#archlinux-ports | Logs for 2017-07-11
Back
[00:36:05] -!- locrian has joined #archlinux-ports
[00:37:00] -!- locrian has parted #archlinux-ports
[00:43:39] -!- isacdaavid_ has joined #archlinux-ports
[00:44:26] -!- isacdaavid has quit [Read error: Connection reset by peer]
[00:49:58] isacdaavid_ is now known as isacdaavid
[00:50:40] -!- isacdaavid has quit [Quit: isacdaavid]
[00:51:04] -!- isacdaavid has joined #archlinux-ports
[04:08:49] alyptik is now known as |bin|ed
[04:12:19] |bin|ed is now known as alyptik
[04:26:12] -!- Faalagorn has quit [Remote host closed the connection]
[04:34:48] -!- isacdaavid has quit [Quit: isacdaavid]
[05:32:24] -!- isacdaavid has joined #archlinux-ports
[05:57:00] -!- dantob has joined #archlinux-ports
[06:19:42] -!- dantob has quit [Quit: dantob]
[06:21:00] -!- deep42thought has joined #archlinux-ports
[06:54:52] <deep42thought> woffs: I "fixed" botan by changing -O3 to -O2 in the Makefile -- we should remove this, once it's fixed upstream
[07:21:27] -!- deep42thought has quit [Quit: Leaving.]
[08:26:52] <brtln> you should probably switch to toolchain from testing and start the rebuild again
[08:27:39] <brtln> although I will be working yet on patching binutils to use full relro by default
[08:49:26] -!- deep42thought has joined #archlinux-ports
[08:57:23] <deep42thought> using the toolchain from testing would require us to compile testing packages (e.g. the toolchain) first
[08:58:02] <deep42thought> I think we don't want that, but maybe we can make an exemption?
[09:15:27] <brtln> I think you should
[09:15:52] <deep42thought> so what packages would that be? All from base-devel?
[09:16:19] <brtln> no, gcc, binutils, devtools, glibc, pacman
[09:16:40] <brtln> I enabled these security flags that Allan promised year ago or so
[09:17:03] <brtln> so these pkgrel bumps that hit repositories now are rebuilds against testing with testing/devtools
[09:17:59] <brtln> strcat gave me a patch that should also enable full relro by default in binutils but I need to fix test suite before using it in the package, and I'll be rather short of time for this in next 2-3 weeks so you dont need to worry about it
[09:23:49] <deep42thought> I'm not really sure, we want the testing toolchain - this calls for a lot of trouble
[09:26:42] <deep42thought> when will those changes approximately move to core?
[09:26:59] <brtln> https://www.archlinux.org when all or >90% of these packages are rebuilt
[09:27:01] <phrik> Title: Arch Linux - Todo: PIE static libraries rebuild (at www.archlinux.org)
[09:37:21] <City-busz> brtln: I think it was a bad idea to add packages to the stable repos that was built with the toolchain from testing. Those packages should be in testing first.
[09:37:57] <brtln> you are free to think so
[09:38:39] <brtln> huge rebuilds like this is never good idea for staying long in [testing]
[09:38:48] <brtln> we have been there with openssl and many times before
[09:39:17] <brtln> and keeping rebuilt packages would be completely pointless
[09:41:14] <brtln> this is only about static libraries, nothing else
[09:41:28] <City-busz> If we want to support automated builds, we should apply stricter rules.
[09:42:57] <brtln> sure, feel free to pay all of us so we stop cutting corners
[09:43:20] <brtln> as long as I do everything in time I could spend with my fiancee or eating lunch at the office, I won't
[09:44:05] <brtln> we have someting in terms of tech debt huge enough to make automatic builds lower on the todo than this
[09:56:33] -!- isacdaavid has quit [Quit: isacdaavid]
[10:41:58] -!- ryuji has quit [Ping timeout: 276 seconds]
[12:43:23] -!- Cthulu201 has quit [Quit: WeeChat 1.9]
[12:58:20] -!- Cthulu201 has joined #archlinux-ports
[13:45:20] <deep42thought> City-busz: when exactly should I "empty" / apply the deletion-list which gets built up during calling "get-package-updates"?
[13:45:31] <deep42thought> This is one thing I couldn't figure out yet
[13:50:55] <City-busz> what?
[13:51:32] <deep42thought> the build master accumulates packages which have been deleted for x86_64 in a deletion-list (your proposal)
[13:51:43] <City-busz> yes
[13:51:47] <deep42thought> at some point, we need to apply this list and delete the packages in our repositories, too
[13:51:54] <deep42thought> my question: when?
[13:52:20] <deep42thought> did you have some condition in mind, which must be met to delete the packages?
[13:52:28] <deep42thought> or is it just "after some time passes"
[13:53:08] <City-busz> this depends on why the package was deleted
[13:53:31] <deep42thought> ok, but that is hard to determine automatically
[13:54:28] <City-busz> yes, it could be a manual task
[13:54:50] <deep42thought> ok, then I will keep it manual for now
[13:55:15] <deep42thought> maybe I'll add some "these packages on the deletion-list are still in the repositories" warning to the buildmasters web site
[13:56:04] <City-busz> e.g. it's possible that a package is replaced by another one, then we should wait until the new package enters to the repo before deleting the old one
[13:58:44] -!- ryuji has joined #archlinux-ports
[13:58:54] <deep42thought> yes, I was thinking of something like "if the package was deleted in revision X and all pending build orders are from after X, we can remove the package"
[13:59:01] <deep42thought> but this has other issues
[13:59:31] <City-busz> yes, this is possible, but not easy
[14:00:06] <deep42thought> the hard part is, that a build order can get replaced by a newer one
[14:02:12] <City-busz> so when we delete a package, then we should check: 1. no replacement package is pending to build, 2. no dependent package is available in the repo
[14:03:03] <deep42thought> we identify "replacement packages" by looking what a package provides?
[14:03:15] <deep42thought> provides=(), pkgname=()
[14:05:13] <City-busz> replaces=() means a replacement package
[14:05:27] <deep42thought> ok
[14:06:08] <deep42thought> I'm currently not extracting that information from the PKGBUILD
[14:06:49] <deep42thought> I would loosen condition 1 a little: a replacement package may still be pending, if an older version also replacing is already built
[14:07:31] <deep42thought> e.g. x-1.0 replacing y is in the repos, x-1.1 is pending, then we still may delete y
[14:08:02] <City-busz> yes, this is true
[14:09:43] -!- Faalagorn has joined #archlinux-ports
[14:22:13] <City-busz> In the most in the most cases it's enough to check the binary repo: check if a package contains replaces=('foo'), and if not, check if any package does not contain {check,make}depends=('foo'). If one of these are true, then the package can be deleted.
[14:23:20] <deep42thought> one can see check-/makedepends in the binary package?
[14:29:04] <City-busz> I think we can't. So we have to check the PKGBUILDs or .SRCINFOs of the built packages.
[15:49:43] -!- deep42thought has quit [Quit: Leaving.]
[17:14:11] -!- phbr has quit [Disconnected by services]
[17:24:23] -!- deep42thought has joined #archlinux-ports
[17:45:54] -!- fhdrin has quit [Quit: Leaving]
[20:22:38] -!- eschwartz has quit [Remote host closed the connection]
[20:25:03] -!- eschwartz has joined #archlinux-ports
[23:18:20] -!- isacdaavid has joined #archlinux-ports
[23:27:27] -!- isacdaavid has quit [Quit: isacdaavid]
[23:28:18] -!- p71 has quit [Read error: Connection reset by peer]
[23:29:02] -!- isacdaavid has joined #archlinux-ports