#archlinux32 | Logs for 2023-07-08

Back
[00:53:20] -!- ConSiGno has joined #archlinux32
[05:47:46] -!- abaumann has joined #archlinux32
[05:47:47] <buildmaster> Hi abaumann!
[05:47:47] <buildmaster> !rq abaumann
[05:47:48] <phrik> buildmaster: <abaumann> YASMSBOP: yet another silly make system based on python
[05:47:53] <abaumann> I cannot reschedule building of packages.
[05:47:54] <abaumann> ./schedule-for-rebuild -w -p '^cython$'
[05:47:54] <abaumann> cython 07da1902cfcbf30a9836ed9ae3adad6f20a9e5b6 78e703995b26dc46b068523c9939dcbc597f6718 community
[05:47:57] -!- abaumann has quit [Client Quit]
[07:05:06] -!- abaumann has joined #archlinux32
[07:05:07] <buildmaster> Hi abaumann!
[07:05:07] <buildmaster> !rq abaumann
[07:05:08] <phrik> buildmaster: <abaumann> ah, cool beans, taking back all my rants (as usual)
[07:10:58] <abaumann> tons of packages missing in the buildmaster.
[07:11:10] <abaumann> also icu-72.1 which especially bad because icu 73 is out.
[07:21:27] <abaumann> ah. what a mess.
[07:24:04] <abaumann> also: when I move things out of the pool and back in again, sanity-check suddenly is happy?
[07:25:51] <abaumann> yep, and every time new packages are not ok.
[07:33:03] <abaumann> ok, we have sanity again. but I'm afraid we are just building non-working packages at the moment or wrong packages (at least for Python).
[07:34:07] <abaumann> mmh. now I have to move broken packages into archive by harnd?
[07:34:31] <abaumann> oh, and maintainance inbound: kernel 6.4.2 is pretty much a must!
[07:38:27] -!- titus_livius has joined #archlinux32
[07:40:29] <abaumann> ok, survived. I leave it running for now this way..
[07:40:31] -!- abaumann has quit [Quit: leaving]
[07:44:34] <girls> ok, finally I have a working instant messaging client again ....
[07:47:30] <girls> I'll go through the logs and comment, as I see fit
[07:48:11] <girls> * forcing a package on a certain machine A does not prevent any other machine B from building it (except, if the package is currently being built on A, then B will not get it - as usual)
[07:51:48] <girls> * why do you want to cache the result of makesourceinfo? IIRC, it should be really only called once, then everything should be in the database. And the build slaves should not call it at all.
[07:53:26] <girls> > I cannot have two package_sources with same git and mod_git revision
[07:53:26] <girls> Why do you need it? If the verbatim package builds, it's unnecessary, if it doesn't, either patch it, or add a (differntly named) shim package in build-support, which provides=(...) the needed stuff, so pacman can install it.
[07:57:24] <girls> schedule-for-rebuild has the -f flag - maybe, I should also add it to get-package-updates? But I would advise to use it carefully: You end up in a "partial upgrade" state (i.e. libraries might be missing, but the recompilation of the dependent package is scheduled)
[07:58:19] <girls> * stale lockfile cannot block anything - it's always using flock, so it only blocks if there's some process behind it.
[07:59:52] <girls> the handling of blacklisted packages is somewhat unfortunate: We cannot ignore them completely, because there's some(tm) dependency tracking in the db, thus we need to get their provides=() and depends=() for the db.
[08:04:56] <girls> return-assignment can lock up things, if the upload did not happen beforehand into the .transfer directory on the master mirror. We could enforce, that the package must be uploaded there (unlockedly) first - then it should not lock that much anymore.
[08:06:48] <girls> the buildmaster takes the repo paths from the db, IIRC
[08:07:15] <girls> mysql buildmaster -e 'select * from git_repositories'
[08:08:38] <girls> build_list_lock_file is set in lib/load-configuration - like every other default, too
[08:08:49] <girls> ah, you found that, too :D
[08:10:36] <girls> mysql_retrieve_static_information() does the setting of repo_paths__*
[08:10:40] <girls> ... and a lot more
[08:11:31] <girls> repo_paths__community and repo_paths__packages should be obsolete
[08:11:38] <girls> I forgot to delete those :)
[08:15:09] <girls> blacklisted packages are not deleted - they are moved to the deletion-list. So they will not be checked on *every* run of get-package-updates, only when they change upstream.
[08:16:05] <girls> > why is the repo part of the build command in devtools
[08:16:05] <girls> It has always been this way: Reasoning: depending on the target repository, you can use packages from different repositories for building
[08:17:35] <girls> re build-support: what do you mean by "overload"? build-support is for completely independent packages (i.e. with a unique name and a complete PKGBUILD). If you want to replace an upstream package (think of devtools or keyring), then just add a patch, that overwrites everything.
[08:19:33] <girls> why do you need get-package-updates in between *our* package builds? If we build a package, this should not be visible for get-package-updates.
[08:48:27] <girls> regarding rescheduling [community] packages: this is a hard one. We cannot build them (as we cannot build old [extra] packages, too). But we cannot remove them from the db, either (or migrate them), as they're old and upstream has a newer version (which we also track). Also, we cannot easily filter them out (at least the [extra] ones) ...
[08:49:08] <girls> I could try to migrate them all in the database - but I might definitely miss some, so this issue can pop up again ...
[08:50:02] <girls> Theoretically, only the least-stable (i.e. the newest) package should be rescheduled, though. And this should always be in a non-legacy repository.
[08:50:13] <girls> I'll check, why this is currently not the case.
[08:53:40] <girls> ah, the check, which package is the one to build, happens after running make_source_info for all of them - which fails for legacy repositories ... :-(
[09:00:06] <girls> I'm afraid, your make_source_info_using_cache is broken - it does not return something useful for "cython a4df68e36683477c6604e46d8e00deafeb8c7d7f 78e703995b26dc46b068523c9939dcbc597f6718 extra" which should work
[09:11:38] <girls> ah, my bad - the arguments were in the wrong order
[09:11:56] <girls> but with the right order, your new curl-error-code-catching seems to stumble over the redirect
[09:15:03] <girls> again my bad: should be fixed, now :)
[09:46:56] <girls> ok, schedule-for-rebuild was broken, should be fixed now
[09:47:29] <girls> In case of still-old packages lying around, it will stumble upon them and fail - in this case, use "-f" to ignore such errors and proceed with the remaining packages
[13:29:04] <KitsuWhooa> Looks like stuff is getting built again
[13:30:55] <girls> \o/
[16:05:13] -!- JerryXiao has quit [Quit: Bye]
[16:05:23] -!- JerryXiao has joined #archlinux32
[16:06:18] -!- JerryXiao has quit [Client Quit]
[16:06:28] -!- JerryXiao has joined #archlinux32
[16:11:43] -!- iamawacko has joined #archlinux32
[17:00:01] -!- abaumann has joined #archlinux32
[17:00:12] <abaumann> deep42thought: thanks for looking at things :-)
[17:00:46] <abaumann> the pkginfo cache for the current script is not helping that much, but there might be other scripts which could benefit.
[17:01:05] <abaumann> aha, so we have to build things in community first so that they hop the repository.
[17:01:18] <abaumann> I'll try again now to push some python things for bootstrapping.
[17:13:07] <abaumann> I feel a clear difference with and without pkginfo cash, but, oh well, I deleted it.
[17:13:14] <abaumann> *cache
[17:15:25] -!- abaumann has quit [Quit: leaving]