#archlinux-ports | Logs for 2017-11-19

Back
[00:11:32] -!- sunshavi has joined #archlinux-ports
[00:13:00] <eschwartz[m]> deep43thought: wait, you didn't do a full rebuild of everything linked to icu, when icu 60 was released?
[00:49:21] -!- isacdaavid has joined #archlinux-ports
[01:05:50] -!- isacdaavid has quit [Quit: Leaving.]
[01:35:27] -!- dopsi has quit [Ping timeout: 240 seconds]
[01:38:00] -!- dopsi has joined #archlinux-ports
[02:09:41] -!- Edu4rdSHL has joined #archlinux-ports
[02:09:45] <Edu4rdSHL> .
[02:38:33] <fsckd> .
[03:09:55] <tyzoid> uh
[03:09:59] <tyzoid> need something?
[03:10:45] -!- Edu4rdSHL has parted #archlinux-ports
[03:11:43] -!- tyzoid has quit [Quit: WeeChat 1.9.1]
[03:39:00] -!- tyler has joined #archlinux-ports
[03:39:40] tyler is now known as tyzoid
[05:53:09] -!- tyzoid has quit [Quit: WeeChat 1.9.1]
[06:07:08] -!- tyler has joined #archlinux-ports
[06:08:01] tyler is now known as tyzoid
[06:49:53] -!- guys has quit [Ping timeout: 248 seconds]
[06:52:02] -!- guys has joined #archlinux-ports
[07:00:29] -!- guys has quit [Ping timeout: 268 seconds]
[07:01:56] -!- guys has joined #archlinux-ports
[07:47:24] -!- and has quit [Ping timeout: 255 seconds]
[07:48:46] -!- and has joined #archlinux-ports
[08:16:41] -!- and has quit [Ping timeout: 240 seconds]
[08:18:29] -!- and has joined #archlinux-ports
[09:56:38] -!- deep42thought has joined #archlinux-ports
[09:57:07] <deep42thought> eschwartz[m]: we rebuilt everything, upstream rebuilt
[09:57:26] <deep42thought> but since not all dependencies are direct, not necessarily in the correct order (e.g. after icu)
[09:57:39] <deep42thought> consider "a -> b -> icu"
[09:57:47] <deep42thought> where a needs rebuild, but b does not
[09:57:57] <deep42thought> then the build master does not "see" the dependency on icu
[09:58:07] <deep42thought> and may schedule a before icu
[10:20:41] -!- and has quit [Ping timeout: 240 seconds]
[10:22:03] -!- and has joined #archlinux-ports
[10:57:43] <deep42thought> eschwartz[m]: this seems to be the case for many packages, e.g.: deepin-music -> qt5-svg -> qt5-base
[10:58:18] <deep42thought> where deepin-music was rebuilt because of a new qt5-base version, but qt5-svg was not
[10:59:18] <deep42thought> I thought, all linked libraries needed to be explicite dependencies, but apparently I'm wrong
[11:44:55] <brtln> they should be, but this is arch, the wild west of packaging
[11:45:09] <brtln> I could rant for hours about this, trust me
[11:45:51] <deep42thought> ok, so I should open bugs for all packages not depending on their linked libraries?
[11:46:24] * deep42thought considers to write a bug submitting script
[11:46:58] <brtln> there is no policy on that
[11:47:16] <brtln> you can try, maybe some maintainers will care
[11:47:24] <deep42thought> :-/
[11:47:26] <brtln> some will "won't fix" and others ignore it
[11:47:53] <deep42thought> hmm, then it's maybe better to consider all dependencies
[11:48:00] <brtln> ideally you should use dynamic linking section for figuring out rebuild order
[11:48:10] <deep42thought> was there a discussion about that on the mailing list?
[11:48:20] <deep42thought> yeah, but I'd need to have built packages for that
[11:48:50] <deep42thought> anyway, I'll go afk, it's lunch time here
[11:48:57] <brtln> you need to know what soname has changed
[11:49:19] <brtln> https://github.com
[11:49:20] <phrik> Title: GitHub - foutrelis/arch-rebuilds: An attempt at automating large rebuilds in Arch Linux (at github.com)
[13:17:19] -!- deep42thought has quit [Remote host closed the connection]
[15:05:21] -!- and has quit [Ping timeout: 248 seconds]
[15:08:02] -!- and has joined #archlinux-ports
[15:23:30] -!- shentey has joined #archlinux-ports
[15:27:17] -!- abaumann has joined #archlinux-ports
[15:36:42] -!- deep42thought has joined #archlinux-ports
[15:38:28] <deep42thought> abaumann: did you have a look at mysql-workbench already?
[15:38:39] <abaumann> no.
[15:39:00] <abaumann> but the name rings a bell.
[15:40:00] <deep42thought> the error is some type conversion problem, which should be easy to solve for you
[15:40:15] <deep42thought> I'm not that familiar with differen int types in c on different architectures
[15:40:21] <abaumann> I'll have a look.
[15:40:26] <deep42thought> thanks :-)
[15:43:39] <abaumann> This looks interesting: https://bugs.mysql.com what worries me, it's a pacht for 6.3.9 and it obviously was not integrated into 6.3.10
[15:43:41] <phrik> Title: MySQL Bugs: #84951: Problem compiling jsonparser.cpp and jsonview.ccp on Slackware (32 bits) (at bugs.mysql.com)
[15:45:01] <deep42thought> I found that bug report, too but missed the patch
[15:45:09] <abaumann> https://bugs.mysql.com
[15:45:10] <phrik> Title: MySQL Bugs: #84886: fix build for 32bit systems (contribution) (at bugs.mysql.com)
[15:45:15] <abaumann> There is a patch file attached.
[15:45:23] <deep42thought> all mysql-workbench binaries are for x86_64
[15:45:43] <deep42thought> but they provide a source "for all platforms"
[15:46:01] <abaumann> Yeah. This will sooner or later be a problem: upstream (aka: orginal author) drops support for 32-bit intel.
[15:46:20] <deep42thought> hmmmm, not good, but I'll go afk for coffe now
[15:50:40] <tyzoid> Just bringing this up in case you haven't seen it yet: https://bbs.archlinux32.org
[15:50:41] <phrik> Title: protobuf-c and protobuf / Creating/Maintaining Packages / Arch Linux 32 Forums (at bbs.archlinux32.org)
[15:57:24] <abaumann> tyz: thanks, I just checked. testing seems ok, the problem is protobuf-c has been pushed to stable, protobuf not.
[15:57:36] <abaumann> I'll comment in the forum..
[15:58:17] <abaumann> We should simply publish protobuf-c, but I cannot do that, I think :-)
[15:58:48] <abaumann> sorry, I meant protobuf of course.
[16:02:14] <tyzoid> I can, but it's via a slow sftp session, and I don't want to do it before deep42thought gets back
[16:18:01] <deep42thought> tyzoid: please don't solely move packages on the master mirror
[16:18:10] <deep42thought> otherwise the build master will stop working
[16:18:15] <deep42thought> (it tracks the movement)
[16:19:53] <deep42thought> ok, I force update protobuf - I'm not sure, what was holding it back
[16:20:03] <deep42thought> potentially breaking more stuff :-/
[16:21:01] <deep42thought> ok, protobuf is now in stable
[16:23:24] <deep42thought> regarding necessary (soname-)rebuilds: What do you think of appending something to the pkgrel enumerating our rebuilds?
[16:24:00] <deep42thought> e.g. bla-1.3-1 (upstream rebuilt), bla-1.3-1.1 (forced rebuild no.1), bla-1.3-1.2 (2nd forced rebuild), ...
[16:24:35] <abaumann> actually a good idea.. the only drawback I see, is people see two different versions on 32 and 64 bit.
[16:25:54] <abaumann> ubuntu does something similar to the debian versioning schema
[16:27:48] <abaumann> I was actually doing something similar for packages on the OBS (open build service), because the caching of depending packages was too agressive. :-)
[16:28:14] <deep42thought> I just figured, we create a lot of cache collissions by not-enumerating our rebuilds
[16:28:32] <deep42thought> I was doing this for my private repo for years, but, well, it's private ...
[16:29:55] <deep42thought> the big advantage I see, is that, if the extra numbering is in place, I can more aggressively automatically schedule rebuilds based on soname changes
[16:31:13] <abaumann> The ARM-Archlinux people don't have a similar problem?
[16:31:26] <deep42thought> WarheadsSE: ^
[16:31:52] <deep42thought> I'll go offline now, but I'll read read the logs when I'm back, so just continue this discussion w/o me :-)
[16:32:01] -!- deep42thought has quit [Quit: leaving]
[16:32:18] <abaumann> While I'm cooking.. sweet. :-)
[16:37:38] <tyzoid> deep42thought: rgr, that's why I wanted to wait until you got back :)
[17:31:25] -!- shentey has quit [Quit: Konversation terminated!]
[17:57:59] -!- isacdaavid has joined #archlinux-ports
[17:58:44] -!- abaumann has quit [Quit: ChatZilla 0.9.92 [SeaMonkey 2.48/20171029041124]]
[18:12:43] -!- abaumann has joined #archlinux-ports
[18:13:35] -!- deep42thought has joined #archlinux-ports
[18:13:39] -!- deep42thought has parted #archlinux-ports
[18:13:48] -!- deep42thought has joined #archlinux-ports
[18:14:52] <tyzoid> wb deep42thought
[18:14:57] <deep42thought> hi tyzoid
[18:15:53] <deep42thought> so what do you think about the sub-pkgrel?
[18:16:26] <tyzoid> I was actually thinking about it before this discussion came up. Can't we have timestamps override?
[18:16:44] <tyzoid> i.e. if the build date is newer, and the version is the same, use the newer one?
[18:17:32] <deep42thought> well, this works on the build master and the mirrors
[18:17:41] <deep42thought> but you will get checksum errors in your package cache
[18:17:59] <tyzoid> so if that's the case, we're forced to use pkgrel, no?
[18:18:07] <deep42thought> and I really like the idea of "one epoch-version-pkgrel is one package"
[18:18:23] <deep42thought> well, we could "live" with the collissions
[18:18:33] <deep42thought> but thats a bad decision, I think
[18:19:25] <deep42thought> It might get problematic, if upstream uses sub-pkgrels, too (they're allowed to be "[0-9a-z.]\+", I think)
[18:19:41] <deep42thought> but they were integer so far ...
[18:20:07] <abaumann> there aren't any git versions in there too?
[18:20:19] <deep42thought> git is usually tracked in the version
[18:20:24] <deep42thought> but let me grep for it
[18:20:56] <tyzoid> deep42thought: can't we just put a decimal on it? iirc pacsort still works with that.
[18:21:18] <deep42thought> yes
[18:21:40] <abaumann> another potential problem: if Archlinux32 is the bases for another Arch-based distribution, and they use pkgrel?
[18:21:44] <deep42thought> but it will break, when we have -1.5-i686.pkg.tar.xz and upstream decides to go from -5 to -5.1
[18:22:10] <deep42thought> ok, pkgrels are integers from 1 to 200, currently
[18:23:36] -!- abaumann has quit [Remote host closed the connection]
[18:23:38] <deep42thought> it looks like alarm uses $num.$num pkgrels
[18:24:38] -!- abaumann has joined #archlinux-ports
[18:25:49] -!- abaumann has quit [Client Quit]
[18:32:00] <eschwartz[m]> pkgrel can be any decimal number, but in practice being based off a group which is based off arch is simply dotty, and arch only uses integers
[18:34:22] <deep42thought> eschwartz[m]: can't it be arbitrarily deep nested, like versions? (e.g. 1.2.3)
[18:34:23] <tyzoid> sticky situation
[18:35:09] <deep42thought> we can just introduce the convention "$pkgver-$upstream_pkgrel.$build_run", starting with $build_run=0
[18:35:22] <deep42thought> so downstream can simply append their numbering
[18:36:29] <tyzoid> deep42thought: if we repo-add the new package, wouldn't the signatures validate after a -Sy?
[18:36:39] <tyzoid> deep42thought: or does it still look in the pacman cache?
[18:37:22] <deep42thought> the signature is read from the repo, but the packages in the cache won't be updated
[18:48:46] -!- abaumann has joined #archlinux-ports
[18:58:59] -!- nobodyknows2202 has joined #archlinux-ports
[18:59:00] -!- nobodyknows2202 has quit [Client Quit]
[19:07:20] <eschwartz[m]> deep42thought: looks like pacman/makepkg accepts arbitrarily-nested pkgrel
[19:08:20] <eschwartz[m]> wait, hold on a minute
[19:08:51] <eschwartz[m]> ==> ERROR: pkgrel must be a decimal, not 1.1.1.
[19:09:25] <eschwartz[m]> makepkg from git is stricter about the philosophical meaning of a pkgrel
[19:10:57] <deep42thought> eschwartz[m]: good to know
[19:11:07] <deep42thought> then we need to patch makepkg ...
[19:11:15] <eschwartz[m]> why?
[19:11:24] <deep42thought> ah, you mean one level is fine?
[19:11:39] <deep42thought> e.g. "1.1" is ok, but "1.1.1" is not?
[19:11:39] <eschwartz[m]> decimal means 1.1 is fine...
[19:11:44] <deep42thought> ok
[19:12:06] <deep42thought> I'm still not used to the concept of "." as a decimal delimiter
[19:12:09] <deep42thought> :-/
[19:12:23] <eschwartz[m]> makepkg simply lints the PKGBUILD to make sure that the documentation on "decimal" does not get circumvented by "collections of numbers and periods"
[19:25:11] <deep42thought> how does pacman sort it - by numerical value or by "version" value (e.g. is 1.10 < 1.9 or 1.10 > 1.9)?
[19:27:58] <brtln> yes, with vercmp
[19:28:12] <deep42thought> ah, good :-)
[19:29:32] <deep42thought> then: why restrict the pkgrel at all?
[19:30:36] <brtln> ask whoever wrote the code
[20:01:06] <WarheadsSE> deep42thought: mathematical decimal number, now logistic semver
[20:01:11] <WarheadsSE> Not*
[20:01:42] <WarheadsSE> It does mathematical sorting, IIRC
[20:01:45] <deep42thought> WarheadsSE: this is in general or alarm-only?
[20:02:39] <WarheadsSE> Both, we don't modify that behavior
[20:03:15] <deep42thought> k, thx
[20:07:36] <eschwartz[m]> Oh, dreisner enforced that in 2012 actually via 708a227578a8d27bb84dd150fec9c2ebc1afd117 and it has moved around in different functions since then.
[20:11:06] <WarheadsSE> Yup. We had input from dreisner over the years. Met him in person at FOSSCON a few years back
[20:22:32] <WarheadsSE> Reason being is that our version comparison will always see -5 > -4.2 so we can do localized fix/patch if absolutely needed
[20:23:01] <WarheadsSE> However, if you do that, it could really mess with downstream projects
[20:24:01] <deep42thought> my idea was to insert exactly one additional "field": -1 upstream becomes -1.0 archlinux32 (and thus -1.0.123p124.23.4 in downstream)
[20:24:19] <deep42thought> this way, downstream pkgrels should sort just fine, too
[20:25:04] -!- zaungast has joined #archlinux-ports
[20:30:10] <brtln> with i686 gone from upstream, there's no reason someone here would do "${pkgrel}.1" for other than testing purposes that won't hit official repositories
[20:30:31] <brtln> very few instances of "dot" pkgrels happened when we bumped package for one arch but not for the other
[20:30:49] <brtln> moot point since we're single arch now
[20:31:09] <brtln> (seriously we should have renamed to "the arch" with deprecation)
[20:34:29] <deep42thought> lol
[20:40:31] <deep42thought> brtln: you're "the arch linux", we're "the other arch linux" and alarm is "arches linux", then?
[20:40:46] <brtln> sounds about right
[20:42:38] <eschwartz[m]> !grab deep42thought
[20:42:38] <phrik> eschwartz[m]: Tada!
[21:17:02] <abaumann> deep42thought: my sed-fu is broken again. :-(
[21:17:25] <abaumann> I checked in a community/mysql-workbench patch, I have an eval prepare -f sed thingy in PKGBUILD, and the patch is ignored.
[21:17:54] <abaumann> Is the 'prepare' function also renamed during makepkg?
[21:18:01] <deep42thought> no
[21:18:52] <abaumann> The problem for me is, I can never test my PKGBUILD without building, I don't see the patched PKGBUILD anywhere. I could try to make a source tarball of the package only, the should be the final PKGBUILD..
[21:19:46] <deep42thought> (. path/to/original/PKGBUILD; . path/to/patched/PKGBUILD; declare -f prepare)
[21:19:51] <deep42thought> shows no difference
[21:20:40] <deep42thought> the problem is, that "declare -f" prints a semi colon at each line end
[21:20:58] <abaumann> ah. thanks, but this is a good testing method..
[21:21:21] <deep42thought> wrap this in another layer of diff and it's a little less to read
[21:22:11] <abaumann> argh. perl '$' end of line in sed. :-)
[21:29:59] <WarheadsSE> brtln: pretty much.
[21:50:00] -!- abaumann has quit [Quit: ChatZilla 0.9.92 [SeaMonkey 2.48/20171029041124]]
[22:44:26] -!- zaungast has quit [Quit: Too late now.]
[22:56:01] -!- yokel has quit [Quit: Lost terminal]
[22:57:01] -!- yokel has joined #archlinux-ports
[23:07:10] -!- deep42thought has quit [Quit: Leaving.]
[23:17:57] -!- and has quit [Ping timeout: 260 seconds]
[23:19:03] -!- and has joined #archlinux-ports
[23:19:41] -!- and has quit [Excess Flood]
[23:20:40] -!- because has joined #archlinux-ports
[23:21:51] because is now known as and