commit @de0a8da broke the wayland backend #1286

KyunLFA · 2024-05-07T10:51:58Z

Self explanatory.

Backtrace:

#0  0x000076579d6a053b in pthread_kill () at /usr/lib/libc.so.6
#1  0x000076579d642048 in raise () at /usr/lib/libc.so.6
#2  0x000076579d624478 in abort () at /usr/lib/libc.so.6
#3  0x000076579da97b6e in __gnu_cxx::__verbose_terminate_handler () at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/vterminate.cc:95
#4  0x000076579daacd4c in __cxxabiv1::__terminate (handler=<optimized out>) at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/eh_terminate.cc:48
#5  0x000076579daacdb9 in std::terminate () at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/eh_terminate.cc:58
#6  0x000076579daad04c in __cxxabiv1::__cxa_throw
    (obj=<optimized out>, tinfo=0x5e04ea277ec0 <typeinfo for std::bad_variant_access>, dest=0x5e04ea079b30 <std::bad_variant_access::~bad_variant_access()>)
    at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/eh_throw.cc:98
#7  0x00005e04ea079aed in std::__throw_bad_variant_access (__what=0x5e04e9ecce26 "std::get: wrong index for variant")
    at /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/13.2.1/../../../../include/c++/13.2.1/variant:1349
#8  0x00005e04ea079a60 in std::__throw_bad_variant_access (__valueless=false) at /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/13.2.1/../../../../include/c++/13.2.1/variant:1357
#9  0x00005e04ea0799ff in std::get<0ul, steamcompmgr_xwayland_win_t, steamcompmgr_xdg_win_t> (__v=std::variant [index 1] = {...})
    at /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/13.2.1/../../../../include/c++/13.2.1/variant:1716
#10 0x00005e04ea0799bd in std::get<steamcompmgr_xwayland_win_t, steamcompmgr_xwayland_win_t, steamcompmgr_xdg_win_t> (__v=std::variant [index 1] = {...})
    at /usr/bin/../lib64/gcc/x86_64-pc-linux-gnu/13.2.1/../../../../include/c++/13.2.1/variant:1155
#11 0x00005e04ea073b3c in steamcompmgr_win_t::xwayland (this=0x5e0513d46240) at ../gamescope/src/steamcompmgr_shared.hpp:138
#12 0x00005e04ea04e60a in determine_and_apply_focus () at ../gamescope/src/steamcompmgr.cpp:3595

bad_variant_access strikes again!

(edit: wow, github does not want to pretty parse the commit in the title. It's this de0a8da .

The text was updated successfully, but these errors were encountered:

sharkautarch · 2024-05-07T13:17:03Z

@KyunLFA
Simple workaround is to just check the type member and not execute the block of code accessing xwayland():
https://github.com/sharkautarch/gamescope/tree/fix_wayland_backend_bug

Not sure if this is a full solution tho, since maybe it's possible that before de0a8da, determine_and_apply_focus(xwayland_ctx_t *,std::vector<steamcompmgr_win_t*>&) might have still been applying mouse cursor warp to the root xwayland window even on the wayland backend??? I have no clue honestly

repomansez · 2024-05-07T17:08:17Z

I applied this patch on my machine but I'm still having issues. Not sure if related, though, here's the full strace if you could maybe have a look:

strace.txt

sharkautarch · 2024-05-07T17:47:26Z

Hmmm weird
does it also crash when you run gamescope -- glxgears?

also, try rebuilding gamescope with these meson flags: -Db_sanitize=address,undefined --buildtype=debugoptimized
also if you are building gamescope w/ PKGBUILD, add in this option line: options=(!strip)

Then when you run gamescope, add in these environment variables: ASAN_OPTIONS="detect_leaks=0 halt_on_error=0" LD_PRELOAD=/usr/lib/libasan.so.8.0.0 (so gamescope -- glxgears would become ASAN_OPTIONS="detect_leaks=0 halt_on_error=0" LD_PRELOAD=/usr/lib/libasan.so.8.0.0 gamescope -- glxgears)
That should provide more info

repomansez · 2024-05-07T18:41:23Z

Yes, it does, in fact it segfaults even if I only run gamescope itself.

I'm using Gentoo, compiled and ran with the flags you asked, and here's the full strace output, let me know if there's anything else I can run that may help debugging :)
obs: for some reason I was not able to compile with -Db_sanitize=address,undefined as it gave me ../gamescope-3.14.12/subprojects/wlroots/backend/wayland/seat.c:49:9: error: ‘memcpy’ specified bound 18446744073709551612 exceeds maximum object size 9223372036854775807 [-Werror=stringop-overflow=]

Not sure if that's a problem with my setup or something else.

strace.txt

sharkautarch · 2024-05-07T20:10:16Z

Oh that's just a weird quirk with the sanitizer
Just add these additional meson flags: -Dc_args="-g1 -Wno-error=stringop-overflow -Wno-error=unused-but-set-variable -Wno-error=unused-variable -fno-omit-frame-pointer" -Dc_link_args="-Wno-error=stringop-overflow -g1 -Wno-error=unused-but-set-variable -Wno-error=unused-variable -fno-omit-frame-pointer" -Dcpp_args="-Wno-error=stringop-overflow -g1 -Wno-error=unused-but-set-variable -Wno-error=unused-variable -fno-omit-frame-pointer" -Dcpp_link_args="-Wno-error=stringop-overflow -g1 -Wno-error=unused-but-set-variable -Wno-error=unused-variable -fno-omit-frame-pointer"

Also, when you run gamescope after compiling it with the sanitizer, don't run it with strace, because it might either conflict with address sanitizer or cause gamescope to exit earlier than usual

repomansez · 2024-05-07T20:22:48Z

Ok, was able to build it now, here's the output of gamescope -- glxgears: https://pastebin.com/raw/w1NXDJZD

Just reminding that this is a build built with fix_wayland_backend_bug patch applied

sharkautarch · 2024-05-07T20:38:35Z

#0 0x7d50079c3506  (/usr/lib64/libdecor/plugins-1/libdecor-cairo.so+0x7506)
#1 0x7d50079c402b  (/usr/lib64/libdecor/plugins-1/libdecor-cairo.so+0x802b)
#2 0x7d500a0a880b  (/usr/lib64/libdecor-0.so.0+0x380b)
#3 0x5c943d4809e3  (/usr/bin/gamescope+0x1c5e9e3)
#4 0x5c943d496f7a  (/usr/bin/gamescope+0x1c74f7a)
#5 0x5c943d09ba2c  (/usr/bin/gamescope+0x1879a2c)
#6 0x5c943d1c0d1d  (/usr/bin/gamescope+0x199ed1d)
#7 0x7d5009eff392  (/usr/lib/gcc/x86_64-pc-linux-gnu/13/libstdc++.so.6+0xe1392)
#8 0x7d50095d3340  (/usr/lib64/libc.so.6+0x88340)
#9 0x7d5009645e1b  (/usr/lib64/libc.so.6+0xfae1b)

HUH, it's segfaulting in /usr/lib64/libdecor-0.so.0...
Ok well I'm stumped...
@Joshua-Ashton do you have any clue about how this could be happening??!?!?!?

also, @repomansez I think your gentoo build thingy is stripping debug info from gamescope
Reason why I'm thinking that's the case is because I don't see any line numbers in the sanitizer logs, which iirc happens when binaries are stripped
it looks like you can configure your build script to not strip gamescope w/ the RESTRICT variable: https://devmanual.gentoo.org/ebuild-writing/variables/#ebuild-defined-variables

repomansez · 2024-05-07T21:15:34Z

Well, I manually set it to restrict stripping, but that apparently did not help as it's outputting basically the same thing:

https://pastebin.com/raw/EzzZ1qDk

sharkautarch · 2024-05-07T21:33:49Z

HUH weird...
what does running file $(which gamescope) give you?

repomansez · 2024-05-07T21:39:40Z

/usr/bin/gamescope: executable, regular file, no read permission

sharkautarch · 2024-05-07T21:50:36Z

ummm I guess gentoo uses a drastically different find program than other distros? because here's an example of what's shown on my archlinux system:
usr/bin/gamescope: ELF 64-bit LSB pie executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=2ddd0073a4f7bb5d71dde186fef4f00703cf147e, for GNU/Linux 4.4.0, not stripped

try running objdump --syms $(which gamescope)
if it doesn't just say something like "no symbols", try also running:
objdump --syms $(which gamescope) | grep debug

repomansez · 2024-05-07T22:06:33Z

objdump --syms $(which gamescope) | grep debug

0000000003edf4e0 l 0000000003edf520 l 000000000162aed9 l 0000000002272920 l 0000000002637b80 l 00000000022419c0 g 0000000003f70020 g 0000000003f72e0a g 0000000003f72e0e g 0000000003f729a0 g 0000000003f72e0f g 0000000001a9c950 g 0000000003f72e0c g 0000000003f72ae0 g 0000000003f72e0d g 0000000002241c20 g 0000000003f70100 g 0000000003f72cc0 g 0000000003f72c20 g 0000000003f72b80 g 0000000001b3a600 g 0000000003f72a40 g 0000000003f72e0b g O .bss 0000000000000001 ZL11debugEvents
O .bss 0000000000000001 ZL10debugFocus
F .text 0000000000000024 push_gles2_debug.cold
F .text 000000000000055f debug_callback
O .rodata 0000000000000048 debug_attribs.5
F .text 0000000000000254 push_gles2_debug
O .bss 0000000000000070 _ZN9gamescope12vblank_debugE
O .bss 0000000000000001 __odr_asan.cv_drm_debug_disable_explicit_sync
O .bss 0000000000000001 __odr_asan.cv_drm_debug_disable_degamma_tf
O .bss 0000000000000070 cv_drm_debug_disable_explicit_sync
O .bss 0000000000000001 __odr_asan.cv_drm_debug_disable_shaper_and_3dlut
F .text 0000000000000395 _Z33vulkan_create_debug_blank_texturev
O .bss 0000000000000001 __odr_asan.cv_drm_debug_disable_output_tf
O .bss 0000000000000070 cv_drm_debug_disable_output_tf
O .bss 0000000000000001 __odr_asan.cv_drm_debug_disable_regamma_tf
F .text 00000000000000af pop_gles2_debug
O .bss 0000000000000001 __odr_asan._ZN9gamescope12vblank_debugE
O .bss 0000000000000070 cv_drm_debug_disable_shaper_and_3dlut
O .bss 0000000000000070 cv_drm_debug_disable_degamma_tf
O .bss 0000000000000070 cv_drm_debug_disable_regamma_tf
F .text 00000000000001cd _ZN8LogScope6debugfEPKcz
O .bss 0000000000000070 cv_drm_debug_disable_blend_tf
O .bss 0000000000000001 __odr_asan.cv_drm_debug_disable_blend_tf

kisak-valve · 2024-05-07T22:07:06Z

Hello @sharkautarch, @repomansez, I'd appreciate if you two stopped creating reply chains of consecutive comments. It adds literal pages of redundant text for no benefit.

KyunLFA · 2024-05-08T00:16:46Z

@sharkautarch That workaround fixed it for me, thanks! Will apply to my fork.

KyunLFA · 2024-05-08T00:37:32Z

Oh wow, that way of checking for xwayland vs wayland was also the fix for the bad variant access when using the wine native wayland driver on embedded!! Still various other bugs, but understanding the steamcompmgr code better like that will help me a ton (in hopes of) getting the Wine wayland.drv plumbed into (my fork of) Gamescope for embedded mode, as nested is already working fine. Tysm @sharkautarch !! Maybe submit a PR for your fix?

sharkautarch · 2024-05-08T00:41:21Z

Sure thing!

MarijnS95 · 2024-05-16T16:10:34Z

Running into the above backtrace down from determine_and_apply_focus() as well whenever starting an external Wayland application inside gamescope (via WAYLAND_DISPLAY=gamescope-0 myapp). After either applying the linked PR or reverting the offending commit 10aad9c (and 2d70c9c first because of conflicts) on top of 312e25b, the app starts in gamescope again.

However a different bad_variant_access via paint_all -> MouseCursor::paint shows up as soon as the cursor is moved:

(gdb) bt
#0  0x00007f8e800c5e44 in ??? () at /usr/lib/libc.so.6
#1  0x00007f8e8006da30 in raise () at /usr/lib/libc.so.6
#2  0x00007f8e800554c3 in abort () at /usr/lib/libc.so.6
#3  0x00007f8e80497b0c in __gnu_cxx::__verbose_terminate_handler ()
    at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/vterminate.cc:95
#4  0x00007f8e804adf1a in __cxxabiv1::__terminate (handler=<optimized out>)
    at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/eh_terminate.cc:48
#5  0x00007f8e8049750a in std::terminate () at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/eh_terminate.cc:58
#6  0x00007f8e804ae1d6 in __cxxabiv1::__cxa_throw
    (obj=<optimized out>, tinfo=0x557db191e9d0 <typeinfo for std::bad_variant_access>, dest=0x557db16cea80 <std::bad_variant_access::~bad_variant_access()>) at /usr/src/debug/gcc/gcc/libstdc++-v3/libsupc++/eh_throw.cc:98
#7  0x0000557db16a2054 in std::__throw_bad_variant_access
    (__what=__what@entry=0x557db183ba38 "std::get: wrong index for variant")
    at /usr/include/c++/14.1.1/variant:1349
#8  0x0000557db16a246c in std::__throw_bad_variant_access (__valueless=<optimized out>)
    at /usr/include/c++/14.1.1/variant:1357
#9  0x0000557db16a3b20 in std::get<0ul, steamcompmgr_xwayland_win_t, steamcompmgr_xdg_win_t>
    (__v=std::variant [index 1] = {...}) at /usr/include/c++/14.1.1/variant:1717
#10 std::get<steamcompmgr_xwayland_win_t, steamcompmgr_xwayland_win_t, steamcompmgr_xdg_win_t>
    (__v=std::variant [index 1] = {...}) at /usr/include/c++/14.1.1/variant:1155
#11 steamcompmgr_win_t::xwayland (this=0x557db273fe60) at ../gamescope/src/steamcompmgr_shared.hpp:138
#12 MouseCursor::paint (this=0x7f8e4c1a0460, window=0x557db273fe60, fit=0x0, frameInfo=0x7f8e594cd480)
    at ../gamescope/src/steamcompmgr.cpp:1701
#13 paint_all (async=<optimized out>) at ../gamescope/src/steamcompmgr.cpp:2287
#14 0x0000557db17037be in steamcompmgr_main (argc=<optimized out>, argv=<optimized out>)
    at ../gamescope/src/steamcompmgr.cpp:7554
#15 0x0000557db1704e3f in steamCompMgrThreadRun (argc=8, argv=0x7ffd28aeec98) at ../gamescope/src/main.cpp:933
#16 0x00007f8e804e0c84 in std::execute_native_thread_routine (__p=0x557db28e0880)
    at /usr/src/debug/gcc/gcc/libstdc++-v3/src/c++11/thread.cc:104
#17 0x00007f8e800c3ded in ??? () at /usr/lib/libc.so.6
#18 0x00007f8e801470dc in ??? () at /usr/lib/libc.so.6

KyunLFA changed the title ~~https://github.com/ValveSoftware/gamescope/commit/de0a8daea6003cbb63fa1ec58bb00e2670f9da44 broke the wayland backend~~ commit #de0a8da broke the wayland backend May 7, 2024

KyunLFA changed the title ~~commit #de0a8da broke the wayland backend~~ commit @de0a8da broke the wayland backend May 7, 2024

sharkautarch mentioned this issue May 8, 2024

steamcompmgr.cpp: Guard against bad variant access #1288

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

commit @de0a8da broke the wayland backend #1286

commit @de0a8da broke the wayland backend #1286

KyunLFA commented May 7, 2024 •

edited

sharkautarch commented May 7, 2024

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited by kisak-valve

repomansez commented May 7, 2024

sharkautarch commented May 7, 2024 •

edited

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited by kisak-valve

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited by kisak-valve

repomansez commented May 7, 2024 •

edited by kisak-valve

kisak-valve commented May 7, 2024

KyunLFA commented May 8, 2024

KyunLFA commented May 8, 2024

sharkautarch commented May 8, 2024

MarijnS95 commented May 16, 2024

commit @de0a8da broke the wayland backend #1286

commit @de0a8da broke the wayland backend #1286

Comments

KyunLFA commented May 7, 2024 • edited

sharkautarch commented May 7, 2024

repomansez commented May 7, 2024 • edited by kisak-valve

sharkautarch commented May 7, 2024 • edited by kisak-valve

repomansez commented May 7, 2024

sharkautarch commented May 7, 2024 • edited

repomansez commented May 7, 2024 • edited by kisak-valve

sharkautarch commented May 7, 2024 • edited

repomansez commented May 7, 2024 • edited by kisak-valve

sharkautarch commented May 7, 2024 • edited by kisak-valve

repomansez commented May 7, 2024 • edited by kisak-valve

sharkautarch commented May 7, 2024 • edited by kisak-valve

repomansez commented May 7, 2024 • edited by kisak-valve

objdump --syms $(which gamescope) | grep debug

kisak-valve commented May 7, 2024

KyunLFA commented May 8, 2024

KyunLFA commented May 8, 2024

sharkautarch commented May 8, 2024

MarijnS95 commented May 16, 2024

KyunLFA commented May 7, 2024 •

edited

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited by kisak-valve

repomansez commented May 7, 2024 •

edited by kisak-valve

sharkautarch commented May 7, 2024 •

edited by kisak-valve

repomansez commented May 7, 2024 •

edited by kisak-valve