With integer subset of ubsan on, Haruna sometimes crashes in libmpv

mpv Information

mpv 0.38.0 Copyright © 2000-2024 mpv/MPlayer/mplayer2 projects
libplacebo version: v7.349.0
FFmpeg version: 7.0.2
FFmpeg library versions:
   libavutil       59.8.100
   libavcodec      61.3.100
   libavformat     61.1.100
   libswscale      8.1.100
   libavfilter     10.1.100
   libswresample   5.1.100

Other Information

• Linux version: Chimera Linux
• Kernel Version: Linux satoru 6.6.44-0-generic #1 SMP PREEMPT_DYNAMIC Sat Aug 3 07:45:14 UTC 2024 x86_64
• GPU Model: Intel Corporation WhiskeyLake-U GT2 [UHD Graphics 620] [8086:3ea0] (rev 02)
• Mesa/GPU Driver Version: 4.6 (Compatibility Profile) Mesa 24.1.5
• Window Manager and Version: mutter 46
• Source mpv: distro package
• Haruna version: 1.1.2
• Introduced in version: no idea

Reproduction Steps

With the system mpv built, with -fsanitize=signed-integer-overflow,integer-divide-by-zero (which Chimera does for everything by default), play the attached video below, like so: haruna x.mkv

Expected Behavior

Haruna plays the file successfully

Actual Behavior

Haruna crashes on startup (sometimes, usually I can reproduce this quite consistently but sometimes it can be harder to get), with this backtrace:

* thread #1, name = 'haruna', stop reason = signal SIGILL: illegal operand
  * frame #0: 0x000071da2dd8321c libmpv.so.2`mp_get_src_dst_rects(log=0x000071da0f464920, opts=<unavailable>, vo_caps=<unavailable>, video=<unavailable>, window_w=<unavailable>, window_h=<unavailable>, monitor_par=<unavailable>, out_src=0x000071da0e8f38c0, out_dst=0x000071da0e8f38b0, out_osd=0x000071da0e8f38e0) at aspect.c:0
    frame #1: 0x000071da2ddb1110 libmpv.so.2`mpv_render_context_render(ctx=0x000071da22be8fb0, params=0x000071da0e8f3a00) at vo_libmpv.c:362:13
    frame #2: 0x000071da34241838 libMpvQt.so.1`MpvRenderer::render(this=0x000071da10d33580) at mpvrenderer.cpp:60:5
    frame #3: 0x000071da30226bc5 libQt6Quick.so.6`QSGFramebufferObjectNode::render(this=0x000071da0f3b7450) at qquickframebufferobject.cpp:210:23
    frame #4: 0x000071da2e9e1408 libQt6Core.so.6`void doActivate<false>(sender=0x000071da34147ef0, signal_index=27, argv=0x000071da0e8f3b18) at qobject.cpp:4098:21
    frame #5: 0x000071da30049e10 libQt6Quick.so.6`QQuickWindowPrivate::renderSceneGraph() [inlined] QQuickWindow::beforeRendering(this=0x000071da34147ef0) at moc_qquickwindow.cpp:625:5
    frame #6: 0x000071da30049dfa libQt6Quick.so.6`QQuickWindowPrivate::renderSceneGraph(this=0x000071da2293ac30) at qquickwindow.cpp:654:13
    frame #7: 0x000071da3022fb8b libQt6Quick.so.6`QSGRenderThread::run() [inlined] QSGRenderThread::syncAndRender(this=0x000071da2d337050) at qsgthreadedrenderloop.cpp:761:12
    frame #8: 0x000071da3022eee9 libQt6Quick.so.6`QSGRenderThread::run(this=0x000071da2d337050) at qsgthreadedrenderloop.cpp:974:13
    frame #9: 0x000071da2eb3d420 libQt6Core.so.6`QThreadPrivate::start(void*) [inlined] QThreadPrivate::start(void*)::$_0::operator()(this=<unavailable>) const at qthread_unix.cpp:326:14
    frame #10: 0x000071da2eb3d2ab libQt6Core.so.6`QThreadPrivate::start(void*) [inlined] void (anonymous namespace)::terminate_on_exception<QThreadPrivate::start(void*)::$_0>(t=<unavailable>) at qthread_unix.cpp:262:9
    frame #11: 0x000071da2eb3d2ab libQt6Core.so.6`QThreadPrivate::start(arg=0x000071da2d337050) at qthread_unix.cpp:285:5
    frame #12: 0x000071da34438cbb ld-musl-x86_64.so.1`start(p=0x000071da0e8f3dc0) at pthread_create.c:208:17
    frame #13: 0x000071da3443bacb ld-musl-x86_64.so.1`__clone + 47

i also noticed that i can reproduce this more consistently when i set my power mode to power saver, and less consistently when i set it to performance

Log File

(note: this is coming from haruna/libmpv, not mpv)

log.txt

Sample Files

https://github.com/user-attachments/files/16515494/video.zip

I carefully read all instruction and confirm that I did the following:

• I tested with the latest mpv version to validate that the issue is not already fixed.
• I provided all required information including system and mpv version.
• I produced the log file with the exact same set of files, parameters, and conditions used in "Reproduction Steps", with the addition of --log-file=output.txt.
• I produced the log file while the behaviors described in "Actual Behavior" were actively observed.
• I attached the full, untruncated log file.
• I attached the backtrace in the case of a crash.

Start with aspect.c at mp_get_src_dst_rects and the attached video, reproducing with signed-integer-overflow and integer-divide-by-zero enabled. Trace the call from vo_libmpv.c:362 through mpv_render_context_render and compare behavior across power modes. Done means the sample no longer triggers the sanitizer-induced SIGILL or crash while Haruna/libmpv renders it.