#archlinux32 | Logs for 2024-03-10

Back
[01:55:59] <KitsuWhooa> abaumann: Also, I built both dbus and dbus-broker for you. Have fun :p
[02:00:33] <KitsuWhooa> oh, and namcap works again
[07:43:32] -!- abaumann has joined #archlinux32
[07:43:32] <buildmaster> Hi abaumann!
[07:43:32] <buildmaster> !rq abaumann
[07:43:33] <phrik> buildmaster: <abaumann> I hope the placebo effect works also for coffeine :-)
[07:43:54] <abaumann> KitsuWhooa: libfido2, thanks for moving. I was under the (wrong) impression, that having systemd depend on it it might be in core.
[07:44:29] <abaumann> OTH I would suggest to merge upstreams repos from core/extra into blob. It represents the state more accurately.
[07:44:56] <abaumann> namcap working and event python.
[07:45:06] <abaumann> hey, Archlinux32 might just make it back to life. :-)
[07:55:21] <abaumann> ok, fixed python-sphinxcontrib-qthelp double repos, for future reference, this seems to work:
[07:55:25] <abaumann> ./delete-packages -i -f i486/extra-staging/python-sphinxcontrib-qthelp
[07:55:49] <abaumann> then you have to execute the intenion logs from the builder/work direcdtory of the buildmaster: . intensions/intenion.2 etc
[07:56:18] <abaumann> make sure you do a ../lib/mysql-functions before because the intension log are scripts which need some functions to be defined
[08:13:39] -!- bdju has quit [Ping timeout: 255 seconds]
[08:15:50] -!- bdju has joined #archlinux32
[08:20:17] <abaumann> yup. things are tucking along nicely now. :-)
[08:20:51] <abaumann> even the backup works again - there is nothing better than mounting a PCI card without any side bars to attack the card to the chassis. ;-)
[08:21:01] -!- abaumann has quit [Quit: leaving]
[08:38:49] -!- gehidore has quit [Ping timeout: 256 seconds]
[10:30:21] -!- zxrom has quit [Quit: Leaving]
[10:57:12] -!- gehidore has joined #archlinux32
[11:02:45] -!- gehidore has quit [Ping timeout: 256 seconds]
[11:04:31] -!- gehidore has joined #archlinux32
[12:33:39] -!- zxrom has joined #archlinux32
[12:39:44] <KitsuWhooa> abaumann: awesome, thanks! It just went insane so I'll fix it
[12:40:01] <KitsuWhooa> I swear I thought -f took a file, but maybe it's only -b that takes a file
[12:40:52] <KitsuWhooa> package multiple times in equally stable repositories: i686/{core-staging,core-staging}/cryptsetup
[12:40:52] <KitsuWhooa> package multiple times in equally stable repositories: pentium4/{core-staging,core-staging}/cryptsetup
[12:41:02] <KitsuWhooa> we're definitely hitting a bug where we build something newer than it expects and it fails
[12:41:10] <KitsuWhooa> I guess it's because we are explicitly telling it what to rebuild
[12:41:15] <KitsuWhooa> but it's not like we have a choice
[13:42:50] <KitsuWhooa> ooooh eurobuild is back
[13:44:50] <KitsuWhooa> At some point I need to try to optimise the buildmaster. The queries take *so* long
[13:44:57] <KitsuWhooa> and the server has so much more ram
[13:45:00] <KitsuWhooa> might add some indexes or something
[13:45:31] <KitsuWhooa> like, we have more builders than it can handle right now
[14:05:43] <KitsuWhooa> 15:45:00 <KitsuWhooa> might add some indexes or something <-- there seem to be indexes everywhere needed, shrug
[14:15:00] <KitsuWhooa> I wonder if it's replication that causes slowdown
[14:16:00] <KitsuWhooa> alternatively I wonder if we can add an SSD to the buildmaster
[16:28:46] <KitsuWhooa> it's impossible to run get-package-updates-ignore with this many builders
[17:30:46] <zxrom> Which version of ghidra ran on a 32-bit Windows?
[17:30:59] <zxrom> or Linux.
[17:46:12] -!- abaumann has joined #archlinux32
[17:46:12] <buildmaster> Hi abaumann!
[17:46:12] <buildmaster> !rq abaumann
[17:46:13] <phrik> buildmaster: <abaumann> wow. I don't know if this library has more bugs or CVEs..
[17:46:35] <abaumann> cryptsetup was the only package which had a pkginfo error, also there something is wrong in the computation of the new git version
[17:46:45] <abaumann> it fails upstream to get the package data
[17:47:07] <abaumann> maybe a miscalculation between the state git repo and a missing reference in the package git repo or so.
[17:48:42] <KitsuWhooa> abaumann: you mentioned something about disabling the PKGBUILD cache yesteday
[17:48:43] <KitsuWhooa> what was that about?
[17:48:50] <KitsuWhooa> (and is it still disabled?)
[17:49:15] <abaumann> no, I reenabled it, because without is was s l o w ...
[17:49:26] <KitsuWhooa> ah :p
[17:49:43] <abaumann> and I had a network outage of my ISP in the middle of a get-udpates
[17:49:53] <abaumann> that accounts for the badness of the buildmaster probably
[17:50:01] <KitsuWhooa> I ran it in screen when I had unstable internet
[17:50:03] <abaumann> (could have used screen/tmux, but maeh)
[17:50:08] <KitsuWhooa> maybe we should make it automatically spin up a screen and run itself inside it
[17:50:16] <abaumann> yes. I usually do.. unless I don't :-)
[17:50:19] <KitsuWhooa> I'm currently running a get-package-updates
[17:50:23] <KitsuWhooa> -ignore
[17:50:44] <abaumann> I forced some core stuff around mkinitcpio etc.
[17:50:51] <abaumann> let's see if everything builds.
[17:51:20] <KitsuWhooa> I got distracted. I tried building llvm, it complained about sphinx, and I said "fuck it, I'm getting sphinx working"
[17:51:52] <abaumann> I built a llvm-mos today, just to find out after 3 hours or so, that some tests fail to build in the end.
[17:51:58] <abaumann> llvm is such a monster
[17:52:07] <KitsuWhooa> sphinx fails because dvipng depends on a libheif that depends on an old libdav1d
[17:52:22] <KitsuWhooa> rebuilding libheif wants ffmpeg, which I found is broken, which wants libplacebo, which wants libdovi which wants rust
[17:52:30] <abaumann> uh. that's some dependency salad there
[17:52:31] <KitsuWhooa> I added a new up to date rust in build-support
[17:52:45] <KitsuWhooa> but then found out that upstream rust has an epoch of 1
[17:52:49] <abaumann> I gave up on documentation tools long time ago.
[17:53:03] <KitsuWhooa> so I added an epoch of 1 to the new rust in build-support, so hopefully it'll pick up on it
[17:53:13] <KitsuWhooa> if I manually inject it into a builder, it's fine
[17:53:23] <KitsuWhooa> and now I'm waiting for get-package-updates-ignore :p
[17:53:28] <abaumann> the problem with epochs is, that now upstream could also use epochs.
[17:53:33] <abaumann> this will bite us in the future
[17:53:35] <KitsuWhooa> upstream already has epoch=1
[17:53:40] <KitsuWhooa> that's why I added it locally
[17:53:42] <abaumann> they do?
[17:53:43] <abaumann> oh.
[17:53:44] <KitsuWhooa> yup
[17:53:53] <KitsuWhooa> https://gitlab.archlinux.org
[17:53:53] <phrik> Title: PKGBUILD · main · Arch Linux / Packaging / Packages / rust · GitLab (at gitlab.archlinux.org)
[17:53:56] <KitsuWhooa> line 15
[17:54:14] <abaumann> well. then. :-)
[17:54:29] <abaumann> we should not have a higher epoch downstream than upstream.
[17:54:31] <KitsuWhooa> And it's only in build support, so whatever. If something goes wrong, we undo it and delete the chroots on the builders
[17:54:35] <KitsuWhooa> yeah, agreed
[17:54:36] <abaumann> similar to pkgver.
[17:54:40] <abaumann> with exceptions, of course
[17:54:50] <KitsuWhooa> yup
[17:55:05] <abaumann> ah, mmh. build-support or build-support-manual should not matter, right.
[17:55:40] <KitsuWhooa> as for the failing llvm test
[17:55:57] <KitsuWhooa> I disabled tests on llvm16 and llvm14(?)
[17:56:08] <KitsuWhooa> but once things are stable, I plan on re-enabling and adding this https://git.alpinelinux.org
[17:56:08] <phrik> Title: allocscore.patch « llvm16 « main - aports - Alpine packages build scripts (at git.alpinelinux.org)
[17:56:33] <KitsuWhooa> it should be that which is failing
[17:56:46] <abaumann> yeah, you can also add a 'make test || true' equivalent
[17:57:07] <abaumann> this would run the test and we have to inspect results manually.. and then ignore some irrelevant tests.
[17:57:07] <KitsuWhooa> that doesn't alert on new test failures though
[17:57:12] <abaumann> true
[17:57:15] <KitsuWhooa> yeah, and no one is going to do that :p
[17:57:27] <abaumann> but adding all those patches might be cumbersome
[17:57:51] <abaumann> for sure not for llvm14 and llvm16 which are only there to keep something alive, like shaders in mesa or zig or odin, or whatever.
[17:57:52] <KitsuWhooa> I already had to fix a few " || true" seds in PKGBUILDs for tests
[17:58:02] <KitsuWhooa> so I think the burdain is already there
[17:58:19] <abaumann> " || true" == yolo
[17:59:19] * abaumann has now two spinning disk motors on his apple machine.. this is not right.
[17:59:48] <KitsuWhooa> Also, I'm going to try switching buildmaster to using bfq instead of mq-deadline
[17:59:51] <KitsuWhooa> maybe this speeds things up a bit
[18:00:54] <abaumann> speed issues are mainly single-threaded cpu things in mysql
[18:01:10] <abaumann> I doubt changing the I/O scheduler changes much here
[18:01:14] <KitsuWhooa> Yeah, I think so too
[18:01:16] <KitsuWhooa> but it's worth a try
[18:01:23] <abaumann> effectively mysql run completly in memory anyway.
[18:01:43] <KitsuWhooa> either the queries are horribly inefficient, or something else is wrong
[18:01:46] <KitsuWhooa> it shouldn't be this slow
[18:01:58] <KitsuWhooa> I was wondering if replication is also slowing it down
[18:02:16] <abaumann> I had an alert on buddyns about using up the quota on dns requests
[18:02:23] <KitsuWhooa> oh?
[18:02:48] <abaumann> and it's not that arch32 suddenly got more popular. :-)
[18:02:50] <KitsuWhooa> buildmaster uses buddydns?
[18:02:54] <KitsuWhooa> oh you mean for the domain
[18:02:57] <abaumann> yes.
[18:03:06] <abaumann> we use it for the public dns zones
[18:03:18] <abaumann> serving a dns server yourself is - well - not a good idea
[18:03:35] <KitsuWhooa> yeah
[18:03:50] <abaumann> oh. Bytedance
[18:04:08] <KitsuWhooa> oh no
[18:04:08] <abaumann> yeah, this crawler tapps on my nerves.. and everything doing SEO optimization
[18:04:35] <abaumann> some 20 years ago I actually wrote web robots in software quality :-)
[18:04:53] <abaumann> nowadays this looks like a big data center with python scripts is thrown at your web server.
[18:05:11] <KitsuWhooa> let alone all the crawlers looking for CVEs
[18:05:35] <abaumann> don't mention those. produsing a dosa and then asking you, if you need their services. :->
[18:05:56] <abaumann> My answer is usually the book of laws and telling them to f*k of.
[18:05:59] <abaumann> *off
[18:06:04] <KitsuWhooa> lol
[18:07:18] <abaumann> people thing, white hacking and cypersecurity and pen testing is fun. I actually think that keeping a system safe on the other side is the true job.
[18:07:34] <KitsuWhooa> I think you need the former for the latter
[18:07:44] <abaumann> and damn hard, because you have externals and internals (also called users). Hard to say, which ones are worse from time to time.
[18:08:50] <abaumann> to be frank. most hipped attacks are not really the dangeerous things. Privilege escalation, yeah, but there is only a user running tomcat/http and root on that system.
[18:09:01] <abaumann> there are by definition no other users around which can escalate to root.
[18:09:14] <KitsuWhooa> ugh
[18:09:21] <KitsuWhooa> schedule-for-rebuild isn't showing the commit I just made
[18:09:30] <abaumann> side-channel attacks: yeah, sure. But I'm not putting a web application of Mr. X along to banking software onto the same server.
[18:09:33] <abaumann> so, irrelevant.
[18:09:45] <abaumann> mmh..
[18:09:58] <KitsuWhooa> > rust176-bin 0000000000000000000000000000000000000000 a3f72677edf33c9dd7b2774ea8484831fb00b044 build-support
[18:10:02] <KitsuWhooa> that's from get-package-updates
[18:10:06] <KitsuWhooa> 1 any rust176-bin 0000000000000000000000000000000000000000 803fb2a3649e3173754bd8e35c792cd8da770812 build-support
[18:10:10] <KitsuWhooa> and this is from schedule-for-rebuild
[18:11:37] <abaumann> urgh, used 75% of the buddyns quota in 1/4 of a month. This could be tight for the rest of 3/4 of the month..
[18:11:46] <KitsuWhooa> ouch
[18:12:08] <abaumann> anyway, what is any robot doing on the buildmaster!!
[18:12:20] <KitsuWhooa> that I do not know
[18:12:27] <abaumann> maybe buildmaster-status leaked.
[18:12:40] <KitsuWhooa> but also
[18:12:44] <KitsuWhooa> why would a robot keep resolving the domain
[18:12:47] <KitsuWhooa> that's insane
[18:13:00] <abaumann> true.
[18:13:16] <abaumann> I'm just following strange behaviour.
[18:13:29] <abaumann> More likely is something doing a lot of DNS requests.. like a mail server.
[18:20:50] <KitsuWhooa> [master@buildmaster ~]$ prioritize-build-list -d -w <(printf '^rust176-bin$' )
[18:20:50] <KitsuWhooa> 81
[18:20:58] <KitsuWhooa> I smell insanity
[19:33:32] -!- abaumann has quit [Quit: leaving]