#archlinux-ports | Logs for 2026-04-20

Back
[01:05:45] -!- Charon77 has joined #archlinux-ports
[02:18:51] <Charon77> firefox build failure: Funny thing, I was able to run the profiling (not trying until completion yet) if I temporarily remove libLLVM.so.22.1, and it started (without having to resort to headless)
[02:19:53] <Charon77> I was able to build llvm-libs with debugging, rebuild firefox, reproduce the crash, and looking at the stacktrace, we got segfault on renderer thread that tries to run llvm code:
[02:19:53] <Charon77>
[02:19:59] <Charon77> /usr/src/debug/llvm/llvm-project-22.1.3.src/llvm/lib/Target/AArch64/AArch64AsmPrinter.cpp:773
[03:25:59] -!- hcmb_ has joined #archlinux-ports
[03:25:59] hcmb is now known as Guest8790
[03:25:59] -!- Guest8790 has quit [Killed (platinum.libera.chat (Nickname regained by services))]
[03:25:59] hcmb_ is now known as hcmb
[07:13:57] -!- meijin007 has joined #archlinux-ports
[07:47:06] -!- stickynotememo has quit [Ping timeout: 255 seconds]
[07:50:35] <linkmauve> bschnei, oh I see, I tend to always use serial because it’s way easier than plugging in a keyboard and a screen —I don’t even have an external keyboard here—, but I understand the need if that’s all you have. And HDMI output is a complicated beast in many SoCs, requiring many parts to be wired up, I think u-boot sometimes merges Linux drivers for that but that’s a lot of work.
[08:03:32] -!- meijin007 has quit [Ping timeout: 245 seconds]
[08:12:52] <linkmauve> fermino, instead of using the proprietary BL31, you might want to build the free software TF-a.
[08:13:23] <linkmauve> It recently gained the ability to enable the crypto extension: https://review.trustedfirmware.org
[08:13:36] <linkmauve> This was the last remaining feature the proprietary one had over it.
[08:15:08] <linkmauve> Also I think the idbloader and itb are old methods, nowadays u-boot directly builds u-boot-rockchip{,-spi}.bin which you can install as is.
[08:15:31] <linkmauve> See https://docs.u-boot.org
[08:15:32] <phrik> Title: ROCKCHIP — Das U-Boot unknown version documentation (at docs.u-boot.org)
[08:17:35] <linkmauve> Also, nothing in that PKGBUILD is specific to your board, instead it is specific to your SoC, the rk356x.
[08:19:10] <linkmauve> Charon77, is that in llvmpipe?
[08:19:41] <linkmauve> You might want to avoid profiling in such a configuration, since it’s extremely unlikely the user will want to use software OpenGL, instead they will likely have a working GPU.
[08:23:56] -!- archmatrix has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- carsme|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- integral|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- anthraxx|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- dvzrv|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- artafinde|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- foxboron|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- C0rn3j|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- antiz|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- bgyorgy|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- jelle|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- tippfehlr|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- solskogen|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- tpkessler|M has quit [Quit: Bridge terminating on SIGTERM]
[08:23:56] -!- heftig has quit [Quit: Bridge terminating on SIGTERM]
[08:24:37] <linkmauve> fermino, also you might want to enable some more configs, for instance I use btrfs everywhere and having u-boot support that fs allows me to have a single partition.
[08:25:07] <linkmauve> Instead of having to have an Ext4 or vfat partition which only hosts the kernel, the dtb and the initramfs.
[08:26:11] -!- archmatrix has joined #archlinux-ports
[08:27:22] <Charon77> are people talking over bridged (matrix?) can't see fermino's chat
[08:28:40] <linkmauve> I talk over XMPP, but I haven’t receivedz a message from them since 00:51:59 CEST.
[08:30:02] <Charon77> ah okay, i thought you're replaying to a recent msg
[08:32:36] <Charon77> anyway linkmauve no, the error is not on llvmpipe (not mesa). it happens on libLLVM
[08:32:36] <Charon77> Isolated it to "AArch64AsmPrinter::emitHwasanMemaccessSymbols".
[08:32:57] -!- drathir_tor has quit [Remote host closed the connection]
[08:33:21] -!- heftig has joined #archlinux-ports
[08:33:22] -!- anthraxx|M has joined #archlinux-ports
[08:33:25] -!- C0rn3j|M has joined #archlinux-ports
[08:33:26] -!- bgyorgy|M has joined #archlinux-ports
[08:33:26] -!- drathir_tor has joined #archlinux-ports
[08:33:26] -!- solskogen|M has joined #archlinux-ports
[08:33:28] -!- dvzrv|M has joined #archlinux-ports
[08:33:29] <linkmauve> Charon77, what part of Firefox is calling it then?
[08:33:31] -!- integral|M has joined #archlinux-ports
[08:33:36] -!- tpkessler|M has joined #archlinux-ports
[08:33:37] -!- artafinde|M has joined #archlinux-ports
[08:33:38] -!- carsme|M has joined #archlinux-ports
[08:33:39] -!- antiz|M has joined #archlinux-ports
[08:33:44] -!- tippfehlr|M has joined #archlinux-ports
[08:33:45] -!- jelle|M has joined #archlinux-ports
[08:33:48] -!- bgyorgy|M has quit [Remote host closed the connection]
[08:33:48] -!- tippfehlr|M has quit [Remote host closed the connection]
[08:33:48] -!- heftig has quit [Remote host closed the connection]
[08:33:48] -!- dvzrv|M has quit [Remote host closed the connection]
[08:33:48] -!- antiz|M has quit [Remote host closed the connection]
[08:33:48] -!- carsme|M has quit [Remote host closed the connection]
[08:33:48] -!- tpkessler|M has quit [Remote host closed the connection]
[08:33:48] -!- archmatrix has quit [Remote host closed the connection]
[08:33:48] -!- solskogen|M has quit [Remote host closed the connection]
[08:33:48] -!- anthraxx|M has quit [Remote host closed the connection]
[08:33:48] -!- integral|M has quit [Remote host closed the connection]
[08:33:48] -!- jelle|M has quit [Remote host closed the connection]
[08:33:49] -!- C0rn3j|M has quit [Read error: Connection reset by peer]
[08:33:51] -!- artafinde|M has quit [Read error: Connection reset by peer]
[08:34:37] <Charon77> It's the renderer thread
[08:34:43] <Charon77> sorry let me elaborate
[08:35:33] <Charon77> renderer thread runs code (presumably llvmpipe), which eventually calls libLLVM
[08:35:33] <Charon77> and got segfault
[08:38:21] <linkmauve> Try to use your GPU driver instead of llvmpipe.
[08:38:46] -!- archmatrix has joined #archlinux-ports
[08:41:48] <BrainDamage> the sw renders are generally quite neglected in mesa
[08:45:19] <Charon77> this is on VM, no GPU :(
[08:45:44] <Charon77> I don't quite understand if there's a bug in mesa or in libLLVM
[08:45:56] <Charon77> but I think you're right, probably doesn't make sense to run PGO without a GPU
[08:46:29] -!- heftig has joined #archlinux-ports
[08:46:30] -!- anthraxx|M has joined #archlinux-ports
[08:46:33] -!- C0rn3j|M has joined #archlinux-ports
[08:46:34] -!- bgyorgy|M has joined #archlinux-ports
[08:46:35] -!- solskogen|M has joined #archlinux-ports
[08:46:36] -!- dvzrv|M has joined #archlinux-ports
[08:46:38] <linkmauve> Report the issue to Mesa with the full traceback, and they’ll tell you if you should report it to LLVM or not afterwards.
[08:46:39] -!- integral|M has joined #archlinux-ports
[08:46:44] -!- tpkessler|M has joined #archlinux-ports
[08:46:45] -!- artafinde|M has joined #archlinux-ports
[08:46:46] -!- carsme|M has joined #archlinux-ports
[08:46:47] -!- antiz|M has joined #archlinux-ports
[08:46:52] -!- tippfehlr|M has joined #archlinux-ports
[08:46:53] -!- jelle|M has joined #archlinux-ports
[08:51:54] <Charon77> even that is difficult because it only happens if I do profiling. Running firefox under xvfb works. I was able to isolate the crash to this line in llvm: https://github.com
[08:51:54] <Charon77>
[08:51:54] <Charon77> if I `return;` before that line, it works.
[08:51:56] <phrik> Title: llvm-project/llvm/lib/Target/AArch64/AArch64AsmPrinter.cpp at llvmorg-22.1.3 · llvm/llvm-project · GitHub (at github.com)
[09:05:57] <Charon77> Okay, I found the fix
[09:06:10] <Charon77> Turns out it's a known bug, there's already a merged github PR on llvm-project
[09:06:53] <Charon77> All I need is this to be cherry-picked https://github.com
[09:06:53] <Charon77> +17 -11 diff
[09:06:55] <phrik> Title: [NFC][AArch64] Allocate AArch64Subtarget on the heap instead of stack by augusto2112 · Pull Request #176488 · llvm/llvm-project · GitHub (at github.com)
[09:07:26] <Charon77> Any recommendation? Should I raise an MR on llvm PKGBUILD, tagging ports?
[09:35:17] <Charon77> Created draft MR. Does anyone know how to assign label on gitlab?
[09:35:33] <Charon77> I saw that I needed to put "pkg::ports" label, but I don't see a way to edit labels
[09:35:43] <Charon77> Here's my MR: https://gitlab.archlinux.org
[09:35:44] <phrik> Title: Draft: [Ports:AArch64] fix AArch64AsmPrinter (!8) · Merge requests · Arch Linux / Packaging / Packages / llvm · GitLab (at gitlab.archlinux.org)
[09:36:10] <Solskogen> this is about firefox, yes?
[09:38:14] <Charon77> yes. firefox's PGO would work (even without headless) if the above changes is done to llvm.
[09:39:30] <Solskogen> It's probably easier to add "export MOZ_HEADLESS=1" to firefox's PKGBUILD until a new llvm with that fix is released.
[09:48:21] <Charon77> I'm a bit conflicted myself.
[09:48:21] <Charon77> If the goal is *just* to build firefox, why not just disable PGO? :D
[09:48:21] <Charon77> If the goal is to build _optimized firefox_, then as linkmauve mentioned, is it meaningful to do PGO with llvmpipe, without GPU?
[09:48:21] <Charon77> If the goal is to make PKGBUILD works regardless, then we should incorporate the patch, it is simple enough, especially since it's already upstream.
[09:48:21] <Charon77> The patch may be needed for anything that does llvm profiling.
[10:03:06] -!- Guest9290 has joined #archlinux-ports
[10:08:28] <Solskogen> I don't know if you get the performance boost PGO gives if firefox is compiled with MOZ_HEADLESS=1
[10:15:53] <linkmauve> What does that do? Can’t it do rendering and stuff even in headless?
[10:24:55] -!- lynxy has joined #archlinux-ports
[10:25:41] -!- filmroellchen has joined #archlinux-ports
[11:58:42] -!- bschnei has quit [Ping timeout: 256 seconds]
[12:01:28] -!- bschnei has joined #archlinux-ports
[12:14:38] <Charon77> <rant>
[12:14:39] <Charon77> I'm so pissed right now. I've spent so much time trying to fix firefox and finally figured out this whole ordeal of having to cherry-pick some commit to fix AArchAsmPrinter.
[12:14:39] <Charon77>
[12:14:39] <Charon77> Then I did pacman -Syu and firefox refused to build. Turns out it's because of rust update.
[12:15:05] <Charon77> I thougt I was crazy, maybe I messed up something. Surely nothing bad would happen just on the day I decided to work on firefox
[12:15:10] <Charon77> yet, I'm not the only one
[12:15:12] <Charon77> https://www.phoronix.com
[12:16:03] <Solskogen> welcome to the world of package building.
[12:17:17] <Charon77> how do you keep your sanity for years
[12:19:26] <Solskogen> who said I had it in the first place?
[12:20:27] <Charon77> :>
[12:20:53] <Charon77> anyway firefox won't build on rust 1.95. Tracking issue here: https://github.com
[12:20:54] <phrik> Title: No longer builds under rust 1.95 · Issue #131 · hsivonen/encoding_rs · GitHub (at github.com)
[12:21:34] <Solskogen> getting to where we are today took a lot of effort, I'll tell you that. Which is why switching to armv8-a isn't just changing some config files and call it a day.
[12:22:03] <linkmauve> Charon77, there was already an open PR fixing that: https://github.com
[12:22:04] <phrik> Title: Add missing Select trait import for Rust 1.95 by any1here · Pull Request #130 · hsivonen/encoding_rs · GitHub (at github.com)
[12:40:11] <Charon77> hmm weird, I couldn't find the announcement... am I just bad at searching?
[12:40:12] <Charon77> https://doc.rust-lang.org
[12:40:13] <phrik> Title: Rust Release Notes (at doc.rust-lang.org)
[12:49:49] <Charon77> I just realized the irony of firefox not building because of a rust update
[13:08:21] <bertptrs> Charon77: I would guess it wasn't announced because the entire module isn't stable yet and therefore subject to change
[13:10:10] <Charon77> oh wait, core::simd is nightly only?
[13:10:35] <bertptrs> correct
[13:10:59] <bertptrs> std::arch::<architecture specific simd bindings> is stable but not as convenient
[13:20:55] -!- TheDcoder has quit [Read error: Connection reset by peer]
[13:21:18] -!- TheDcoder has joined #archlinux-ports
[13:28:34] <Charon77> https://gitlab.archlinux.org
[13:28:35] <phrik> Title: Draft: Fix rust encoding_rs not building since Rust 1.95 (!4) · Merge requests · Arch Linux / Packaging / Packages / firefox · GitLab (at gitlab.archlinux.org)
[13:28:53] <Charon77> oh wait, this is not ports specific
[13:42:57] <bertptrs> not ports specific but still useful. Thanks!
[13:46:05] <bschnei> Charon77: nice work!
[14:37:06] -!- heftig-weechat has joined #archlinux-ports
[14:37:22] <heftig-weechat> ping
[14:37:32] <heftig> pong
[14:38:00] <heftig> Solskogen: say something, pleaes
[14:38:07] <heftig> s/pleaes/please/
[14:38:47] <heftig> ah, whoops, Solskogen
[14:43:40] <solskogen|M> ping
[14:44:06] -!- heftig-weechat has quit [Quit: WeeChat 4.9.0]
[14:51:54] -!- Charon77 has quit [Ping timeout: 255 seconds]