem to work, are we missing some configuration to make it possible? 2022-05-28 08:02:56 you mean setting it to false still does not let you install unsigned addons? 2022-05-28 08:03:01 yeah 2022-05-28 08:03:03 let me see 2022-05-28 08:04:46 could you send me a sample unsigned .xpi (pinky swear no virus, etc) 2022-05-28 08:05:09 https://mirror.drewdevault.com/himitsu-firefox.xpi 2022-05-28 08:08:57 i can't seem to be able to add it with any of those options, but there is a workaround that does let it be loaded 2022-05-28 08:09:10 oh? 2022-05-28 08:09:14 about:debugging has a 'load temporary' 2022-05-28 08:09:18 loads it unsigned 2022-05-28 08:09:20 yeah I'm aware of this 2022-05-28 08:09:22 mhm 2022-05-28 08:09:31 but we should be able to permanently install unsigned addons on a correctly configured firefox dev edition 2022-05-28 08:10:39 think it needs MOZ_REQUIRE_SIGNING= at build time 2022-05-28 08:10:45 i can test in a moment 2022-05-28 08:10:59 ah, yep, it's the one flag i forgot 2022-05-28 08:11:46 i'll give it a test for sure then push it 2022-05-28 08:11:53 ty! 2022-05-28 08:11:53 give it like an hour or so 2022-05-28 08:51:40 psykose: fyi I built it myself with export MOZ_DISABLE_SIGNING=1 in build() and it does not seem to have fixed the issue 2022-05-28 08:51:47 I dunno if you did something different or no 2022-05-28 08:51:58 MOZ_REQUIRE_SIGNING= 2022-05-28 08:52:00 no 1 2022-05-28 08:52:07 ah, rip 2022-05-28 08:52:12 not REQUIRE != disable 2022-05-28 08:52:17 good thing I wasted all of that I/O and CPU time 2022-05-28 08:52:32 s/not// 2022-05-28 08:52:32 ikke meant to say: REQUIRE != disable 2022-05-28 08:52:44 also that, haha 2022-05-28 09:11:55 yep, with that added, and then also signatures.required=false, it lets you: https://img.ayaya.dev/1qFVTPGuGYBI.png 2022-05-28 09:12:17 pushed, thanks 2022-05-28 09:22:45 ah, i think harec is stuck on aarch64 2022-05-28 09:22:57 yep, can reproduce locally 2022-05-28 09:23:13 gets stuck on 34-declarations, and grows forever https://img.ayaya.dev/7OHI0Oa1H794.png 2022-05-28 09:24:50 and now it passed 2022-05-28 09:24:53 huh, strange 2022-05-28 09:25:07 wonder why it took 12 minutes 2022-05-28 09:29:37 psykose: rebuilt with export MOZ_REQUIRE_SIGNING= and it still has the same issue 2022-05-28 09:30:11 well, if you got it working I'll assume your fix is better than mine and await it from the builders 2022-05-28 09:30:30 re: 34-declarations; dunno, I could reproduce it as well but only in abuild. Since it finishes in a *somewhat* reasonable timeframe I deferred the investigation 2022-05-28 09:30:31 that's weird, because for me it definitely fixed it, maybe you didn't upgrade to the locally built one? 2022-05-28 09:30:44 hm, could be 2022-05-28 09:30:45 without a pkgrel bump it wouldn't 2022-05-28 09:30:55 (unless you manually add it from the local filepath) 2022-05-28 09:31:06 aha there we go 2022-05-28 09:31:09 cheers :) 2022-05-28 09:31:11 :) 2022-05-28 09:31:47 now my cool add-on totally works and I'm happy :D 2022-05-28 09:36:50 ^~^ 2022-05-28 17:45:00 similar to that, does chromium-like browsers have such option? I remember that whenever I tried to make extension I could load it but it would scream about being dev and insecure and not from google bla bla 2022-05-29 01:11:45 i think the chromium dev version (+2 versions) has them maybe 2022-05-29 12:29:02 I wonder why rspamd-proxy is a subpackage as it only contains config files 2022-05-29 12:30:36 i dunno either 2022-05-29 12:34:45 its trying to split as there are different daemons, explained in the patch. As it requires extra patching as well: https://git.alpinelinux.org/aports/tree/community/rspamd/30-conf-split-workers.patch But it seems to create more work for little gain 2022-05-29 12:36:35 Yeah 2022-05-29 12:36:54 Installing rspamd-proxy somewhere does not result in a smaller installation or anytthign 2022-05-29 12:37:01 you can rspamd + 2 config files 2022-05-29 12:37:07 s/can/get/ 2022-05-29 12:37:07 ikke meant to say: you get rspamd + 2 config files 2022-05-29 12:37:47 Only advantange might be that it starts only a proxy worker 2022-05-29 12:38:30 I think that is still the rspamd executable tho 2022-05-29 12:38:37 yes 2022-05-29 12:39:05 But it uses packages to deploy specific configurations 2022-05-29 12:41:05 it’s not really worth changing unless the patch needs changing as most of the Dockerfiles I have seen using rspamd include all of the worker sub-packages 2022-05-29 12:42:02 I was not aware of this split, so was wondering why the default config was different from the documentation 2022-05-29 12:42:33 I manually downloaded the original proxy config file 2022-05-29 12:52:12 the proxy config shouldn’t be changed, just split out into worker-proxy.conf which includes worker-proxy.inc 2022-05-29 12:52:39 yes, like I said, I was not aware of this split-up 2022-05-29 13:01:26 This first commit in the git log for APKBUILD is move from testing, the second is cleanup config https://gitlab.alpinelinux.org/alpine/aports/-/commit/7cbc22c7328da6cb1d6bd77052c8efedebc2bc19 which already has a proxy sub package 2022-05-29 13:07:01 38565d1a7242a50c23eeb8bc3f74681d30f42eda 2022-05-29 13:14:08 it looks like before that it was still split into normal controller and fuzzy, so maybe always split? 2022-05-29 13:28:17 a16bitsysop: I just started using rspamd 2022-05-29 13:28:54 I didn't look at all the subpackages that existed 2022-05-29 13:29:33 does something bad happen if the files merely exist? seems pointless to split like 12kb of text 2022-05-29 13:30:00 It would start all the workers specified in those files 2022-05-29 13:30:03 ah 2022-05-29 13:34:11 it depends on your config, it has local.d and override.d as well which are two separate locations for user config files. If you were following a guide for another distro it would be easy to miss the subpackages. worker-fuzzy is just a few configs as well, worker-controller has web interface files as well. 2022-05-29 13:35:08 yes, I found them now 2022-05-29 13:35:24 And adjusting my docker config accordinglu 2022-05-29 13:40:33 a16bitsysop: I looked at upstream documentation btw 2022-05-29 13:42:19 For something like Arch Linux, I would look for documentation on the archlinux wiki 2022-05-29 13:42:26 Anything archlinux specifc would be documented there 2022-05-29 13:42:39 the upstream docs include using stunnel to create an encrypted tunnel for redis 2022-05-29 13:43:16 I haven't come accross that 2022-05-29 13:44:33 https://rspamd.com/doc/tutorials/stunnel_setup.html 2022-05-29 13:45:13 looks fine to me 2022-05-29 14:21:50 a16bitsysop: got it working now 2022-05-29 14:21:53 including contorller 2022-05-29 14:21:55 controller 2022-05-29 14:22:16 at least, I think so :P 2022-05-29 14:23:59 great, the web interface is really handy for stats and checking errors 2022-05-29 14:24:30 How do the controller and the proxy communicate? Through redis? 2022-05-29 14:25:21 through tcp ports 2022-05-29 14:25:52 So if I have them in separate containers, I need to configure where to connect to? 2022-05-29 14:28:27 yes, default is localhost https://github.com/rspamd/rspamd/blob/master/conf/worker-proxy.inc 2022-05-29 14:29:33 I have self_scan = yes; for worker-proxy 2022-05-29 14:30:42 https://rspamd.com/doc/workers/rspamd_proxy.html#self-scan-mode 2022-05-29 14:31:54 you can see my config here: https://gitlab.com/container-email/rspamd/-/tree/main/local.conf 2022-05-29 14:35:46 a16bitsysop: so I only have a controller + proxy in self_scan mode. 2022-05-29 14:35:52 at least, atm 2022-05-29 14:35:58 and redis 2022-05-29 14:42:13 i think self_scan and proxy are two different setups? It looks like I don’t use it and I use proxy 2022-05-29 14:42:36 a16bitsysop: according to the documentation, you can have 2 setups 2022-05-29 14:42:44 proxy + normal workers 2022-05-29 14:42:49 or proxy in self_scan mode 2022-05-29 14:42:58 The latter is a lighter setup 2022-05-29 14:43:40 piping all my saved spam messages to rspamd now :) 2022-05-29 14:44:10 its lighter but if it crashes thats it, proxy workers spawn a new one. Not that it crashes much now 2022-05-29 14:45:04 It's for a small personal mail server, so I can live with that :) 2022-05-29 14:45:16 and if it plain crashes, docker will restart it 2022-05-29 14:45:58 okay 2022-05-29 15:00:47 ikke: what would make it really great is using redis for postfix as well! https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/33818 2022-05-29 15:02:02 a16bitsysop: natanael is the maintainer, but my feeling is that he would not be in favor of such a patch 2022-05-29 15:02:40 This adds major functionality as a patch, something we try to avoid 2022-05-29 15:05:11 Is it based on this? https://github.com/titusjose/postfix_redis 2022-05-29 15:06:30 it looks much more updated than that 2022-05-29 15:07:18 right 2022-05-29 15:07:35 https://marc.info/?l=postfix-devel&m=161441647826361&w=2 2022-05-29 15:07:51 Apparently Duncan updated it 2022-05-29 15:15:38 hah 2022-05-29 15:16:09 cheeky :3 2022-05-29 15:47:04 I have an open MR in the titusjose repo to update it but it remains unmerged 2022-05-30 07:43:30 omni: Hello 2022-05-30 07:48:35 Oops, never mind, misread something, sorry for the highlight 2022-05-30 07:52:42 rbq: hi! =) 2022-05-30 09:06:13 omni: You can close your imapsync MR (28902), it's been update :) 2022-05-30 09:20:42 rbq: nice, thanks! 2022-05-30 09:29:43 is that rubicon :) 2022-05-30 09:29:51 Hello, psykose 2022-05-30 09:30:09 heya 2022-05-30 09:31:08 Any advice for !34890 ? :) 2022-05-30 09:31:17 (I know, I should squash my commits) 2022-05-30 09:31:29 > Changes: 199 2022-05-30 09:33:05 you can set pipeline exec time in your repo settings 2022-05-30 09:35:17 btw. could someone look at !33367 2022-05-30 09:35:26 that's a lotta rebuilds 2022-05-30 09:35:45 currently i am bashing my head against libexecinfo being completely worthless 2022-05-30 09:39:36 insert "first time?" gif 2022-05-30 09:40:54 if you pass more than 52 as size it segfaults, which is amazing 2022-05-30 09:41:18 and it segfaults at anything over 17 or so in gdb, so i'm not sure exactly why 2022-05-30 09:41:32 i hate this shit with a passion but some programs actually call backtrace for some reason, and it breaks them 2022-05-30 09:44:04 Thanks for the information, panekj, I'll have a look at that 2022-05-30 09:44:49 quite literally it just fails in some shit like `return __builtin_frame_address(56);` i have no idea how you're meant to make things like this work 2022-05-30 09:45:16 the asm is especially funny `mov (%rax),%rax` 2022-05-30 09:47:50 I'm rebuilding 20 aports that depend on libperl.so directly, and another 177 that place .so files in /usr/lib/perl5/vendor_perl/auto (according to https://pkgs.alpinelinux.org/contents)...I think for Perl 5.32.0, over 700 aports were rebuilt, I won't be doing that :) 2022-05-30 09:49:44 not sure how perl works, or if the rebuilds are needed 2022-05-30 09:59:37 oh, and it pretty much always segfaults with -O1/s/2. *sometimes* it doesn't with -Os, and then it returns 0 addresses, the only way to get it to do anything whatsoever is with -O0 2022-05-30 09:59:40 hate this 2022-05-30 10:03:07 if i just copypaste the glibc one and pinky swear it's actually bsd licenced can i just put it into aports /s /s /s 2022-05-30 10:09:58 this program has a custom backtrace printing thing that then also optionally links into libbacktrace and there is a _backtrace_get with a fat `char message[12000];` at the top of the function 2022-05-30 10:09:59 i love software 2022-05-30 10:19:28 ACTION hugs psykose 2022-05-30 10:20:01 i successfully excised it from the program and it doesn't segfault anymore when trying to get backtraces i guess 2022-05-30 10:20:07 though it's quite funny now 2022-05-30 10:20:24 'your program ran into an issue! plz report this to the developers: ' 2022-05-30 10:24:07 i wonder how well works in comparison 2022-05-30 10:24:55 and i also wonder how much of a shitshow c++23 will be with std::stacktrace_entry 2022-05-30 10:28:00 ACTION hugs panekj 2022-05-30 11:22:43 psykose: on the one hand, stacktrace works better than libexecinfo, on the other hand... it's C++ so good luck with the traces :P 2022-05-30 11:23:07 hehe, just a curiosity 2022-05-30 11:26:46 psykose: the glibc one is dependent on glibc rtld internals afaik 2022-05-30 11:26:54 very unfortunate 2022-05-30 11:27:25 in practice, it is possible to use dlinfo(3) and stack walking to write a backtrace function that will work for dynamically linked programs 2022-05-30 11:27:45 (tbh all of this backtrace stuff is undefined behavior on a static binary) 2022-05-30 11:28:11 in this case nothing is static 2022-05-30 11:28:21 i just wish calling backtrace() didn't segfault and gave you nothing consistently 2022-05-30 11:28:39 er, 'at least' gave you nothing instead of crashing 2022-05-30 11:29:26 you need a little bit of target-specific asm to do the unwind, but i could write something akin to libucontext for this 2022-05-30 11:30:28 if you can write a replacement for whatever the hell main/libexecinfo is that actually works i would be very happy (though not because i use it anywhere or anything, as i don't) 2022-05-30 11:30:55 just annoying that random programs expect the functionality and then one has to wonder which way to patch all of it over and over again.. 2022-05-30 11:32:15 and in similar news fortify-headers has also needed fixing for a while, though it's hard to make minimum reproduction of the inlining failure for vsnprintf 2022-05-30 11:32:36 because it relies on way-too-much stuff happening in c++ world to fail in lto 2022-05-30 11:34:59 and is also seemingly fixed with a -fno-ipa-cp in most cases, but that is not great 2022-05-30 13:03:21 Hi, I'd like to work on proposing a new community package for alpine. 2022-05-30 13:03:21 I've browsed the website and wiki but did not found any guidelines 2022-05-30 13:03:21 Do you have any documentation pointers ? 2022-05-30 13:03:21 I'd lile to get familiar with the process of creating and testing such package 2022-05-30 13:04:32 oh, it seems I found sth : https://wiki.alpinelinux.org/wiki/Category_talk:Developer_Documentation#Building_from_source_and_creating_packages isn't it ? 2022-05-30 13:06:31 ty for the vulkan bump psykose 2022-05-30 13:20:55 setop: it's usually: 1. make package in testing 2. move from testing if deemed fine 2022-05-30 13:22:52 most if not all new package maintainers just look at other packages as examples since those can be in more up-to-date format 2022-05-30 13:24:08 https://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package is fairly up-to-date 2022-05-30 13:24:28 @panekj, from a complete newbee it really looks like "draw a owl in two steps" :) (that one : https://external-preview.redd.it/DodWFQ9mQkVyWoKFa0ZIu12PYrPo3P2T0taaK-lgJCo.png?width=530&height=277.486910995&auto=webp&s=c48c5b38cad1e966b705e080307aa58ade47bc83) 2022-05-30 13:24:30 and if something is wrong, one of the devs who will review your MR will help 2022-05-30 13:24:50 But I'll go through the doc and come back if sth not clear 2022-05-30 13:24:54 thank for your message. 2022-05-30 13:25:26 page I linked is 90% of what you really need to know 2022-05-30 13:25:35 :+1 2022-05-30 13:26:09 then there are: https://gitlab.alpinelinux.org/alpine/tsc/-/issues/2 https://gitlab.alpinelinux.org/alpine/tsc/-/issues/22 2022-05-30 13:26:24 but those are ongoing things so they are not really documented (yet) 2022-05-30 13:27:31 (https://gitlab.alpinelinux.org/alpine/tsc/-/merge_requests/5/diffs) 2022-05-30 14:26:20 bl4ckb0ne: :) seemed like everything worked, so i threw it in 2022-05-30 14:26:24 hopefully i don't regret it 2022-05-30 14:27:57 heh its on you now 2022-05-30 14:30:38 ~ 2022-05-30 14:35:36 Oops, someone asked a question in my Perl 5.36.0 MR that I don't really know the answer to, and haven't thought about yet 2022-05-30 14:36:21 As Perl is in main/, this mean if it is upgraded, it will have to be upgraded in Alpine 3.16 too, right? 2022-05-30 14:36:41 Only if it's security/bugfix 2022-05-30 14:37:00 Ok 2022-05-30 17:41:59 ncopa: could you please take a look at this at some point soon-ish (no big rush, just has been in the backlog for a while): https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/31213 2022-05-30 18:02:26 the cp pkgdir/etc just makes an empty directory, is that what you intended 2022-05-30 18:06:14 ah no, that's not intentional. I'll fix it 2022-05-30 18:11:52 aside from that looks fine :) 2022-05-30 18:13:41 :) 2022-05-30 20:51:31 Hi! I'm reviewing https://security.alpinelinux.org/branch/3.16-main, trying to find status for CVE-2021-35940. I'm looking at https://pkgs.alpinelinux.org/package/v3.16/main/x86_64/apr, and I'm missing the patch here. Looks like a vanilla 1.7.0, missing patch https://dist.apache.org/repos/dist/release/apr/patches/apr-1.7.0-CVE-2021-35940.patch. 2022-05-30 20:52:36 I'm wondering, why is the package not listed as a vulnerable package, and how to help to get this patch included in alpine. Thanks! 2022-05-30 20:52:38 > a16bitsysop: natanael is the maintainer, but my feeling is that he would not be in favor of such a patch 2022-05-30 20:52:38 It doesn't change existing behaviour, only if you choose to use redis 2022-05-30 20:53:18 it adds 'large' functionality 2022-05-30 20:58:02 Is Natanael in this channel? 2022-05-30 20:58:34 I'm not sure about the reference to redis, the problem seems not related to redis? 2022-05-30 20:59:08 that's not related to what you said :) 2022-05-30 20:59:40 oic 2022-05-30 21:03:00 it has been pretty well tested, and someone is using in a VM as well and compiling themselves. Still not in postfix github repo 2022-05-30 21:05:13 jalberti: fixed in every branch 2022-05-30 21:09:50 psykose: I'm clearly slow today ... please elaborate, yes, the patch is already in the 1.7 branch of apr, but no 1.7.1 release has been made, the 1.7.0 src downloaded during the apk build does not include the patch, does it? 2022-05-30 21:10:15 jalberti: what i mean is that i pushed that patch you linked to every branch, and it is building/has finished building 2022-05-30 21:10:31 e.g. https://git.alpinelinux.org/aports/commit/?id=905ae740f87a 2022-05-30 21:11:19 branch as in the alpine release branches 2022-05-30 21:11:27 psykose: I did not dare to believe that this was what you meant ;) 2022-05-30 21:11:33 thx 2022-05-30 21:15:34 psykose: how long does it take roughly for the apk to show up in 3.16 via apk -U upgrade? 2022-05-30 21:15:46 depends on the mirror 2022-05-30 21:15:57 for dl-cdn.. roughly 15 minutes at most after the builder is done 2022-05-30 21:16:02 usually 5-10 2022-05-30 21:16:09 for the others, shrug 2022-05-30 21:16:23 some sync every 30 minutes, some every 6 hours 2022-05-30 21:16:54 edge and 3.16 specifically will take a while, because it's building ceph in the meantime :) 2022-05-30 21:18:21 ok, but "a while" is still measured in hours, not days, can I assume that 2022-05-30 21:18:46 did your company cto tell you to get this fixed doublefast :p 2022-05-30 21:18:52 but yes, about ~1 hour for x86_64 2022-05-30 21:20:01 lol ... nope, just curious, trying to learn something new 2022-05-30 21:20:14 or, rather, x86_64 is done actually, i'm just bad at looking 2022-05-30 21:20:16 too many things on screen 2022-05-30 21:21:39 thanks a ton, nice to meet you ... and bye for now. :) 2022-05-30 21:21:47 you can always look at the builders yourself too, the usual page is https://build.alpinelinux.org/ 2022-05-30 21:21:53 have a great day :) 2022-05-30 21:22:08 u2, thx! 2022-05-31 07:10:41 Running the Perl testsuite in parallel shortened the time needed to finish building it from 20 minutes to 4 :) 2022-05-31 07:12:21 threads are fun things 2022-05-31 07:17:26 until you have to debug something 2022-05-31 07:17:37 Perl :/ 2022-05-31 07:20:39 unless 2022-05-31 07:22:13 therefore 2022-05-31 07:24:14 perchance 2022-05-31 07:32:27 omg, wouldn't've been surprised if there was a perchance or therefore statement in perl 2022-05-31 07:33:15 make sure you're doing it for openssl too 2022-05-31 07:33:20 perchance is definitely in british PHP 2022-05-31 07:33:21 you have to use, like perl, some random variable for it 2022-05-31 07:36:35 i actually ran into giga-slow openssl tests before 2022-05-31 07:36:38 what is the variable 2022-05-31 07:38:31 oh crap it's only for openssl 3 2022-05-31 07:38:33 but it's HARNESS_JOBS 2022-05-31 07:43:59 aw 2022-05-31 08:55:05 ikke: friendly ping on the appstream infra MRs :) It'd be nice to have 3.16 available and the fixes from the latest release in place. Not super urgent though if it is not possible 2022-05-31 13:05:35 damn psykose stop stealing my thunder :P 2022-05-31 13:05:41 :) 2022-05-31 13:05:56 you can press merge on !30596 if you want 2022-05-31 13:06:57 mergered 2022-05-31 13:07:29 ty 2022-05-31 13:08:18 btw for !34928 it's a known issue for pre RDNA amd hw 2022-05-31 13:08:44 ah, makes sense 2022-05-31 13:08:51 i think it should work on my laptop with iris though 2022-05-31 13:09:00 i'll try that when i have more than negative energy 2022-05-31 13:09:04 doenst work on my RX580 here but works fine on my x230 2022-05-31 13:09:18 some zen apu? 2022-05-31 13:09:30 x230 is haswell or ivy bridge 2022-05-31 13:09:34 ah 2022-05-31 13:09:40 right, good to know 2022-05-31 13:09:56 should work fine on intel 2022-05-31 13:10:25 https://gitlab.freedesktop.org/mesa/mesa/-/issues/5882 there you go 2022-05-31 13:10:28 GFX8 and older 2022-05-31 13:10:47 ahh 2022-05-31 13:10:54 https://www.x.org/wiki/RadeonFeature/ 2022-05-31 13:11:11 GFX9 is polaris (RX4xx and RX5xx) 2022-05-31 13:11:15 8* 2022-05-31 13:11:56 iirc its not blocked by HW caps but by a shitton of refactoring 2022-05-31 13:12:25 one day, one day.. 2022-05-31 13:12:35 not sure if there are even any benefits to using the vulkan wlr backend 2022-05-31 13:12:54 fwiw you're probably fine with the gles2 one but it's a nice to have 2022-05-31 13:13:08 i was mostly a bit curious because my gpu is a little fucked and there's some weird artifacting in specific cases, but i can't even test it 2022-05-31 13:13:15 its good to run it and give feedback since its pretty new 2022-05-31 13:13:17 and for intel.. well, i have never had an intel igpu fail at anything 2022-05-31 13:14:29 the intel driver is quite good 2022-05-31 13:15:17 mhm 2022-05-31 15:03:09 fucking bazel 2022-05-31 15:08:04 when i see bazel i just give up and turn around 2022-05-31 15:09:27 my fan just turned on 2022-05-31 15:25:37 psykose: it's fine, most people do so inside google too 2022-05-31 15:25:56 don't blame em 2022-05-31 15:38:56 what are you trying to build? 2022-05-31 15:39:57 i'm pretty sure i found a small issue with this -stage0 providers stuff in abuild 2022-05-31 15:39:59 joy 2022-05-31 15:40:28 tensorflow 2022-05-31 15:40:30 I know, I know 2022-05-31 15:40:39 hah 2022-05-31 15:40:43 i tried to do that just the other week 2022-05-31 15:40:59 then i went 'wait how do i patch these 50 million vendored things' and gave up 2022-05-31 15:41:13 psykose: what did you find? 2022-05-31 15:41:21 pypy depends on pypy-stage0 2022-05-31 15:41:26 which makes it uninstallable 2022-05-31 15:41:28 try yourself 2022-05-31 15:41:33 the reason why is the tracing is broken 2022-05-31 15:41:40 i'm pretty sure the error is in uhh 2022-05-31 15:42:33 about :1430 2022-05-31 15:42:50 the issue is it adds the names of the packages that 'own' some .so files as a depends= 2022-05-31 15:43:06 and for whatever reason, one of those files inside the pkgdir, is returned as owned by -stage0, so.. 2022-05-31 15:43:11 https://tpaste.us/QNzb 2022-05-31 15:43:26 i should probably build this with some debugging in abuild 2022-05-31 15:43:52 Seems like they both provide cmd:pypy 2022-05-31 15:43:57 which conflicts 2022-05-31 15:44:05 well yeah, they're the same package 2022-05-31 15:44:09 it's a stage0 bootstrap 2022-05-31 15:44:33 the reason that happens is a bug in abuild, not any other reason 2022-05-31 15:45:58 pypy should not depend on pypy-stage0 indeed 2022-05-31 15:46:15 This one? I don't see stage0 here: https://gitlab.alpinelinux.org/alpine/aports/-/blob/master/testing/pypy/APKBUILD 2022-05-31 15:46:37 It's a dedicated package: https://gitlab.alpinelinux.org/alpine/aports/-/blob/master/testing/pypy-stage0/APKBUILD 2022-05-31 15:46:44 I almost wish google *didn't* publish open source software, so that we wouldn't have to deal with it 2022-05-31 15:48:14 the only shimmy fix is overriding trace_apk_deps() and manually adding in the needed libs, then it becomes correct, but of course that is quite scuffed 2022-05-31 15:49:01 i don't think apk is doing anything wrong if my theory is right, so abuild just needs to have a special options= for this case or something 2022-05-31 15:49:25 So pypy has to be built in two passes? 2022-05-31 15:49:44 the slightly-dirty-but-actually-small fix is `if option stage0; delete stage0 from deps;`, and that would be fine 2022-05-31 15:49:47 Ermine: yes, as it depends on python2 2022-05-31 15:50:26 psykose: do you know what part of that function adds -stage0 as a direct dep? 2022-05-31 15:50:40 the only place it can be added is the apk output 2022-05-31 15:50:54 deppkgs= is apk --who-owns 2022-05-31 15:50:55 Ermine: so liske provided the stage0 package, which first builds python2, then builds pypy 2022-05-31 15:51:06 then autodeps="$autodeps $deppkgs" 2022-05-31 15:51:16 i mean, i could be completely wrong 2022-05-31 15:51:29 Hi, is there process to get review/commit for https://gitlab.alpinelinux.org/alpine/infra/aports-turbo/-/merge_requests/47 2022-05-31 15:51:30 but i looked at every part of abuild that modifies the depends; and the other parts don't apply 2022-05-31 15:51:50 the so: ones would be named different, the symlink ones are something else, etc 2022-05-31 15:51:55 it's the only part of the code that can add that 2022-05-31 15:52:11 so once i have 5 hours to build pypy i guess i will add 20 printfs in there and confirm it 2022-05-31 15:54:26 Hmm, that is based on rpaths? 2022-05-31 15:54:32 https://gitlab.alpinelinux.org/alpine/abuild/-/blob/master/abuild.in#L1184 2022-05-31 15:58:29 the rpaths don't matter, it just gets basically every dir that has a .so in it, then checks the .so 2022-05-31 15:58:40 and there are no rpaths, theyre' just $ORIGIN 2022-05-31 16:00:18 does pypy provide=pypy-stage0? 2022-05-31 16:00:29 no 2022-05-31 16:00:32 it should allow the solver to cancel out some edges when solving 2022-05-31 16:00:41 and is what we do for rust vs rust-bootstrap 2022-05-31 16:00:44 No, only pypy-bootstrap 2022-05-31 16:00:53 ok 2022-05-31 16:00:57 it's the same as every other -stage0 implementation, they both provide -bootstrap with one being higher 2022-05-31 16:01:14 in this case it explicitly adds -stage0 as a dep, not anything else to do with virtual stuff 2022-05-31 16:01:26 let me take a look 2022-05-31 16:01:32 the solver would always fail if you add -stage0 into pypi (to install), so i don't think that's the problem 2022-05-31 16:01:38 https://tpaste.us/Ke0q 2022-05-31 16:01:39 unless it can do stuff regarding --who-owns 2022-05-31 16:02:33 the problem is that the non-virtual name pypy-stage0 is preferenced over pypy-bootstrap when evaluating the solution 2022-05-31 16:02:45 if you rename pypy-stage0 to pypy-bootstrap, it should fix it 2022-05-31 16:03:27 is that a general improvement for everything that currently uses -stage0 too? it seems like it would apply to every case 2022-05-31 16:04:03 yes, the solver is not even getting to the point where it considers pypy-bootstrap at all in these cases 2022-05-31 16:04:08 Ariadne: but is it correct that pypy-stage0 is added as a dependency to pypy? 2022-05-31 16:04:17 it's not, that's separate 2022-05-31 16:04:20 and purely an abuild bug 2022-05-31 16:04:26 but if we can sidestep it like this, that works 2022-05-31 16:04:48 ikke: abuild is doing that because the provider initially chosen for pypy-bootstrap is pypy-stage0 2022-05-31 16:05:20 because then in any case where abuild adds -stage0 by accident, it will instead add -bootstrap, which will already be virt-provided by the same package, so it will never conflict like this 2022-05-31 16:05:23 sure, i will rename it 2022-05-31 16:05:29 apk --who-owns is entirely unaware of virtuals, it just tells you what the non-virtual owner of the file is 2022-05-31 16:05:39 mhm 2022-05-31 16:06:01 surprised this is the first instance of this bug, hah 2022-05-31 16:07:50 you don't really need provider_priority here either 2022-05-31 16:07:56 just make pypy-bootstrap provide=pypy 2022-05-31 16:08:08 well, provide=pypy-${pkgver}-r${pkgrel} 2022-05-31 16:08:44 because a real package will always win over a virtual 2022-05-31 16:08:54 then you can do makedepends=pypy in pypy 2022-05-31 16:09:00 that's how we do rust anyway 2022-05-31 16:09:15 Ariadne: isn;t that disallowed? 2022-05-31 16:09:25 which is why -bootstrap virtuals were used 2022-05-31 16:09:59 no, the -bootstrap virtuals are provided to satisfy the real package until it is available 2022-05-31 16:10:24 afaik there was also an issue that a package could not makedepends on itself 2022-05-31 16:11:08 i mean, you can explicitly depend on the virtual if you want 2022-05-31 16:11:11 but apk can solve it 2022-05-31 16:12:01 https://gitlab.alpinelinux.org/alpine/aports/-/blob/master/community/ghc/APKBUILD#L35 2022-05-31 16:12:05 ah, right 2022-05-31 16:17:31 well the answer there is "abuild is doing something it shouldn't" :p 2022-05-31 16:18:55 tbf i prefer the two-aport version anyway 2022-05-31 16:19:06 Ariadne: also thanks for the fast debugging :) hope it works 2022-05-31 16:19:28 psykose: you mean rather than 2 aport + provides? 2022-05-31 16:19:55 ghc has no -bootstrap/-stage0 or whatnot, it's just one aport 2022-05-31 16:20:16 Yes, but it requires manual intervention on bootstrap 2022-05-31 16:20:40 ah, i guess that's true 2022-05-31 16:20:42 Meaning someone needs to explicitly install it from edge on bootstrap 2022-05-31 16:21:00 the stage0 part was meant to fix that, and make the bootstrap origin explicit 2022-05-31 16:21:06 the two-aport version does not work for ghc or rust, there would have to be dozens of aports 2022-05-31 16:21:27 yeah 2022-05-31 16:21:33 or, well 2022-05-31 16:21:45 technically it can be +1. with.. 80 source urls, and a nice loop :) 2022-05-31 16:21:51 (and the price of your sanity if it crashes) 2022-05-31 16:22:36 and that's why people hate bazel 🙃 2022-05-31 16:23:00 and for pypy it means you don't have to build python2 on each upgrade 2022-05-31 16:23:25 because you get ✨ magic ✨ from having 80 intermediary artifacts that all have to work out :p 2022-05-31 16:24:31 hey you made it sparkly 💫 magic ✨ 2022-05-31 16:24:34 i like sparkly magic 2022-05-31 16:24:39 i love bazel! 2022-05-31 16:25:09 It's an ilusion, don't fall for it 2022-05-31 16:34:51 a16bitsysop: The way rspamd works, it makes less and less sense to me to separate out the config in subpackages. I've asked some questions in #rspamd, and they mentioned for example you'd run a controller wherever you have a worker. 2022-05-31 16:37:03 I first split up the controller from the worker (proxy), but that resulted in the controller not getting any starts from the worker 2022-05-31 16:37:12 stats* 2022-05-31 16:42:39 (also if you think bazel is bad, wait until you hear about GRTE 😵‍💫) 2022-05-31 16:47:29 is it worse than android tree building 2022-05-31 16:47:48 it's the Google Runtime Environment 2022-05-31 16:47:53 Ariadne: is that supposed to be pronounced "grate"? lol 2022-05-31 16:47:54 so much stuff packed in there 2022-05-31 16:48:31 it's basically a linux distribution without actually being a linux distribution 2022-05-31 16:48:59 hey that describes android too /s 2022-05-31 19:26:03 > I first split up the controller from the worker (proxy), but that resulted in the controller not getting any starts from the worker 2022-05-31 19:26:03 I am not sure how the stats are passed, they are stored locally in `/var/lib/rspamd/stats.ucl ` 2022-05-31 19:26:54 Yes, and I mounted that directory in both containers 2022-05-31 19:27:18 but the proxy said it didn't find a controller in the config so it would do some controller tasks itself 2022-05-31 19:28:29 Once I installed both rspamd-controller and rspamd-proxy in the same container, everything started to work 2022-05-31 19:29:33 have you set the controller config? you can do `rspamadm configdump worker` to see what the final settings used are 2022-05-31 19:30:08 a16bitsysop: what controller config exactly? 2022-05-31 19:31:02 that gives controller fuzzy normal and proxy and basicaly lists the settings 2022-05-31 19:32:50 https://tpaste.us/jPpq 2022-05-31 19:33:07 The controller part is there because I installed (and additionally configured) rspamd-controller 2022-05-31 19:34:11 secure_ip 0.0.0.0 is super bad and allows any ip to connect without a password to the controller 2022-05-31 19:37:18 Not without a /0 suffix 2022-05-31 19:38:45 why not set to an actual ip? my rspamd_proxy hosts has a port eg 127.0.0.1:11333 2022-05-31 19:40:02 It's routed internally in docker 2022-05-31 19:43:42 I guess the you would have a different config for the proxy and controller if running in separate containers? 2022-05-31 19:44:38 from what I understood, that setup does not work 2022-05-31 19:45:07 "If a node that runs a normal worker does not run a controller, it means you cant incldue its data/history/statistics in the webui" 2022-05-31 19:46:11 So I think at least each worker needs a controller 2022-05-31 19:46:50 that doesn't seem to save any resources! 2022-05-31 22:22:42 I think I liked it more when the main page linked to git.a.o over gitlab.a.o, but perhaps I'm just conservative like that sometimes 2022-05-31 23:06:23 omni: issues/merge requests are on gitlab, so it makes more sense for beginners; pkgs.a.o still links to cgit though :)