#archlinux-ports | Logs for 2017-11-12
Back
[00:00:06] <eschwartz[m]> rewbycraft, deep42thought: also consider using chanop.py if your IRC client is weechat, it wraps various operator actions through chanserv as /oop etc.
[00:01:31] -!- shentey has quit [Quit: Konversation terminated!]
[00:08:14] <fsckd> so useful stuff for ops: https://wiki.archlinux.org https://archwomen.org https://wiki.archlinux.org
[00:08:15] <phrik> Title: User:Tigrmesh - ArchWiki (at wiki.archlinux.org)
[00:14:01] <fsckd> (most ops in the #archlinux- namespace use weechat, which is why all the commands given are for that. however, many of the things listed can be adapted to other clients.)
[00:21:34] because is now known as and
[01:46:14] -!- isacdaavid has quit [Quit: Leaving.]
[02:04:32] -!- and has quit [Ping timeout: 240 seconds]
[02:06:42] -!- because has joined #archlinux-ports
[02:41:17] -!- isacdaavid has joined #archlinux-ports
[03:41:48] because is now known as and
[04:40:08] -!- {levi} has quit [Quit: Leaving]
[04:54:57] -!- and has quit [Ping timeout: 240 seconds]
[04:56:57] -!- and has joined #archlinux-ports
[05:31:02] -!- Polichronucci has quit [Ping timeout: 268 seconds]
[05:33:47] -!- Polichronucci has joined #archlinux-ports
[06:10:44] -!- isacdaavid has quit [Quit: Leaving.]
[08:03:15] -!- deep42thought has joined #archlinux-ports
[08:03:15] -!- mode/#archlinux-ports [+o deep42thought] by ChanServ
[10:11:49] -!- City-busz has quit [Read error: Connection reset by peer]
[10:11:50] -!- anyone has quit [Remote host closed the connection]
[10:19:25] -!- City-busz has joined #archlinux-ports
[10:19:27] -!- ztrawhcse has joined #archlinux-ports
[10:19:44] -!- City-busz has quit [Changing host]
[10:19:44] -!- City-busz has joined #archlinux-ports
[10:53:09] -!- abaumann has joined #archlinux-ports
[10:54:40] -!- zaungast has joined #archlinux-ports
[11:05:11] <zaungast> Right now there are localization packages 5.4.3 of LibreOffice fresh in testing while the software itself is still at 5.4.2. Just saying. Arch Linux 32 (but I guess this channel isn't used for anything else right now anyway).
[11:08:20] <deep42thought> the build of libreoffice-fresh is broken
[11:08:35] <deep42thought> but it should not have moved its localization files to testing, then
[11:09:44] <deep42thought> hmm, formally spoken, there is no dependency between libreoffice-fresh and its localization packages :-/
[11:11:17] <deep42thought> I'm not sure what the error is here, probably the localization packages should depend on libreoffice-fresh?
[11:11:47] * zaungast is having a look at the upstream packaging.
[11:12:57] -!- zaungast_ has joined #archlinux-ports
[11:15:57] -!- zaungast has quit [Ping timeout: 240 seconds]
[11:20:54] <deep42thought> Is this common to Zaungästen - one is replaced by another? ;-)
[11:23:29] <zaungast_> No. I was temporarliy kicked out of the network.
[11:25:04] zaungast_ is now known as zaungast
[11:40:29] <zaungast> So the language packages are indeed depending on libreoffice-fresh only, not on some particular version. If stating something like
[11:40:29] <zaungast> depends='libreoffice>=${pkgver}'
[11:40:29] <zaungast> solved the problem we're discussing right now, it would probably be worthwhile. Should be rather addressed upstream, though, IMO.
[11:42:43] <zaungast> s/libreoffice/libreoffice-fresh/. Also, by upstream I mean Arch x86_64.
[11:42:58] <deep42thought> wait, they depend on libreoffice-fresh?
[11:43:04] <deep42thought> then the build master should see it ...
[11:43:30] <deep42thought> depending on a special version neither helps nor is necessary
[11:43:55] <zaungast> You may want to search for "depends" in https://git.archlinux.org
[11:48:13] <deep42thought> It might be, that 'makepkg --printsrcinfo' does not catch the depends=... inside the _package() function
[11:48:24] <deep42thought> (why is it inside that function in the first place?)
[11:50:16] <deep42thought> granted: It's quite hidden beneath a layer of eval
[11:50:51] <zaungast> I guess it's inside the function as the actual build is done in that for _lang... loop in order to make building that plethora of packages in a not too complicated manner.
[11:51:57] <deep42thought> I'm not sure at which point, we should fix it
[11:52:18] <deep42thought> we could grep more thoroughly for "depends" (instead of calling 'makepkg --printsrcinfo')
[11:52:31] <deep42thought> or we could add an explicit "depends=..." line to the PKGBUILD
[11:52:47] <deep42thought> or we could file a bug upstream to do either of both
[11:52:51] <jelle> yeah there is no real nice way to generate package functions I think
[11:54:13] <zaungast> Lacking deep knowledge about your build system the lack of some particular dependency seems like a plausible reason for the problem we're discussing to me: build master checks dependencies → finds libreoffice-fresh → thinks everything is fine, builds the localization packages and pushes them to the mirrors, ignoring any version of the corresponding libreoffice-fresh.
[11:54:13] <zaungast> So I'd still think depending on some particular version is the easiest solution, given it would prevent the Arch Linux 32 build system from behaving like it does right now.
[11:55:13] <deep42thought> zaungast: seeing the dependency is necessary, depending on a specific version is not
[11:55:35] <deep42thought> e.g. only sets of packages with no out-side dependencies are moved
[11:55:52] <deep42thought> where "out-side" still needs to be in the set of considered packages (e.g. staging + build list)
[11:56:01] <deep42thought> sry, I'm afk, lunch time
[11:58:14] <zaungast> > "depending on a specific version is not"
[11:58:14] <zaungast> Well, it may not be necessary in terms of the localization packages do build when the corresponding version of libreoffice-foo itself is missing. But it does not make any sense to build localization packages of a more recent version than the software suite itself. So I don't see any downside due to depending on some particular version.
[11:58:14] <zaungast> Guten Appetit.
[11:59:22] <deep42thought> Danke
[12:02:45] <abaumann> En Guete :-)
[12:30:22] -!- privilege has quit [Ping timeout: 268 seconds]
[12:43:04] -!- privilege has joined #archlinux-ports
[12:57:11] <deep42thought> yet another option would be to assume that every $pkgname-i18n run-depends on $pkgname
[12:57:19] <deep42thought> This is currently the best I can come up with
[12:57:38] <deep42thought> but I'll leave now and have a look into that issue this evening
[12:59:49] <deep42thought> https://buildmaster.archlinux32.org
[12:59:49] <phrik> Title: Todos in the build scripts (at buildmaster.archlinux32.org)
[12:59:59] -!- deep42thought has quit [Quit: Leaving.]
[13:05:03] -!- n0x90 has joined #archlinux-ports
[13:20:08] -!- n0x90 has quit [Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net
[13:51:42] -!- shentey has joined #archlinux-ports
[14:01:51] -!- p71 has quit [Read error: Connection reset by peer]
[14:40:15] -!- privilege has quit [Ping timeout: 248 seconds]
[14:52:59] -!- privilege has joined #archlinux-ports
[15:10:07] -!- privilege has quit [Ping timeout: 248 seconds]
[15:23:30] -!- privilege has joined #archlinux-ports
[17:37:31] -!- p71 has joined #archlinux-ports
[17:38:36] <rewbycraft> eschwartz[m]: chanop is indeed very handy and I use it on places where I have op capabilities. But since I don't here it's not super relevant information.
[18:19:41] -!- abaumann has quit [Quit: ChatZilla 0.9.92 [SeaMonkey 2.48/20171029041124]]
[18:48:06] -!- isacdaavid has joined #archlinux-ports
[18:59:13] ztrawhcse is now known as anyone
[19:16:23] -!- krumelmonster has joined #archlinux-ports
[19:32:11] -!- krumelmonster has quit [Ping timeout: 250 seconds]
[19:43:05] -!- deep42thought has joined #archlinux-ports
[19:43:10] -!- mode/#archlinux-ports [+o deep42thought] by ChanServ
[19:51:50] -!- krumelmonster has joined #archlinux-ports
[19:56:21] -!- krumelmonster has quit [Ping timeout: 258 seconds]
[20:01:44] -!- krumelmonster has joined #archlinux-ports
[20:07:57] -!- abaumann has joined #archlinux-ports
[20:10:14] <deep42thought> hmm, not all $pkgname-i18n packages should depend on $pkgname
[20:10:41] <deep42thought> "python-oslo-i18n" should not depend on "python-oslo" as the latter does not exist
[20:10:59] <deep42thought> and "eric-i18n" should rather depend on "eric-common"
[20:11:06] <abaumann> :-)
[20:11:42] <deep42thought> I think, it's safest, just to add explicit dependencies for the other *-i18n packages
[20:12:25] <deep42thought> these would be: libreoffice-still-i18n, thunderbird-i18n, firefox-i18n, libreoffice-fresh-i18n, kompozer-i18n
[20:20:07] -!- krumelmonster has quit [Ping timeout: 258 seconds]
[20:20:21] <deep42thought> on the other hand, modifications to the PKGBUILD will not only influence the build master's logic, but also enter the built package
[20:20:28] <deep42thought> hmmm
[20:23:51] <deep42thought> on the other hand: eric depends on eric-common; and depending on a non-existent python-oslo won't hurt either
[20:24:42] <deep42thought> I think, I'll go for the "each $pkgname-i18n depends on $pkgname" approach
[20:25:04] <deep42thought> (I'm talking about dependencies, the build master considers, not about dependencies entering the actual built package)
[21:08:04] -!- yokel has quit [Quit: Lost terminal]
[21:08:43] -!- yokel has joined #archlinux-ports
[21:24:47] -!- kl1602 has joined #archlinux-ports
[21:27:22] -!- jazztickets has joined #archlinux-ports
[21:27:39] -!- jazztickets has quit [Client Quit]
[21:33:59] <privilege> why does the build master consider the dependencies at all, if they are not in the deptree?
[21:36:02] <deep42thought> they are
[21:36:18] <deep42thought> but "makepkg --printsrcinfo" has limited capabilities and does not show them
[21:36:53] <privilege> what is it missing then, we should fix that!
[21:36:58] <deep42thought> we could/should switch to looking into the *.pkg.tar.xz at some point, maybe
[21:37:09] <privilege> (It should only miss conditional logic in a package() function)
[21:37:15] <deep42thought> "makepkg --printsrcinfo" cannot parse full bash syntax
[21:37:15] <privilege> (which is gross anyway)
[21:37:49] <deep42thought> have a look at https://git.archlinux.org
[21:37:49] <phrik> Title: PKGBUILD\trunk - svntogit/packages.git - Git clone of the 'packages' repository (at git.archlinux.org)
[21:38:28] <deep42thought> _package() { depends=(bla); }; eval "package() { _package; }"
[21:39:01] <deep42thought> so it's either the separate _package function or the eval
[21:39:05] <deep42thought> (or both)
[21:40:54] -!- abaumann has quit [Quit: ChatZilla 0.9.92 [SeaMonkey 2.48/20171029041124]]
[21:42:11] <privilege> Oh, that is pretty gross, but it works fine in the linux PKGBUILD...
[21:49:25] <privilege> You could consider asking heftig to make this PKGBUILD more --printsrcinfo friendly. It might require sed'ing though, instead of definining each function as something which calls _package
[21:50:13] <deep42thought> well, three of the other four pkgbuilds suffer the same problem
[21:55:17] <deep42thought> and honestly, I find the sed magic more cumbersome than artificially inserting the dependency $pkgname for all packages $pkgname-i18n
[21:55:52] <deep42thought> but I think, the king's way would be to look at the built packages in the end
[21:58:13] <privilege> and that would be rather cumbersome in the sense that you shouldn't rely on foreign binaries to determine build order. But I don't see why it matters -- those "dependent" packages don't seem to be makedepends...
[21:58:27] <privilege> So who cares what the build order is?
[21:58:39] <deep42thought> ah you're right
[21:58:46] <deep42thought> I was thinking of our binary packages
[21:59:02] <deep42thought> and considering these dependencies for moving packages around between staging, testing and stable
[21:59:54] <privilege> Well, consider the deptree graph of built packages when moving packages around the repos. Consider the makedepends graph *and top-level depends* when considering build master ordering.
[22:01:17] <deep42thought> yes
[22:01:21] <deep42thought> this is still a todo
[22:01:56] <deep42thought> currently, the built packages are handled by the run-dependencies of the build-assignment (e.g. all parts of a split packages are handled as one undividable set)
[22:02:43] <deep42thought> this is, because there is (currently) only one set of metadata - which is extracted from the PKGBUILD right at the beginning
[22:03:56] <deep42thought> e.g. https://buildmaster.archlinux32.org
[22:03:58] <phrik> Title: Todos in the build scripts (at buildmaster.archlinux32.org)
[22:05:57] -!- krumelmonster has joined #archlinux-ports
[22:12:21] -!- shentey has quit [Quit: Konversation terminated!]
[22:23:21] -!- krumelmonster has quit [Ping timeout: 250 seconds]
[22:29:46] -!- zaungast has quit [Quit: Enough for now.]
[23:02:21] -!- {levi} has joined #archlinux-ports
[23:17:39] <deep42thought> ok, I added tyzoid's and rewbycraft's master keys and abaumann's packager key
[23:18:20] <tyzoid> :)
[23:18:30] <tyzoid> deep42thought: I assume this means abaumann's key is now trusted?
[23:18:49] <deep42thought> when the archlinux32-keyring package is built and people installed it: yes
[23:45:51] -!- kl1602 has quit [Quit: Leaving]
[23:53:55] -!- kendidy has joined #archlinux-ports
[23:54:23] -!- kendidy has parted #archlinux-ports