#archlinux32 | Logs for 2022-09-23
Back
[01:32:46] -!- lithiumpt has joined #archlinux32
[03:21:28] -!- bill-auger has joined #archlinux32
[06:17:53] <bill-auger> i beleive that firefox-esr 102 compiled successfully for i686 (ran out of disk space during the package() phase)
[06:18:43] <bill-auger> so it is probalby some different in v105, which is causing the trouble
[07:29:09] -!- GNUtoo has quit [Ping timeout: 258 seconds]
[07:36:07] -!- GNUtoo has joined #archlinux32
[10:09:06] -!- abaumann has joined #archlinux32
[10:09:06] <buildmaster> Hi abaumann!
[10:09:06] <buildmaster> !rq abaumann
[10:09:07] <phrik> buildmaster: <abaumann> reality is more cruel than cruel jokes about reality.. ;-)
[10:09:29] <abaumann> bill-auger: I thing an option is also to disable SIMD altogether, disable JIT, disable WASI, etc.
[10:09:48] <abaumann> this is probably all the new stuff they work on and which constantly breaks..
[10:10:47] <abaumann> and Rust definitions of architectures supporting whatever craziness Intel comes up with seems also rather shacky..
[10:11:08] <abaumann> ..especially if you want to build things for a machine _not_ having all this AVX(n) stuff
[10:11:51] <abaumann> ah, and then there is skia (grmpf)
[10:17:36] <bill-auger> i had to disable SIMD for x86_64 to build icecat this time
[10:17:50] <bill-auger> i686 always adds: ac_add_options --disable-rust-simd
[10:20:47] <bill-auger> i think this is just a regression in v105 that mozilla did not notice - they probably only target distros on a long release cycle like ubuntu/fedora
[10:24:23] <bill-auger> _rust-simd_ i had to disable for x86_64 - rust is just too brittle - backward compatibility from one release t te next is very low
[10:24:23] <bill-auger> two days ago i did not need to disable it for v105, and i did not need to disable it for v102 when that was the rolling version
[10:26:44] <bill-auger> most likely because v102 is compatble with a previous rust, and v105 is compatble with the latest rust (3 moths difference)
[10:28:36] <bill-auger> even 3 years of difference between GCCs does not break stuff like rust does in 3 months
[10:31:03] <bill-auger> and now they want to put rust in linux - not good news for rolling distros
[11:13:12] <abaumann> yeah, but I think they have to create severe damage to the Linux ecosystem first, before people start to realize and react..
[11:14:23] <abaumann> the idea with firefox-esr is good, IIRC, I hit another bug around linking there..
[11:15:07] <abaumann> An #undef MOZILLA_MAY_SUPPORT_AVX2 in SIMD_avx2.cpp seems to work..
[11:15:10] <abaumann> ..I get further..
[11:19:36] <abaumann> Upstream might get into the same trouble, firefox is built for AVX2 now.. maybe it doesn't really show in normal browsing?
[11:22:48] <abaumann> error: ‘MOZ_NEVER_INLINE’ does not name a type
[11:22:57] <abaumann> ok, jit is completely breaking with this hack.
[11:25:29] <bill-auger> naively, i wonder if it has to do with the PGO phase - we skip that for 32bit and go directly to the "optimized build" phase
[11:26:00] <bill-auger> maybe that skipped phase is where some magic happens, to determine what CPU features are supported
[11:27:07] <bill-auger> presumably mozilla nevr skips that phase for production builds, and maybe building without it are not well tested
[11:28:10] <abaumann> could be. but this pgo phase cases nothing but trouble, I thing, we disabled it for a long time already..
[11:29:11] <bill-auger> yes, it is trouble for x86_64 also - half of the release will not complete that phase unless networking is active in the build chroot (which we never do)
[11:30:42] <bill-auger> now arch has enabled build-time pip , which strictly requires internet access - presumably to downlaod sources which will not be in the source pacakage, nor in the makedepends
[11:31:06] <abaumann> oh, cool.
[11:31:23] <bill-auger> we were able to disable that - all of the pip package that it needs are already packaged in arch - but who knows if that will remain true
[11:31:57] <abaumann> I'm loosing python anyway at the moment on i486, rust is creeping in. :-)
[11:32:09] <abaumann> and I still don't have a workin rust compiler for 486..
[11:32:35] <abaumann> I also have to face the facts: things will get unbelievably hard to support..
[11:34:59] <bill-auger> i think many distro packagers would already not beleive how much effort you guys need to put into this
[11:35:53] <abaumann> yeah, and they send us Voidlinux patches to help :-)
[11:36:16] <abaumann> I know, I can copy certain things from somewhere else.. but..
[11:36:23] <abaumann> a) I loose the ability to do it mysqlf
[11:36:27] <bill-auger> i think ubuntu and fedora dropped 32bit arches - few distros still bother - debian, suse, and slackware are the only major strains i know of
[11:36:46] <abaumann> b) If everybody does that, patches will not be published anymore, or everybody stops patching.
[11:37:09] <abaumann> void, alpine, T2SDE
[11:37:54] <abaumann> It would be great, if there would be more collaboration on patches.. something like a github for patches..
[11:38:04] <abaumann> ..actually.. this should be github maybe?
[11:38:25] <abaumann> I just don't like the fact, that the central infrastructure for open source software is owned by a certain company..
[11:40:21] <bill-auger> people do that to themselves by following trends - theres no necessesity for it
[11:40:43] <abaumann> true :-)
[11:41:01] <bill-auger> GNU savannah could be "the central infrastructure for open source software" - it existed long before github - its just not as trendy
[11:41:20] <bill-auger> and/or the FSF has not as much advertising cash as github did
[11:42:59] <bill-auger> github attracted all the new young people away from savanmnah and sourceforge by "gamifying" software development - not because they offered a better service
[11:43:16] <bill-auger> in reality, they offered much fewer services
[11:43:31] <abaumann> yeah, visibility and rating (stars) is one of the main attractions, it seems..
[11:44:23] <abaumann> so.. another firefox is running.. now I have to do some shopping before the rain.. :-)
[11:45:30] <bill-auger> right, and savanmnah and sourceforge stayed with the old-school crowd who preferred simple tools SSH, FTP, email, and such, rather than using a web browser for everything
[11:46:13] <bill-auger> without "using a web browser for everything", that trendy popularity-contest stuff is not possible
[11:46:37] <abaumann> Well, you can add rating to FTP servers, if you like.. ;-)
[11:47:11] <bill-auger> yea but no jumping blinking smiley faces - those are important
[11:52:13] <bill-auger> nothing demonstrates professionalism in devel discussions, quite like winking yellow smiley faces and cartoon thumbs
[11:56:24] <bill-auger> https://archlinux.org
[11:56:24] <phrik> Title: Arch Linux - News: Removing python2 from the repositories (at archlinux.org)
[12:00:04] <bill-auger> i am probably going to need to spend a week dropping packages
[12:02:37] <bill-auger> yep about 100 packages
[12:20:41] <abaumann> oh, good.
[12:21:07] <Foxboron> :)
[12:21:40] <Foxboron> One Arch derivative has picked up every python2 package we have ever dropped, removed attribution and then redistributed everything
[12:28:02] <bill-auger> these are mainly stuff from the AUR that users requested (5-10 years ago, most of them) - probably no major losses, just a lot of housekeeping
[12:28:21] <bill-auger> 'trytond' accounts for about 20 of them
[12:28:27] <abaumann> So far I found only a local patch to ptyhon2 itself in our repos..
[12:28:51] <bill-auger> yea most of it was removed from arch about 2 years ago
[12:29:04] <Foxboron> abaumann: if you want to keep python2 you really need to pick up the backported gentoo patches
[12:29:10] <Foxboron> there are several security issues there
[12:29:56] <abaumann> Yes, my bigger problem is that python2 remains in the archive, so people will otherwise just use those, I suppose.
[12:31:21] <abaumann> uh, I'm still using offlineimap for backuping email :-)
[12:31:22] <abaumann> fun.
[12:32:03] <Foxboron> offlineimap was fixed a long time ago
[12:32:23] <abaumann> yeah, I'm running it on ArchArm ARMv6..
[12:32:32] <abaumann> ..there is bigger problem here. :-)
[12:32:33] <Foxboron> Why would that still depend on python2?
[12:33:09] <abaumann> ARMv6, I have to migrate half of my backup infrastructure (bacula, offlineimap) away from that machine first..
[12:36:10] <abaumann> makedepends_i486=(python2) in nmap, ok, that one was from bootstrapping, I suppose..
[14:18:06] -!- GNUtoo has quit [Remote host closed the connection]
[14:23:47] -!- GNUtoo has joined #archlinux32
[15:51:47] -!- bill-auger has quit [Ping timeout: 252 seconds]
[17:32:49] -!- GNUtoo has quit [Remote host closed the connection]
[17:33:17] -!- GNUtoo has joined #archlinux32
[18:19:17] -!- GNUtoo has quit [Ping timeout: 258 seconds]
[18:26:07] -!- GNUtoo has joined #archlinux32
[18:42:11] -!- drathir_tor has quit [Read error: Connection reset by peer]
[19:02:48] -!- drathir_tor has joined #archlinux32
[19:20:39] -!- lithiumpt has quit [Ping timeout: 265 seconds]
[19:31:47] oaken-so1rce is now known as oaken-source
[19:43:20] -!- lithiumpt has joined #archlinux32
[20:21:16] -!- abaumann has quit [Quit: leaving]