#archlinux-ports | Logs for 2017-11-25

Back
[00:12:20] -!- zaungast has quit [Quit: Too late now.]
[00:33:10] -!- deep42thought has joined #archlinux-ports
[00:34:04] <tyzoid> wb deep42thought
[00:34:08] <deep42thought> hi
[00:34:14] <tyzoid> you're on early
[00:34:14] <deep42thought> I'm reading the logs
[00:34:18] <deep42thought> ... late
[00:34:21] <tyzoid> ah
[00:34:22] <deep42thought> actually
[00:34:23] <tyzoid> lol
[00:34:33] <deep42thought> "long night of science"
[00:34:38] <tyzoid> how'd it go?
[00:34:46] <deep42thought> nice :-)
[00:35:11] <deep42thought> but sometime, I want to attend as a visitor, not as a presenter
[00:38:35] <tyzoid> sounds good.
[00:38:48] <tyzoid> hey deep42thought, just wrote up an API frontend to pacman's db
[00:38:59] <tyzoid> https://pkgapi.arch32.tyzoid.com to get the list of repos
[00:39:07] <tyzoid> and https://pkgapi.arch32.tyzoid.com to list packages in the repo
[00:39:21] <tyzoid> :/ except ssl is broken
[00:39:24] <tyzoid> it works over http
[00:39:52] <deep42thought> you only show non-empty repositories?
[00:40:29] <deep42thought> hmm, no
[00:40:34] <tyzoid> I manually limit it to those four.
[00:40:40] <deep42thought> ah, ok
[00:40:41] <tyzoid> I can track more, but I use pacman's local cache for it.
[00:40:56] <deep42thought> ok, to your questions:
[00:41:13] <deep42thought> the build master has no cache of PKGBUILDs - they're created on demand
[00:41:34] <deep42thought> but it has a cache of dependencies, built packages, groups, and other stuff it's interested in
[00:41:44] <deep42thought> this is per PKGBUILD
[00:42:20] <deep42thought> then it keeps track of packages in the repositories, but it only knows of "staging/community-staging" vs. "testing/community-testing"
[00:42:39] <deep42thought> e.g. it doesn't know what packages are in core/extra/community
[00:43:51] <deep42thought> w.r.t. PKGBUILD, I'd suggest to simply concatenate them - it's almost not human-readable, but it's the only way you catch all cases - PKGBUILDs are valid bash and tend to be quite complex
[00:43:57] <tyzoid> so a couple of follow-up questions. Where is the code to generate the pkgbuild/is it possible to do this on another arch system on the fly?
[00:44:12] <deep42thought> that's easy: just concatenate
[00:44:41] <deep42thought> https://github.com
[00:44:42] <phrik> Title: builder/common-functions at master · archlinux32/builder · GitHub (at github.com)
[00:44:46] <tyzoid> Ok. Is there any authoritative master database which tracks what packages are where and in which state?
[00:44:56] <tyzoid> or is it just whatever's in the master repo's repo-db
[00:45:23] <deep42thought> staging and testing are tracked, the rest is assumed from the master mirror
[00:45:36] <tyzoid> alright, makes sense
[00:45:42] <deep42thought> but we could start tracking other repos, too
[00:45:46] <tyzoid> btw, https is working.
[00:45:50] <deep42thought> it's not that hard, actually
[00:46:05] <tyzoid> I'm still looking at making that package bug tracker
[00:47:18] <tyzoid> and I'm thinking it should be extended to handle most things regarding packages
[00:47:18] <tyzoid> including tracking versions and livecycles
[00:47:18] <tyzoid> then, we could use that to build up an archweb like interface.
[00:47:18] <tyzoid> but yea, scope creep is a thing
[00:47:18] <tyzoid> so we may need to switch to listerine.
[00:47:34] <deep42thought> ah, to use extract_source_directory(), you will find find_pkgbuilds() useful
[00:51:35] <deep42thought> hmm, actually, I think, the build master is suited better to track package states and movement than the/a bugtracker is.
[00:51:56] <tyzoid> With said scope creep, it's less of a bugtracker than a replacement for archweb
[00:51:57] <deep42thought> bugtracker and buildmaster should communicate in some way
[00:52:00] <tyzoid> which can also track bugs
[00:52:11] <deep42thought> ah, hmm
[00:52:26] <deep42thought> why not run archweb?
[00:52:37] <tyzoid> because our build infra. is entirely different?
[00:53:14] <tyzoid> I can look at seeing what it needs
[00:53:23] <deep42thought> yes, but we could probably use quite a portion of it
[00:53:26] <tyzoid> but that's going to be hours of looking through source code
[00:53:30] <tyzoid> if not day
[00:53:31] <tyzoid> day*
[00:53:33] <tyzoid> days*
[00:53:47] <deep42thought> well, better than hours/days of writing source code?
[00:54:02] <tyzoid> ehh, it's active time, instead of passive time.
[00:54:13] <tyzoid> I get bored quickly just looking at code.
[00:54:19] <deep42thought> :-)
[00:54:33] <tyzoid> doing feels good.
[00:54:44] <tyzoid> And may be occasionally fun.
[00:54:54] <tyzoid> All part of a balanced breakfast.
[00:55:00] <deep42thought> well, maybe you're right
[00:55:26] <deep42thought> at least we already set up some parts of archweb separately, which would be depressing to tear down
[00:55:29] <tyzoid> that said, if it's 1day of reading vs 14days of writing, there's a difference.
[00:55:29] <deep42thought> ;-)
[00:55:46] <tyzoid> the only part of archweb we have, iirc, is the forum
[00:55:57] <tyzoid> unless I'm mistaken
[00:56:22] <deep42thought> semi-automatically updated download site
[00:56:30] <tyzoid> ah
[00:56:43] <deep42thought> well, it's not much
[00:57:42] <deep42thought> but maybe, it's better to implement every feature we want than to revise the whole archweb and remove every feature we don't want/need or can't use 1:1 from upstream
[00:58:10] <tyzoid> That's part of why I wanted to create a requirements spec
[00:58:40] <deep42thought> can we just use the wiki on github for that?
[00:58:57] <tyzoid> sure, we should probably create an arch32web if we haven
[00:59:01] <tyzoid> haven't already
[00:59:22] <deep42thought> I would call it "archweb32" for consistency
[00:59:29] <deep42thought> but: yes
[00:59:33] <tyzoid> btw, https://pkgapi.arch32.tyzoid.com now works
[00:59:47] <tyzoid> via https
[00:59:55] <deep42thought> yes, I've seen
[01:00:04] <deep42thought> but I can't get any details on one package
[01:00:07] <deep42thought> https://pkgapi.arch32.tyzoid.com
[01:00:13] <deep42thought> or is this not, how it should be?
[01:00:28] <tyzoid> Yup. I only implemented the package list and repo list.
[01:00:34] <deep42thought> ah, ok :-)
[01:00:48] <tyzoid> The reason is I'm feeding the data directly through the shell, and I want to get input validation right before I open it to arbitrary packages.
[01:02:36] <tyzoid> deep42thought: https://github.com
[01:02:56] <phrik> Title: Requirements · archlinux32/archweb32 Wiki · GitHub (at github.com)
[01:03:15] <deep42thought> thanks :-)
[01:05:04] <tyzoid> deep42thought: np. We can continue the discussion in the morning.
[01:05:15] <tyzoid> since I suspect you've had a long day.
[01:05:32] <deep42thought> nah, just 19h ;-P
[01:05:51] <tyzoid> ok, so you've got another 7 hours to go :P
[01:06:00] <deep42thought> I'll empty my brain into the new wiki and then go to bed ...
[01:06:06] <deep42thought> huh? why 7?
[01:06:17] <tyzoid> you don't work 28 hours a day 8 days a week?
[01:06:31] <deep42thought> only when I have beam time at the laser ...
[01:06:46] <tyzoid> :)
[01:07:56] <tyzoid> deep42thought: Be sure to put the 'nice to haves' in there, so we can keep that in mind if we ever get through all the 'need to haves'
[01:11:11] <deep42thought> well, the most stuff is only "nice to have"
[01:16:24] <deep42thought> ok, I filled some stuff into the wiki
[01:16:34] <deep42thought> feel free to add more and/or shift stuff up/down
[01:19:38] <deep42thought> good night/evening!
[01:20:03] -!- deep42thought has quit [Quit: Leaving.]
[02:16:58] -!- ianaca has quit [Remote host closed the connection]
[02:17:33] -!- ianaca has joined #archlinux-ports
[02:23:24] -!- ianaca has quit [Quit: Oops, my znc went out from under me...]
[02:26:05] -!- ianaca has joined #archlinux-ports
[02:42:09] -!- isacdaavid has quit [Quit: Leaving.]
[05:58:18] -!- hringriin_ has joined #archlinux-ports
[06:01:19] -!- hringriin has quit [Ping timeout: 248 seconds]
[06:01:20] hringriin_ is now known as hringriin
[08:03:03] -!- abaumann has joined #archlinux-ports
[08:22:46] -!- deep42thought has joined #archlinux-ports
[08:29:00] <abaumann> deep42thought: good morning :-)
[08:29:32] <deep42thought> good morning!
[08:30:53] <deep42thought> Is everything or did I leave some questions unanswered?
[08:31:12] <deep42thought> inser "clear" at the right position
[08:31:24] <abaumann> The only thing was the libre repo of parabola.
[08:31:58] <abaumann> I really wonder how they want to keep 32-bit alive.
[08:32:07] <deep42thought> they should just concat the PKGBULDs
[08:32:16] <abaumann> ..and this raised the question of how to publish the PKGBUILDs of Archlinux32.
[08:32:18] <deep42thought> and replace all other source files
[08:32:29] <deep42thought> yeah, I read that
[08:33:05] <abaumann> Because to be fair: digging into a git and reading sed commands is maybe not that obvious. :-)
[08:33:29] <deep42thought> yes, but every other way I can think of, might miss some corner cases
[08:33:40] <abaumann> true.
[08:34:15] <deep42thought> but maybe, we should provide a "more human readable PKGBUILD" somewhere next to the original one
[08:34:52] <abaumann> This would most likely require rules how to write PKGBUILDs..
[08:35:28] <abaumann> Unless there is a fancy way of getting bash to spit out the whole modified script while running. :-)
[08:36:01] <deep42thought> just some logic to (. PKGBUILD; declare -f prepare; declare -f build; ...; printf 'depends=(\n'; printf '%s\n' "${depends[@]}"; printf ')'; ... ) and alike
[08:37:17] <deep42thought> so, actually a .SRCINFO plus prepare/build/package-functions and in a slightly different format (e.g. bash) would do
[08:38:10] <abaumann> right.
[08:38:41] <abaumann> I think far more important is the dependency mechanism right now.
[08:38:52] <abaumann> I still get libreoffice-still to croak about icu.59. :-(
[08:39:16] <abaumann> Namcap is not doing a sanity check if all binaries and shared libraries of the package are found?
[08:39:34] <deep42thought> only pretty generic
[08:39:51] <deep42thought> I'll go afk for breakfast
[08:40:05] <abaumann> ok. me also.. coffee. :-)
[08:42:59] -!- abaumann has quit [Remote host closed the connection]
[08:43:57] -!- abaumann has joined #archlinux-ports
[09:42:31] <deep42thought> abaumann: libreoffice-still uses "--without-system-icu"
[09:42:36] <deep42thought> that's probably the crux
[09:42:54] <abaumann> oh.
[09:44:43] <abaumann> pacman -Ql libreoffice-still | grep icu
[09:44:54] <abaumann> then there should be a local icu inside the package, but there isn't
[09:46:55] <deep42thought> I wonder why x86_64 works, then
[09:49:30] <abaumann> bizare: the 64-bit version depends on a local icu 58 inside /urs/lib/libreoffice/program.
[09:50:01] <abaumann> our libreoffice delendes mostly onicu 59, but also libicudata and libuuc on icu 60
[09:50:32] <abaumann> maybe there alre old icu59 symlinks on build slaves?
[09:51:06] <deep42thought> oh
[09:53:07] <deep42thought> nope, all fine
[09:54:11] <abaumann> mmh.
[09:55:42] <deep42thought> the file list of the i686 and x86_64 packages differ only by an empty directory
[10:01:38] <abaumann> https://git.archlinux.org use bundled icu..
[10:01:39] <phrik> Title: svntogit/packages.git - Git clone of the 'packages' repository (at git.archlinux.org)
[10:02:25] <abaumann> So, it still uses that one for building?
[10:03:40] <abaumann> The question is: what happens if a build slave contains icu already by accident? Is libreoffice changing it's mind then?
[10:05:26] <deep42thought> hmm, dunno
[10:05:33] <deep42thought> but I also thought, pacman depends on icu somehow
[10:05:44] <abaumann> btw. where are the build logs of libreoffice-still?
[10:05:46] <deep42thought> but maybe, it only makedepnds on icu
[10:05:52] <abaumann> I just see namcap logfiles on the buildmaster
[10:05:57] <deep42thought> there are none, if it is not "broken"
[10:06:16] <abaumann> for certain values of "broken" :-)
[10:06:27] <abaumann> keeping everything is painfuil, I agree..
[10:06:33] <deep42thought> well, if the build creates a package it is not considered 'broken' anymore
[10:06:46] <abaumann> sure thing. :-)
[10:06:46] <deep42thought> well, we could resort to keeping logs for staging and testing, too
[10:06:59] <deep42thought> I'll add a TODO
[10:07:10] <abaumann> otherwise you have to rebuild the package just to see what's going on.
[10:07:13] <abaumann> (y)
[10:09:02] <deep42thought> https://buildmaster.archlinux32.org
[10:09:03] <phrik> Title: Todos in the build scripts (at buildmaster.archlinux32.org)
[10:09:43] <deep42thought> I'll go afk and be back in a few ten minutes - cu later!
[10:09:52] <abaumann> cu
[11:03:26] -!- abaumann has quit [Remote host closed the connection]
[11:21:51] -!- abaumann has joined #archlinux-ports
[11:32:49] <deep42thought> hmm, it might indeed be a problem if icu is installed on the build slave while libreoffice-still is compiled
[11:39:43] <deep42thought> hmm, icu get's installed on the build slave if libreoffice-still is being compiled
[11:43:07] <deep42thought> icu <- boost-libs <- libixion <- liborcus <- libreoffice-still
[11:47:39] <abaumann> Was suspecting something like that.
[11:48:01] <abaumann> We could revert the "local ICU patch" from upstream and use the system ICU instead (bein icu60)
[11:48:20] <abaumann> BTW: I think I killed the buildmaster :-)
[11:48:36] <abaumann> nvidia and nvidia-lts were running on crashing slaves of mine. *sorry*
[11:49:52] <abaumann> upstream nvidia is releasing broken 32-bit code in kernel modules, not sure I can fix those.
[11:50:20] <abaumann> https://bugs.archlinux32.org
[11:50:21] <phrik> Title: FS#17 : [nvidia] and [nvidia-lts] kernel modules build failure on 32-bit (at bugs.archlinux32.org)
[12:25:54] <deep42thought> I'm not sure, what's up with the build master - it errors when built packages are uploaded. But I guess, I'll first get some sleep and then look into this ;-)
[12:37:24] -!- abaumann has quit [Quit: ChatZilla 0.9.92 [SeaMonkey 2.49.1/20171107234153]]
[12:43:09] -!- abaumann has joined #archlinux-ports
[13:47:54] -!- jelle has joined #archlinux-ports
[14:00:25] -!- shentey has joined #archlinux-ports
[15:42:00] <deep42thought> abaumann: you didn't break the build master, you simply enabled a bug which has been there since the beginning (in the build slaves)
[15:45:35] <deep42thought> s/enabled/triggered/
[15:45:38] <abaumann> oh. :-)
[15:45:57] <deep42thought> the build slaves expect an error code 0 from ssh when reporting failed builds
[15:46:12] <deep42thought> but the build master answers with 2 if the package is outdated
[15:46:40] <deep42thought> which is the case if you push a patch to libreoffice-still while it is being built somewhere
[15:49:11] <deep42thought> ah, no it was nvidia
[15:49:26] <abaumann> yep. right. I pushed nvidia and nvidia-lts
[15:49:33] <abaumann> so they end up on my slave.
[15:49:50] <abaumann> then virt-manager decided to play funny tricks with my video card and my machine crashed. :-)
[15:50:16] <deep42thought> this should not be a problem
[15:50:39] <deep42thought> except that the assignment won't be handed out to a different build slave
[16:02:37] <deep42thought> abaumann: I fixed that bug, but you should update your build slave to enable the fix
[16:03:50] <deep42thought> abaumann: also, did you read the error message from the nvidia build?
[16:04:17] <deep42thought> I guess, you made a change to devtools32, which got overwritten when I updated it.
[16:09:11] <abaumann> ok. I update the slave..
[16:11:12] <deep42thought> when keeping logs of staging and testing packages, I can safely remove logs of "tested" packages, right?
[16:11:16] <abaumann> Ah yes. quite a mess. :-)
[16:11:41] <abaumann> They break anyway later in 64-bit specific code..
[16:17:32] <abaumann> deep42thought: yeah, I think it's safe to remove those log files
[16:17:45] <deep42thought> ok
[16:18:26] <deep42thought> in case the nomenclature is unclear: "tested" are packages in "testing" or "community-testing" which have been reported as working and therefor may be moved to stable
[16:20:35] <abaumann> BTW: I tested a manjaro32 today (latest 32-bit ISO) with Archlinux32 mirrors. Looks really good.
[16:20:45] <deep42thought> :-)
[16:22:35] <abaumann> Parabola on the other hand looks a little bit dead to me. :-(
[16:23:18] <deep42thought> haven't you been on their channel?
[16:23:52] <abaumann> yep. but there were just one or two users there
[16:23:56] <deep42thought> oh
[16:24:11] <abaumann> and I run the latest Parabola iso..
[16:24:48] <abaumann> the install.sh was missing. so basically the "experience" was a "plain Archlinux" one. :-)
[16:25:12] <abaumann> maybe we should write some of the developers?
[16:25:38] <deep42thought> yes
[16:26:05] <abaumann> oh, we have a Wikipedia page: https://en.wikipedia.org
[16:26:07] <phrik> Title: Arch Linux 32 - Wikipedia (at en.wikipedia.org)
[16:26:12] <abaumann> :-)
[16:26:25] <deep42thought> uii
[16:26:33] <abaumann> not my doing. :-)
[16:27:02] <abaumann> Arch Linux 32 is aimed at low- and high-end 32-bit CPUs
[16:27:05] <abaumann> *snigger*
[16:32:49] -!- T`aZ has quit [Ping timeout: 248 seconds]
[16:34:53] -!- T`aZ has joined #archlinux-ports
[16:37:09] <deep42thought> we should create mirror usage statistics to support the claim that our user base is growing (if it is)
[16:37:33] <abaumann> to get rid of the {citation needed} :-)
[16:40:33] <deep42thought> yes
[17:19:36] <deep42thought> abaumann: What do you think of the idea to explicitely add the sha sum of the x86_64 config to linux and linux-lts, so it breaks when upstream changes that file?
[17:20:41] <deep42thought> or do we rely on a breaking build if the source is new and the confid is old?
[17:41:56] <abaumann> We could do that. The eralier we detect it, the better. And breaking kernels are dangerous.
[17:47:26] <rewbycraft> tyzoid, deep42thought: Ah crud. I'll fix the mirror to not redirect. I've not had time this week to look at it because I've been pulling 19 hour days and wanted to sleep
[17:48:35] <deep42thought> rewbycraft: no hurry - I modified the dns update script to only add working mirrors :-)
[17:49:22] <rewbycraft> Fixed
[17:49:33] <rewbycraft> PRoblem was that I had nginx configured to redirect to https
[17:49:46] <rewbycraft> But I used the generic form of the redirect
[17:49:51] <rewbycraft> So it redirected to $host/$uri
[17:50:12] <rewbycraft> And when coming from the pool dns, $host is the pool dns and then things get iffy
[17:55:47] <rewbycraft> Should be fixed now though
[18:03:29] -!- shentey has quit [Quit: Konversation terminated!]
[18:05:31] -!- shentey has joined #archlinux-ports
[19:11:55] -!- isacdaavid has joined #archlinux-ports
[19:23:03] -!- shentey has quit [Quit: Konversation terminated!]
[19:54:26] <deep42thought> libreoffice-still even makedepends on icu itself O.o
[20:17:45] <tyzoid> rewbycraft: no worries :)
[20:37:52] -!- shentey has joined #archlinux-ports
[20:55:33] -!- abaumann has quit [Quit: ChatZilla 0.9.92 [SeaMonkey 2.49.1/20171107234153]]
[20:59:27] -!- zaungast has joined #archlinux-ports
[21:03:45] -!- zaungast has quit [Client Quit]
[21:17:38] -!- zaungast has joined #archlinux-ports
[21:23:37] <zaungast> Attempting to install gtk-vnc yields
[21:23:37] <zaungast> [...]
[21:23:37] <zaungast> (1/1) checking keys in keyring
[21:23:37] <zaungast> [#######################################################] 100%
[21:23:37] <zaungast> downloading required keys...
[21:23:37] <zaungast> :: Import PGP key 2048R/8218F88849AAC522E94CF470A5E9288C4FA415FA, "Jan Alexander Steffens (heftig) <jan.steffens@gmail.com>", created: 2011-08-25?
[21:23:39] <zaungast> [Y/n] n
[21:23:42] <zaungast> here. But a key download like this should be necessary any longer, should it?
[21:23:57] <zaungast> s/should be/should not be/
[21:24:24] <deep42thought> huh?
[21:26:03] <deep42thought> hmm, might be, that this package is one of the original ones
[21:26:12] <deep42thought> the file is from may 2017
[21:26:49] <deep42thought> we didn't rebuild _all_ packages - only the new ones
[21:28:09] -!- deep42thought has quit [Quit: Leaving.]
[21:55:34] -!- isacdaavid has quit [Read error: Connection reset by peer]
[22:03:41] -!- xes_ has joined #archlinux-ports
[22:06:09] -!- xes has quit [Ping timeout: 248 seconds]
[22:45:14] -!- Barralastic has joined #archlinux-ports
[22:51:35] -!- sunshavi has quit [Remote host closed the connection]
[23:16:52] -!- sunshavi has joined #archlinux-ports
[23:41:44] -!- shentey has quit [Quit: Konversation terminated!]
[23:43:18] -!- zaungast has quit [Quit: Too late now.]