2019-01-01 02:09:03 ncopa: questions on apk (I'll start writing related sections tomorrow, so no need to hurry re: answering): 2019-01-01 02:09:49 1. what exactly is in the APKINDEX (and similar things) right now? (this won't go directly into the user handbook, but is rather here so I understand apk better, which then helps communicating the general usage stuff) 2019-01-01 02:10:20 2. repositories support http and https - do they support anything else (e.g rsync?), and are there external helpers involved at any point? 2019-01-01 02:10:58 3. I noticed searching for, say, a binary, or a library name will often yield the desired result - I suspect this is because search can parse APKINDEX - what are the specifics on that (what can search access, and what parts of it are in APKINDEX) 2019-01-01 02:11:42 3. does any of this propagate to add? (I recall there being a case where `add x` added package y that contained x, but a quick attempt to do this with a shared object quickly failed, meaning it does not match up exactly to what search can do) 2019-01-01 02:11:47 s/3/4/ 2019-01-01 02:11:47 SpaceToast meant to say: 4. does any of this propagate to add? (I recall there being a case where `add x` added package y that contained x, but a quick attempt to do this with a shared object quickly failed, meaning it does not match up exactly to what search can do) 2019-01-01 02:12:22 5. what is the "official" way to downgrade from a higher release? (I've done it edge -> 3.8 before, but it seemed quite hacky, if there is no official way, just say so) 2019-01-01 02:12:46 6. what information (`apk info`) is available before installing a package, vs after? 2019-01-01 02:13:15 7. what's the exact syntax of the world file? will it work with things that are not package names but `add` will support? how about virtuals? 2019-01-01 02:13:54 8. currently, signing keys are all RSA 2048 - is it possible to have different keytypes, and if so, what algorithms are we currently restricted to? what's /etc/apk/keys/ parsing like? 2019-01-01 02:14:06 (and that is all, I'ma start settling in for sleep now, thanks in advance :) ) 2019-01-01 09:16:19 https://wiki.adelielinux.org/wiki/APK_internals#.PKGINFO 2019-01-01 19:50:47 SpaceToast: 1: its an compressed tarball with lists like this: http://0x0.st/sRzb.txt 2019-01-01 19:51:32 you can see that *.so library names are in there, which might explain 3 2019-01-01 19:52:17 6: before, only the data that is in the APKINDEX, after installing, the actual package has been fetched so its now possible to know the file list, for example 2019-01-01 19:55:50 dragons be at 7 2019-01-01 20:00:47 in general, the 'binary' things are usually gzipped tarballs, so its possible to inspect them with standart tools 2019-01-01 22:58:32 thank you danieli, AinNero - I'm still missing 2, 4, 5 and 8, but that should be enough to write at least an initial "user"-oriented piece 2019-01-01 23:02:40 external helpers, i believe not, apk seems to have support for ftp and http(s) through its libfetch 2019-01-01 23:05:08 i don't believe there is any supported or official way to downgrade, someone will have to confirm that 2019-01-02 04:07:45 ncopa: are release versions strictly following semver (or at least "strictly enough" that I can say we follow it), or should it be explained separately? (and, if so, how do we handle release versions?) 2019-01-02 08:50:30 SpaceToast: I think we can say that we follow semver 2019-01-02 08:50:49 hmm 2019-01-02 08:51:19 i think our versions are MAJ.MIN.PATCH 2019-01-02 08:51:41 when PATCH increases, everything should work without any changes in configs 2019-01-02 08:52:01 when MIN increases, you may need change configs etc 2019-01-02 08:52:26 but it should be binary compatible, so it is possible to upgrade 2019-01-02 08:53:10 when MAJ increases, it is no longer binary compatible, and re-install is needed for upgrade 2019-01-02 08:53:32 for example, when we swtiched from uclibc to musl libc, we changed MAJ=2 to MAJ=3 2019-01-02 08:54:10 im dont think that is how semver works 2019-01-02 08:54:58 re 2. only http/https is supported. as danieli said, ftp may work 2019-01-02 08:55:37 i think we at some point used wget as a fallback external helper 2019-01-02 08:55:41 but i dont think we do that anymore 2019-01-02 08:56:09 nope. we dont use external helpers 2019-01-02 08:57:48 re 4 and apk add 2019-01-02 08:58:00 some packages has "provides" 2019-01-02 08:58:46 so you can do things like: apk add cmd:htop 2019-01-02 08:59:06 because: 2019-01-02 08:59:09 $ apk info --provides htop 2019-01-02 08:59:09 htop-2.2.0-r0 provides: 2019-01-02 08:59:09 cmd:htop 2019-01-02 08:59:37 apk add can also install by file: apk add ./file.apk 2019-01-02 08:59:42 but im not sure we want support that... 2019-01-02 09:00:22 there is also: apk add --virtual mytempname dep1 dep2 dep3 2019-01-02 09:00:47 where mytempname is a "virtual" or empty package which depends on dep1 dep2 and dep3 2019-01-02 09:01:07 the purpose is to lock dep[1-3] so they are not accidential removed 2019-01-02 09:01:19 for example as build time dependencies 2019-01-02 09:01:55 and it also means that you can uninstall dep1, dep2 and dep3 in one go with: apk del mytempname 2019-01-02 09:03:44 `apk search` and `apk add` does not take same args. apk search can do globs: `apk search "qemu*"` 2019-01-02 09:03:49 apk add cannot do that 2019-01-02 09:04:14 there are no official way to downgrade 2019-01-02 09:06:00 re 8. I don't know. It is possible that it other key types works, but i have not tested it 2019-01-02 11:04:09 ncopa: thanks! very helpful :) (esp. re: 4) 2019-01-03 09:56:07 SpaceToast, we have smart support :) 2019-01-03 14:53:37 "smart support?? 2019-01-03 14:53:38 "? * 2019-01-03 14:54:38 danieli, https://git-scm.com/book/en/v2/Git-Internals-Transfer-Protocols 2019-01-03 14:54:43 ah right 2019-01-03 14:54:53 i forgot about that for a sec 2019-01-03 15:18:01 clandmeter: nice! So now we can generate beta.docs.a.o the "proper" way (no more workspace/ stuff, just run against site.yml) 2019-01-03 15:18:22 (also, antora 2.0.0 is fully out now, for a few days, so README should be complete) 2019-01-03 15:22:07 SpaceToast, yes we can 2019-01-03 15:22:15 what is the proper way from beta to stable? 2019-01-03 15:22:43 I'm not entirely sure on the specifics, I simply uninstalled and reinstalled using yarn 2019-01-03 15:22:50 though I suspect just performing an update would be fine 2019-01-03 15:25:22 i have no idea about node and try to keep it that way :) 2019-01-03 15:25:56 my primary gripe with node is... node itself - the implementation 2019-01-03 15:27:14 practically, variants of ecmascript have been used for much longer for scripting than node's existed, and html is very flexible as a UI toolkit (e.g the Adobe AIR era really shows this off), it's just that v8 sucks and the people behind node don't seem to be interested in fixing it :/ 2019-01-03 15:30:09 antora --version 2019-01-03 15:30:09 2.0.0 2019-01-03 15:30:23 \o/ 2019-01-03 15:30:39 is there a lightweight adoc to html converter? 2019-01-03 15:31:20 there's asciidoctor (and its family), and the original asciidoc implementation in python (which doesn't support a lot of the extensions that have grown since) 2019-01-03 15:31:29 I suppose it depends on your definition of lightweight 2019-01-03 15:31:37 antora does a bit more than just convert adoc to html though 2019-01-03 15:31:58 during development, I usually use the asciidoctor.js-based browser extension with auto-reload to see just the page I'm working on with the default CSS 2019-01-03 15:32:39 cgit can also do adoc if we really wanted. 2019-01-03 15:32:52 aha 2019-01-03 15:33:00 if there was some lightweight option i would take it into consideration 2019-01-03 15:33:04 asciidoctor would be the go-to implementation 2019-01-03 15:33:15 but thats ruby right? 2019-01-03 15:33:17 yeah 2019-01-03 15:33:25 ok ill keep md only :p 2019-01-03 15:33:28 it runs regularly (e.g for my blog) on my cheapo vps without any visible impact 2019-01-03 15:33:33 that's fine :) 2019-01-03 15:33:40 I already converted the READMEs into md anyway 2019-01-03 16:04:31 SpaceToast, docs should now update on push to any branch for user-handbook 2019-01-03 16:05:36 I guess we also need to upgrade docs.a.o on push? 2019-01-03 16:05:59 technically, yeah 2019-01-03 16:06:17 the idea is on any docs/* push, docs.a.o should be rebuilt (but it shouldn't take too long) 2019-01-03 16:06:37 (and only docs.a.o need be explicitly checked out, now that we have smart http) 2019-01-03 16:06:53 yep, it works 2019-01-03 16:07:01 i had to kill the cache though. 2019-01-03 16:07:08 it was complaining 2019-01-03 16:07:31 btw, cgit is no long the official prefix on git.a.o 2019-01-03 16:07:39 but still supported 2019-01-03 16:07:46 ah, what should I change it to? 2019-01-03 16:07:52 (I'll push for a rebuild while I'm at it) 2019-01-03 16:07:54 / 2019-01-03 16:08:07 i updated the uris on cgit 2019-01-03 16:09:13 ok, should be updated now - I can also build it locally 2019-01-03 16:09:33 it did? 2019-01-03 16:10:56 ? 2019-01-03 16:11:04 ah you updasted docs 2019-01-03 16:11:17 i thought you updated handbook 2019-01-03 16:11:29 to verify if it would update 2019-01-03 16:11:51 ill add a check for git/# later 2019-01-03 16:11:54 oh, sorry, still waking up, I assumed we already had full docs/* 2019-01-03 16:11:59 err 2019-01-03 16:12:08 git/docs/# 2019-01-03 16:12:36 and if the topic is docs.a.o ill pull the repo 2019-01-03 16:12:48 im going to go home. 2019-01-03 16:12:50 \o 2019-01-03 18:45:26 your nick lengths is the same! 2019-01-03 21:04:07 :) 2019-01-03 21:04:18 SpaceToast, fix auto update 2019-01-03 21:04:27 s/fix/fixed 2019-01-03 21:04:27 clandmeter meant to say: SpaceToast, fixed auto update 2019-01-03 21:11:23 \o/ 2019-01-03 21:12:56 if you push to handbook it will rebuild 2019-01-03 21:13:26 if you push to docs it will update git repo clean the cache and build 2019-01-03 21:14:20 do you want a nice irc message saying new version is live? ;-) 2019-01-04 03:31:20 I think I'm good ;) 2019-01-04 03:31:34 (sorry for delay, I passed out and then forgot this was going on) 2019-01-07 21:37:34 as a side-update, AsciiDoc is now being standardized ( https://asciidoctor.org/news/2019/01/07/asciidoc-spec-proposal/ ) - once this is done, the "lightweight implementation" clandmeter wanted could become a reality (since there will be a hard spec to conform to) \o/ 2019-01-07 21:38:02 <_ikke_> nice 2019-01-07 21:38:06 :) 2019-01-08 08:20:49 wow 2019-01-08 08:20:53 that's great news 2019-01-09 21:04:02 what is the bets place for user FAQs? 2019-01-09 21:04:37 for things like: "what is the license for Alpine Linux?" 2019-01-09 21:05:06 "where can i find source for package X?" 2019-01-09 21:10:52 <_ikke_> I would say the website 2019-01-09 21:20:42 under what page or section? 2019-01-09 21:20:53 or on docs.a.o? 2019-01-09 21:50:38 probably its own page, perhaps the about apge 2019-01-09 21:50:39 page* 2019-01-09 21:50:49 I think the About page can fit it, and that's where I'd look if there was no separate FAQ page 2019-01-09 22:27:06 if we put it on www i guess community would be best. 2019-01-09 22:28:11 or else a new section called faq 2019-01-09 22:28:21 +1 2019-01-09 22:29:03 hiding it in the docs is probably not a good option. 2019-01-09 22:56:49 ^ 2019-01-09 22:56:52 kinda wanna show it off 2019-01-09 23:32:46 ^ I don't think a FAQ belongs in docs 2019-01-10 16:25:05 Hey clandmeter, settermj (he's been here before) is interested in writing the UI bundle/theme for beta.docs.a.o (we clearly need it; default has all sorts of funky stuff e.g in the top navbar) 2019-01-10 16:25:27 he's asking for a repo to push it to; I'll tell him to start over at github, but closer to completion it might be something to think about :) 2019-01-10 16:25:46 ok 2019-01-10 16:25:54 sounds good 2019-01-10 16:26:10 i think you have permissions to create new repo's right? 2019-01-10 16:26:20 yeah, but I don't think I have permissions to add new keys 2019-01-10 16:26:37 right 2019-01-10 16:27:43 lets do it first via github and see if it works out. 2019-01-10 16:27:53 yeah, that's what I asked them to do 2019-01-10 16:28:04 let me know when you think its ok to move. 2019-01-10 16:28:08 once it's building, deployable, and looks ok I don't think anyone would object to bringing it over to docs/ 2019-01-10 16:28:11 alright :) 2019-01-10 16:28:54 ok i need to go and prepare dinner :) 2019-01-10 16:28:59 enjoy :D 2019-01-10 16:29:01 else the wife will slap me 2019-01-18 03:43:01 ncopa: for when you have some time - if someone does `apk add so:libsomething.so.0`, and the package eventually moves to provide libsomething.so.1, how will apk handle that? 2019-01-18 03:43:42 I can see this going two ways - either the virtual for so:libsomething.so.0 will resolve once (when adding), so it'll just be the equivalent of the package name, or it'll keep looking up the specific library name in APKINDEX 2019-01-18 03:43:53 regardless of which one it is, it should likely be documented :) 2019-01-18 05:38:57 <_ikke_> SpaceToast: If I do apk add so:libfcgi.so.0, that's also what gets added to /etc/apk/world, so I don't think it gets resolved to a package 2019-01-18 09:11:46 that's kind of the point, so the question is whether or not that breaks once libfcgi.so.1 is out :) 2019-01-18 09:12:16 <_ikke_> right, was just trying to rule out one of the options 2019-01-18 09:12:25 <_ikke_> rule out / confirm 2019-01-18 09:13:58 my understanding is that so: and cmd: are considered virtuals, I haven't looked into what usermade virtuals look like that, so the consideration is that it might be stored in the same way (and thus resolved at add-time) 2019-01-18 09:15:43 <_ikke_> You could possibly ask fabled (not in this channel) 2019-01-18 09:16:18 I'm fine with waiting :) my response mostly happened because I woke up at random (it's 4 am here) 2019-01-18 09:42:06 ppl actually use apk add so:xxx? 2019-01-18 09:42:27 no idea, but it's getting documented, so the pitfalls should get documented too :) 2019-01-18 09:42:49 (I actually do use apk add cmd:xxx, so:xxx seems like a bother because you have to fill the soname out at all) 2019-01-18 09:42:50 it would probably fail if its upgraded 2019-01-18 09:43:52 hmm 2019-01-18 09:44:02 i never really looked in apk too deep 2019-01-18 09:44:23 i just complain to fabled if it doesnt work :) 2019-01-18 14:55:05 SpaceToast: apk will not upgrade the package in the mentioned case 2019-01-18 14:55:14 ncopa: ok, thanks :) 2019-01-18 14:55:51 if pkgsomething depends on so:libsomething.so.0 2019-01-18 14:56:20 and gets installed, and later libsomething is upgraded to so:libsomething.so.1, it will keep the old version 2019-01-18 14:56:25 or it will break 2019-01-18 14:56:43 I'm documenting apk-add(1) right now (well, right now I'm at $dayjob, but still) 2019-01-18 14:56:49 if you do: apk add so:libsomething.so, it will be added to "world", and "wold" will depend on that 2019-01-18 14:57:08 yeah, so: and cmd: are virtuals, and now there's confirmation that they're resolved every time 2019-01-18 14:57:33 yes 2019-01-18 14:57:49 ty ^^ 2019-01-18 18:15:40 What are considered virtuals? 2019-01-18 18:16:51 my understanding is that anything that isn't literally a package name is a virtual, so we have so:, cmd:, and anything created using apk add [-t | --virtual] 2019-01-18 18:17:27 It refers to what a package provides 2019-01-18 18:17:49 Not sure that's called virtuals 2019-01-18 18:18:01 yes, but they're also virtual packages (in the sense that you can add them); you can do `apk add so:libfoo.so.0` 2019-01-18 18:18:15 technically, based on the above, that makes the world virtual depend on those things, but that's the same in practice 2019-01-18 18:18:43 We have virtuals but they work differently 2019-01-18 18:19:04 Anyway I'm not an expert 2019-01-18 18:19:13 So don't listen to me 2019-01-18 18:19:33 even if technically these things aren't virtual packages, they're effectively treated like it 2019-01-18 18:19:40 for example, consider `apk info cmd:cpm` 2019-01-18 18:20:14 this resolves to diffutils, and I doubt you can (conceptually) explain "this examines the dependency" 2019-01-18 18:20:30 s/cpm/cmp/ 2019-01-18 18:20:30 SpaceToast meant to say: for example, consider `apk info cmd:cmp` 2019-01-18 18:21:03 Did you check out apkindex? 2019-01-18 18:21:12 Our* 2019-01-18 18:21:21 yes, I'm aware that the reason this works is because of the provides lines 2019-01-18 18:21:27 Ok 2019-01-18 18:21:50 Then we are in sync 2019-01-18 18:22:06 but to me that's an implementation detail - for example, you don't expect to be able to do `apk add file:/var/lib/lxc` 2019-01-18 18:23:04 or, to be more specific 2019-01-18 18:23:11 to me, a virtual package is something that refers to a (group of) package(s), without being a physical package itself 2019-01-18 18:23:19 which certainly includes this *and* apk add --virtual 2019-01-18 18:23:27 What I'm saying is, calling it a virtual could be confusing. 2019-01-18 18:24:02 explaining repository indexes to standard users (which is what I'm writing now, developer docs will obviously go into more details) is even worse 2019-01-18 18:24:18 and, conceptually, I think my generic definition for "virtual package" works 2019-01-18 18:24:29 it's just a superset of the technical definition specific to apk 2019-01-18 18:27:26 For me it's more like a reference 2019-01-18 18:29:11 well, you don't normally install/uninstall/examine references with your package manager (in the generic sense) :) 2019-01-18 18:29:52 would you be more comfortable if I added a footnote saying that technically, "virtual package" refers purely to [-t | --virtual] in apk? 2019-01-18 18:34:19 Not sure. Do what you think is best. It's just details. 2019-01-18 18:35:02 Just wanted to share my view which possibly is not perfect. 2019-01-18 18:35:37 And I'm on my phone, not the best way to communicate 2019-01-19 01:05:08 SpaceToast: i think you are right, the technical term is "virtual" 2019-01-19 01:05:41 i think what clandmeter is worried about is confusion with 'apk add --virtual ...' 2019-01-19 01:19:48 I'm documenting apk add --virtual in the same category, but explained separately :) 2019-01-19 01:22:31 sounds good 2019-01-21 15:59:55 hi 2019-01-21 16:00:34 <_ikke_> hey 2019-01-21 16:10:24 hi 2019-01-21 16:10:27 :) 2019-01-21 16:10:59 they appear to have left, but hi anyway! :D 2019-01-21 16:16:34 SpaceToast: I'd like to tweet about beta.docs.alpinelinux.org. where can users provide feedback? 2019-01-21 16:16:38 irc channel? 2019-01-21 16:17:02 also, do you have a twitter account i could mention? 2019-01-21 16:19:40 my twitter account is https://twitter.com/5paceToast , but I only use it for announcements now 2019-01-21 16:19:51 I can tweet it out as one, and you can retweet it as @alpinelinux if you'd like 2019-01-21 16:24:46 better! 2019-01-21 16:27:08 https://twitter.com/5paceToast/status/1087386089680240641 enjoy :) 2019-01-21 16:30:15 hum, i wonder if i brought the password for @alpinelinux account 2019-01-21 16:30:53 <_ikke_> heh 2019-01-21 16:31:11 i think i have it stored in macos 2019-01-21 16:31:14 need to reboot 2019-01-23 20:32:41 hm, forgot to rejoin after rebooting 2019-01-23 20:33:13 SpaceToast: how do the contributions work here? 2019-01-23 20:33:24 are you doing it all alone or is some patch/PR channel present? 2019-01-23 20:33:37 <_ikke_> currently it's just SpaceToast afaik 2019-01-23 20:33:42 <_ikke_> maybe one other person 2019-01-23 20:35:26 i used alpine for some months until i learned of 'update-conf' 2019-01-23 20:35:44 just felt the urge to make sure this is documented 2019-01-23 20:36:34 I'm doing it all solo for the first draft 2019-01-23 20:36:45 I have someone helping make the UI (though I haven't heard from them) 2019-01-23 20:36:56 once everything is set up, I'll try and get a patch/PR channel 2019-01-23 20:37:00 for now, suggestions go here :) 2019-01-23 20:37:10 also, I'll make sure to add `update-conf` to the section on updating packages 2019-01-23 20:37:39 I'm right on it as-is, so this was good timing :) 2019-01-23 20:40:58 alright! 2019-01-23 20:41:24 you ppl are awesome! 2019-01-23 20:42:15 thank you https://pa1.narvii.com/5678/d8168e8ab67ab982cf109996769c299ad241120b_hq.gif 2019-01-23 20:43:59 quite the reaction for a "I'll add a few sentences talking about tool that will be documented in the manual pages anyway" ;) 2019-01-23 21:04:42 speaking of - once the user handbook is done (or, rather, in review (by me and a few others)), I'm planning to take a break before starting the dev handbook to catch up on my user-aports, and submit some fixes (e.g to baselayout) for issues that I came accross while documenting the current situatino 2019-01-23 22:08:56 Great work SpaceToast! 2019-01-24 01:01:20 clandmeter: I was noticing an issue with beta.docs.a.o not updating - turns out --pull isn't a "force" flag, but is actually required for *any* fetching to happen. This may also fix our caching issues (in general). Any chance at an update? 2019-01-24 10:15:42 SpaceToast, ok now? 2019-01-24 10:16:37 <_ikke_> clandmeter: SpaceToast is in us timezone, so he's probably still sleeping :-) 2019-01-24 13:46:27 clandmeter: it seems to have updated; I'll get back to you after next time I push to confirm that the solution is long term :)