2025-03-10 16:34:23 the build of linux-lts 6.12.18 failed on loongarch: error: implicit declaration of function 'annotate_reachable' 2025-03-10 16:34:30 anyonly has time to help with that? 2025-03-10 16:34:45 <@ikke> Probably caused by https://lore.kernel.org/all/20250305174506.868421566@linuxfoundation.org/ 2025-03-10 16:35:06 i suppose we could just add a revert commit of that 2025-03-10 16:36:02 or maybe not... 2025-03-10 16:36:21 commit message says: 2025-03-10 16:36:24 And the annotate_unreachable() usage in unreachable() is plain wrong; 2025-03-10 16:36:24 it will hide dangerous fall-through code-gen. 2025-03-10 17:28:31 there are other patches in the series (https://lore.kernel.org/all/20241128094312.235637588@infradead.org/), maybe the backported patch depends on one 2025-03-10 17:44:16 https://lore.kernel.org/all/20241128094312.133437051@infradead.org/ should fix it 2025-03-10 18:02:57 ncopa: https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/81128 2025-03-10 18:03:35 is also already queued for 6.12.19: https://lore.kernel.org/stable/20250310170458.155042546@linuxfoundation.org/ 2025-03-10 19:22:57 thanks! 2025-03-11 00:50:45 Good morning! 2025-03-11 00:51:14 Thanks ncopa fossdd[m] 2025-03-11 08:54:05 ncopa: same happens with kernel 6.13.6 2025-03-11 09:37:26 Yeah, the patch mentioned by fossdd[m] can be used as a fix. https://lore.kernel.org/all/20241128094312.235637588@infradead.org/ 2025-03-11 09:41:26 https://lore.kernel.org/loongarch/20250309195622.1541936-1-ojeda@kernel.org/T/#t 2025-03-11 10:07:09 https://lore.kernel.org/all/20250310170447.729440535@linuxfoundation.org 2025-03-11 10:07:20 huajingyun: ^ 2025-03-11 10:07:50 6.13.7rc1 2025-03-13 10:55:43 In file included from /home/mps/aports/community/telegram-desktop/src/tdesktop-5.12.4-full/Telegram/lib_webview/webview/webview_dialog.cpp:13: 2025-03-13 10:55:46 /home/mps/aports/community/telegram-desktop/src/tdesktop-5.12.4-full/Telegram/lib_ui/ui/widgets/buttons.h:276:17: error: 'nullptr_t' has not been declared 2025-03-13 10:55:49 276 | nullptr_t, 2025-03-13 10:55:52 | ^~~~~~~~~ 2025-03-13 10:55:54 ninja: subcommand failed 2025-03-13 10:56:22 what should be included to fix this 2025-03-13 10:56:52 which file i mean 2025-03-13 12:12:50 Maybe need to add #include in Telegram/lib_ui/ui/widgets/buttons.h 2025-03-13 12:19:01 Aha, same error on x86_64 2025-03-13 12:19:47 yes, error is on all arches 2025-03-13 12:26:15 #include didn't helped 2025-03-13 12:26:23 The following changes seem to work: 2025-03-13 12:26:23 2. Change nullptr_t to std::nullptr_t in Telegram/lib_ui/ui/widgets/buttons.cpp, 2025-03-13 12:26:23 1. Add #include in Telegram/lib_ui/ui/widgets/buttons.h and change nullptr_t to std::nullptr_t, 2025-03-13 12:26:49 ok, will try 2025-03-13 12:26:51 Do you have time to try it? 2025-03-13 12:27:01 ok 2025-03-13 12:31:24 ACTION wonders why i take to maintain any pkg written in C++ 2025-03-13 12:32:46 also above don't work. nvm, it is not important to upgrade tdesktop 2025-03-13 12:34:46 and i don't understand C++ nor its namespace 2025-03-13 12:37:47 Well, I'm not sure, but let me try 2025-03-13 12:39:55 i will try again on clean source tree 2025-03-13 12:42:32 huajingyun: looks like it works 2025-03-13 12:42:41 thanks 2025-03-13 12:43:36 i will try on aarch64 also to if binary work 2025-03-13 12:48:27 Ok, you're welcome 2025-03-13 12:53:19 Wonder why there is an error on alpine, maybe g++ is different in some places 2025-03-13 12:55:34 mps: I'll be going AFK soon,hopefully binary will work too 2025-03-13 12:55:53 Bye 2025-03-13 12:56:18 huajingyun: thanks, i will report here later 2025-03-13 16:46:04 kernel 6.13.7 builds on loongarch64. i guess also 6.12.19 will 2025-03-13 16:48:38 huajingyun: telegram-desktop works with your proposed fixes. diff is here https://tpaste.us/Kx0o 2025-03-14 01:09:34 mps:That's good,thanks for telling me:) 2025-03-14 06:33:10 huajingyun: np, you found fix, thanks again 2025-03-14 06:43:33 You're welcome:) 2025-03-17 00:16:19 i'm preparing an upgrade for gutenprint. the new release has a more recent config.sub and the current aport updates this. the most probable issue i see is the lack of loongarchx32 in the new config.sub in the source. doen't seem to me that this should pose a problem, but maybe someone here knows better. 2025-03-17 00:23:55 here is the diff between the abuild config.sub and the new one in the source: https://tpaste.us/1yBQ 2025-03-17 01:32:52 jvvv: Yeah, since the updated config.sub is already supported in the new gutenprint release, I think "update_config_sub" in gutenprint aports can be removed 2025-03-17 01:53:16 Yes, the latest 5.3.5 release source code is already compatible; update_config_sub has become redundant after the upgrade and can be removed. 2025-03-17 02:39:46 community/miller test seems to be failing at the moment on loongarch64, has anyone come across this before? https://build.alpinelinux.org/buildlogs/build-edge-loongarch64/community/miller/miller-6.13.0-r3.log 2025-03-17 02:41:31 it passed on other arches 2025-03-17 02:44:34 (it came up during a rebuild with go 1.24.1) 2025-03-17 02:58:33 DWwanghao: Do you have time to look at community/miller? 2025-03-17 03:06:11 huajingyun:Yach, I'm looking into it. 2025-03-17 03:08:36 Thanks 2025-03-17 03:09:27 thanks! 2025-03-17 03:39:59 mio:It appears to be a precision issue in floating-point calculations. I have already reported it to the LoongArch Go compiler team, 2025-03-17 03:40:14 and I will continue to follow up. I will provide you with an update once there are any results. 2025-03-17 03:40:52 DWwanghao: thanks for the quick reply and diagnosis! 2025-03-17 03:41:12 You're welcome:) 2025-03-17 03:41:50 :) 2025-03-17 22:09:35 in some cases (such as viewing PDFs), i am seeing graphical corruption in firefox on loongarch. the nature of the graphical corruption i am seeing indicates to me that some image tiles are 2x wider than they should be 2025-03-17 22:09:56 how vetted is some of this LASX code? 2025-03-17 22:16:13 example of corruption https://usercontent.irccloud-cdn.com/file/V9MKaVdC/Screenshot_20250317_151548.png 2025-03-17 22:29:20 correction, that actually looks like it is 4x wider than it should be 2025-03-17 22:29:37 which reinforces the likelihood that this is a vectorization issue 2025-03-17 22:30:02 (4bpc vs 1bpc for 32-bit color) 2025-03-17 22:43:24 you can also download the original PDF here: https://rippling2.imgix.net/Complaint.pdf it's a great read despite the graphics corruption 2025-03-17 22:59:17 hm, not sure which patch could be responsible for this 2025-03-17 23:00:44 though as far as i can tell, both larger lasx patches were at some point touched by me 2025-03-17 23:00:50 so it's probably something i broke 2025-03-17 23:01:03 is there a reliable way to emulate loong64 on amd64 2025-03-17 23:12:35 this smells like it is in Skia 2025-03-17 23:29:38 i don't think it is any edits you've made, tab previews have had similar graphical corruption since day 1 for me. i just don't care about tab previews 2025-03-17 23:51:20 digging further, it appears the corrupted images are PNGs 2025-03-18 00:02:24 hmm, no. they are all jpegs according to my copy of acrobat 2025-03-18 00:02:33 BUT they are being resized with a mask 2025-03-18 00:02:50 i am blaming skia here 2025-03-18 04:49:15 DWwanghao: discussed the situation of miller with miller's maintainer, will temporarily skip the test cases to unblock the builder. it would be still be great if there's an update or fix, whenever it happens 2025-03-18 04:51:37 Ariadne: I've noticed the same before on Gentoo, but I gather some other distros do not have this problem 2025-03-18 04:52:27 i’m almost certain something is wrong in skia 2025-03-18 04:52:28 I'll try to debug in a few days or maybe 2 weeks, it's most likely Skia but miscompilation could also be a problem 2025-03-18 04:53:11 it seems like it is effectively 4x’ing row-stride. i used to see this with altivec code a lot back in the day. 2025-03-18 04:53:18 yeah, I've tried to pinpoint the issue but the Skia test suite (dm) doesn't expose the problem 2025-03-18 04:53:52 one option to verify the theory might be to build skia with the vector code disabled 2025-03-18 04:54:05 during my work on https://skia-review.googlesource.com/c/skia/+/908137 I compared the results on loongarch64 with amd64 and they're mostly the same 2025-03-18 04:54:30 wrt disabling SIMD, you're right, I didn't test this 2025-03-18 04:54:52 thanks for the tip, I'll try it shortly and see ;-) 2025-03-18 04:56:06 unrelated: strangely the entirety of firefox segfaults if i try to configure a lenovo t14s in the lenovo store website 2025-03-18 04:56:26 will try to get a backtrace 2025-03-18 05:02:20 I created a mr to fix hugo test failure on loongarch64 !81502 2025-03-18 05:05:20 it would be great if someone is willing to help merge it. The builder is also blocked by hugo 2025-03-18 05:08:21 CI failure? 2025-03-18 05:42:15 merged 2025-03-18 06:04:08 Ariadne: thanks! 2025-03-18 07:21:41 thanks for fixing hugo! 2025-03-18 07:34:28 Hi, ncopa, you’re welcome! 2025-03-18 07:36:53 Thanks also to Ariadne for helping with the timely merge 2025-03-18 07:38:51 mio:Okay, any further fixes or updates will be posted in the IRC in a timely manner. 2025-03-18 07:38:58 Thank you for your help in keeping the build server running. 2025-03-18 13:02:08 DWwanghao: great, thanks for following up on the progress 2025-03-21 11:41:21 Ariadne: it seems the graphic corruption issue is hairier than I expected 2025-03-21 11:41:30 webkit-gtk 2.46.5 with Skia m129 (2024-08-12) turned out fine 2025-03-21 11:41:59 but webkit-gtk 2.48.0 with Skia m135 broke in another way 2025-03-21 11:42:17 I have recorded my investigations so far at https://github.com/loongson-community/discussions/issues/87 2025-03-21 16:10:51 interesting 2025-03-28 07:31:01 hi! anyone available to help solve this build failure? https://build.alpinelinux.org/buildlogs/build-edge-loongarch64/community/wpewebkit/wpewebkit-2.46.6-r0.log 2025-03-28 07:38:18 ncopa: did you see the recent email from loongarch? 2025-03-28 07:42:25 ncopa:Let me take a look. 2025-03-28 07:57:52 I took a quick look and remembered that a similar error occurred in webkit2gtk-4.1 before 2025-03-28 07:58:01 Clang doesn't seem to have added full support for float16 yet 2025-03-28 07:58:35 DWwanghao: The changes in https://git.alpinelinux.org/aports/commit/?id=fefa3d040402 (loongarch64-fix-simde.patch) should fix the issue, please verify, thanks 2025-03-28 08:03:43 It looks like the issue is due to the lack of Float16 support.let me verify it. 2025-03-28 08:04:58 Carlo,ncopa: Thanks! And yes, I will be on vacation for a few months starting next week. If anyone have any questions during my absence, please contact my colleague DWwanghao:) 2025-03-28 08:05:41 Thanks DWwanghao ! 2025-03-28 09:49:23 hi ncopa:After applying the patch provided by Jingyun, the issue with Float16 was resolved 2025-03-28 09:49:33 However, a new problem occurred during the compilation process—Clang got stuck in an infinite loop while compiling LowLevelInterpreter.cpp.o 2025-03-28 09:50:15 This issue has already been reported to the Clang team, and I will keep you updated on any progress. 2025-03-28 09:50:54 huajingyun1: enjoy your vacation! 2025-03-31 06:06:59 hi ncopa:This mr (https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/82142) fixes the wpewebkit build failure issue.Thank you for reporting the issue! 2025-03-31 10:14:18 thanks!