#archlinux32 | Logs for 2024-08-03
Back
[04:42:23] -!- ssserpent has joined #archlinux32
[07:05:00] -!- abaumann has joined #archlinux32
[07:05:01] <buildmaster> Hi abaumann!
[07:05:01] <buildmaster> !rq abaumann
[07:05:02] <phrik> buildmaster: <abaumann> no harm, if I patch things ahead, I presume? ;-)
[07:06:25] <abaumann> I just want to mention that the reason for building i486 in i486 virtual machines is because most packages don't to proper probing and setting of CPU features nowadays. While building pentium4 with SSE2 is much closer to the current ISA-level of 32-bit builds, i486 isn't. Switching to chroots in 486 leads to much more patching required (so much, that I think, it's unfeasible).
[07:06:42] <abaumann> i486 is broken, because nobody really maintains it at the moment. :-)
[07:07:45] <abaumann> I doubt, that a 32-bit process in a container or in a virtual machine can get more than 4 GB of RAM. The only way to use more memory during build is to actually cross-compile. Which is even harder to do, especially if you have to follow the upstream package dependencies (which are far from ideal for cross-compiling).
[07:08:00] -!- abaumann has quit [Client Quit]
[10:01:51] <KitsuWhooa> abauamann: in a 32 bit environment a process can't allocate the full 32 bit address space
[10:02:02] <KitsuWhooa> inside a chroot with a 64 bit kernel it can
[10:02:09] <KitsuWhooa> so it is still restricted by the 32 bit address space, but it has more of it
[10:20:58] <KitsuWhooa> and I still think we can patch the cpu detection in i486. In fact, I was going to do that after I was done with experimenting with i686 to stop sse2 from creeping in
[10:21:05] <KitsuWhooa> but too bad buildmaster is broken
[10:21:13] <KitsuWhooa> (yes I am still trying to fix it)
[10:22:07] <KitsuWhooa> abaumann: If I were to attempt to restore an older version of the database, say from a month ago, where would I find the backups?
[10:22:18] <KitsuWhooa> I assume they are sql files that I just pipe into the mysql command
[10:22:54] -!- texus has joined #archlinux32
[10:45:14] -!- texus has quit [Quit: WeeChat 4.3.5]
[11:04:09] <KitsuWhooa> I was going to add the decomissioned server as a builder, but there's no point now that buildmaster is broken. Oh well
[11:12:17] -!- ssserpent has quit [Quit: WeeChat 4.3.5]
[11:16:38] -!- ssserpent has joined #archlinux32
[11:24:04] -!- ssserpent has quit [Quit: WeeChat 4.3.5]
[11:30:47] -!- ssserpent has joined #archlinux32
[11:41:35] -!- abaumann has joined #archlinux32
[11:41:36] <buildmaster> Hi abaumann!
[11:41:36] <buildmaster> !rq abaumann
[11:41:37] <phrik> buildmaster: <abaumann> we should aim at less broken packages anyway ;-)
[11:43:01] <abaumann> KitsuWhooa: there are backups on /data/backup/mysql, but this is for the whole database, not just the buildmaster (so also fluxbb, flyspray, some house keeping stuff (mirror stats).
[11:43:10] <KitsuWhooa> Ah that's okay
[11:43:16] <KitsuWhooa> I'll figure out how to split it
[11:43:24] <abaumann> Furthermore, mysql on the buildmaster is master to the slave on archlinux32.org
[11:43:41] <abaumann> so you have to wipe the slave state there and then resync.
[11:44:01] <abaumann> there is a script re-bootstrap-slave-db for this as root.
[11:44:21] <abaumann> yes, the buildmaster is broken. I personally don't think there is much hope to fix it.
[11:44:36] <abaumann> The code is too convoluted for my brain. :-)
[11:45:04] <abaumann> Another issue: the database of the buildmaster is then out of sync with the status of the web mirror and the archive, which will cause additional problems.
[11:45:13] <abaumann> there is no easy way to sync that.
[11:45:46] <abaumann> Is going back to an older state really the solution?
[11:46:05] <abaumann> Maybe it's just some simple git states not updating or broken pkginfo data halting the current process?
[11:47:40] -!- texus has joined #archlinux32
[11:49:37] <abaumann> oh. I hope it's ok I give you access to archlinux32.org
[11:49:47] <KitsuWhooa> Ah I see
[11:50:01] <KitsuWhooa> I think something happened that made build master not mark packages as broken
[11:50:09] <KitsuWhooa> I don't think it's some bad pkginfo
[11:50:22] <KitsuWhooa> And I wonder if it had something to do with my builders dying
[11:50:51] <KitsuWhooa> I haven't checked the dates yet but it's possible something broke when the builder crashed at "the right moment"
[11:50:58] <KitsuWhooa> I'm speculating but who knows
[11:51:28] <KitsuWhooa> On the other hand, if it is some bad pkginfo it'll break instantly again
[11:51:32] <KitsuWhooa> So we'll know
[11:51:54] <KitsuWhooa> <abaumann> yes, the buildmaster is broken. I personally don't think there is much hope to fix it. <-- it's not like we have a choice :p
[11:53:20] <abaumann> ssh -p 9586 tasos@archlinux32.org
[11:53:27] <abaumann> sudo su - should work
[11:53:38] <abaumann> last time I checked there was _NO_ backup for this machine, so careful :-)
[11:54:02] <abaumann> Though, no. Actually, I do have a backup, but I don't have easy access to a KVM or something
[12:10:19] <KitsuWhooa> That makes me wary of touching it :p
[12:16:47] -!- texus has quit [Quit: WeeChat 4.3.5]
[12:17:24] -!- ssserpent has quit [Ping timeout: 252 seconds]
[12:19:30] -!- ssserpent has joined #archlinux32
[12:22:22] -!- texus has joined #archlinux32
[13:03:03] -!- ssserpent has quit [Ping timeout: 252 seconds]
[13:04:33] <abaumann> yeah, we could also disable the sync for a while and let archlinux32.org run on an older version of buildmaster. on archlinux32.org it only use the buildmaster database. The fluxxbb and flypray databases are also synched, just to have a failover.
[13:04:48] -!- ssserpent has joined #archlinux32
[13:17:40] <KitsuWhooa> I won't touch anything yet
[13:17:46] <KitsuWhooa> I'll let you know if I decide on something
[13:17:52] <KitsuWhooa> thanks for helping
[13:21:22] <abaumann> at the moment it feels more like throwing things over the wall in your direction. :-)
[13:22:06] <KitsuWhooa> that works :p
[13:28:26] -!- texus has quit [Quit: WeeChat 4.3.5]
[13:56:40] -!- abaumann has quit [Quit: leaving]
[16:40:01] -!- phrik has quit [Read error: Connection reset by peer]
[16:41:05] -!- phrik has joined #archlinux32
[17:39:00] -!- ssserpent has quit [Quit: WeeChat 4.3.5]
[20:42:42] <KitsuWhooa> 22:43:36 * buildmaster goes insane.
[20:42:46] <KitsuWhooa> how does it keep breaking itself
[20:42:51] <KitsuWhooa> more and more
[20:42:56] <KitsuWhooa> I didn't even thouch it
[20:42:58] <KitsuWhooa> *touch it
[20:47:02] <KillerWasp> KitsuWhooa: you already check the log and bash history?
[20:47:21] <KitsuWhooa> nope :p
[20:47:53] <KillerWasp> maybe exist other users that are touch it.
[20:48:22] <KitsuWhooa> Nothing in bash_history
[20:49:07] <KillerWasp> like clean? strange...
[20:49:15] <KitsuWhooa> no, not clean
[20:49:17] <KitsuWhooa> it
[20:49:25] <KitsuWhooa> *it's just my commands from last time I ran something
[20:49:38] <KitsuWhooa> The following packages appear on the build- and deletion-list: python-pygraphviz
[20:49:49] <KitsuWhooa> this isn't even loading https://buildmaster.archlinux32.org
[20:50:41] <KillerWasp> KitsuWhooa: the problem is with mysql package?
[20:50:57] <KitsuWhooa> nah, but something broke in the database
[20:50:59] <KitsuWhooa> https://buildmaster.archlinux32.org
[20:51:00] <phrik> Title: result of archlinux32 build master's sanity check (at buildmaster.archlinux32.org)
[20:53:25] <KillerWasp> say sanity check failed, but same show like passed and continue it...
[20:53:45] <KitsuWhooa> oh that's just buggy code :p
[20:53:54] <KitsuWhooa> it's been like that since forever I think
[20:54:32] <KillerWasp> i guess the script are wrong and check better the result for stop it
[20:57:03] <KillerWasp> and must* check
[21:19:25] -!- polyzen has joined #archlinux32
[21:21:39] -!- dmc has quit [Ping timeout: 260 seconds]
[21:44:31] <KitsuWhooa> well, I fixed it. We'll see what happens
[21:44:36] -!- polyzen has quit [Quit: WeeChat 4.3.4]
[21:45:48] <KitsuWhooa> nope, we're back to the same crap of building the same packages over and over
[21:49:28] -!- dmc has joined #archlinux32
[22:39:40] <KitsuWhooa> Okay, so, the script does correctly detect the failure. return-assignment gets called with "error"