While reporting an issue for gui-wm/hyprland-0.22.0 upstream (https://github.com/hyprwm/Hyprland/issues/2016), I noticed that the Crash Report has all its symbols stripped. It's the first time I stumble with something like this, so I'm not really sure which files I need to share to track this down.
The maintainer @vaxerski suggested: "make sure -rdynamic is passed to ld and that there is no symbol stripping process anywhere"
Please see https://wiki.gentoo.org/wiki/Debugging#Per-package.
Thanks Sam as always to point me to the right direction. I tried following the steps: 1. I installed debugedit and gdb 2. I pasted to /etc/portage/env/debugsyms the snippet CFLAGS="${CFLAGS} -ggdb3" CXXFLAGS="${CXXFLAGS} -ggdb3" # nostrip is disabled here because it negates splitdebug FEATURES="${FEATURES} splitdebug compressdebug -nostrip" 3. I wrote to /etc/portage/package.env/hyprland the line gui-wm/hyprland debugsyms. 4. I run emerge --ask --oneshot gui-wm/hyprland I then proceeded to provoke the crash. After inspecting, the report still shows stripped symbols. Is there anything I forgot to do?
You're very welcome :) Could you try commenting out/removing the package.env entry and run: FEATURES="nostrip installsources" CFLAGS="-O2 -ggdb3" CXXFLAGS="-O2 -ggdb3" emerge -v1 gui-wm/hyprland? This bypasses all of the fancy separate debugging info stuff we normally do and keeps it all in the binaries themselves. If that command works, we can transform it into an env file for you then.
I recompiled with that command, but I still get stripped symbols when I provoke a crash
(In reply to ignamartinoli from comment #5) > I recompiled with that command, but I still get stripped symbols when I > provoke a crash Can you run file /usr/bin/hyprland or whatever the path to the binary is, and show me the output?
Created attachment 859830 [details] /usr/bin/Hyprland output
(In reply to ignamartinoli from comment #7) > Created attachment 859830 [details] > /usr/bin/Hyprland output can you show me `file /usr/bin/Hyprland` too? (file is a command too, it's a bit confusing)
I added the output as an attachment
(In reply to ignamartinoli from comment #9) > I added the output as an attachment That's the output of Hyprland, not the file command run on Hyprland
Aaaah sorry, I got confused, here is the output sicro@sicro ~ $ file /usr/bin/Hyprland /usr/bin/Hyprland: ELF 64-bit LSB pie executable, x86-64, version 1 (GNU/Linux), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 3.2.0, with debug_info, not stripped It says that it's not stripped, but today I made it crash again and the Crash Report Backtrace was #0 | Hyprland(+0x113758) [0x556273e65758] ?? ??:0 #1 | Hyprland(+0x57179) [0x556273da9179] ?? ??:0 #2 | /lib64/libc.so.6(+0x38a00) [0x7fb16de73a00] ?? ??:0 #3 | Hyprland(+0x59a83) [0x556273daba83] ?? ??:0 #4 | Hyprland(+0x1281a0) [0x556273e7a1a0] ?? ??:0 #5 | Hyprland(+0x90d9a) [0x556273de2d9a] ?? ??:0 #6 | Hyprland(+0x9b32c) [0x556273ded32c] ?? ??:0 #7 | /usr/lib64/libwayland-server.so.0(wl_signal_emit_mutable+0x7c) [0x7fb16e773b1c] ?? ??:0 #8 | Hyprland(+0x147a58) [0x556273e99a58] ?? ??:0 #9 | /usr/lib64/libdrm.so.2(drmHandleEvent+0xf7) [0x7fb16e74fc17] ?? ??:0 #10 | Hyprland(+0x149a69) [0x556273e9ba69] ?? ??:0 #11 | /usr/lib64/libwayland-server.so.0(wl_event_loop_dispatch+0xd2) [0x7fb16e775b22] ?? ??:0 #12 | /usr/lib64/libwayland-server.so.0(wl_display_run+0x25) [0x7fb16e773555] ?? ??:0 #13 | Hyprland(+0x487c6) [0x556273d9a7c6] ?? ??:0 #14 | /lib64/libc.so.6(+0x238ca) [0x7fb16de5e8ca] ?? ??:0 #15 | /lib64/libc.so.6(__libc_start_main+0x85) [0x7fb16de5e985] ?? ??:0 #16 | Hyprland(+0x4bed1) [0x556273d9ded1] ?? ??:0
pfff. I feel borderline stumped. I think it might be easier if we try debug this on IRC interactively?
Alright, never used IRC but I can try. Can you send me the link?
Note that the latest version of Hyprland is 0.24.1. I updated the ebuild in GURU recently, but it hasn't been merged in master yet, so only 0.22.0 is available for now. I'm sure upstream will ask you to reproduce on the latest release so it might be pertinent try this bug again on 0.24.1 once it's available.
The latest version 0.24.1 is now published in GURU, so you can update to it
If you can reproduce the bug on 0.24.1, can you try to put the following in /etc/portage/env/hyprland.conf: CFLAGS="${CFLAGS} -ggdb3" CXXFLAGS="${CXXFLAGS} -ggdb3" LDFLAGS="${LDFLAGS} -rdynamic" FEATURES="-splitdebug -compressdebug nostrip" You may have to create the `env` directory if it doesn't exist Once that file is created put the following in /etc/portage/package.env: gui-wm/hyprland hyprland.conf After that, rebuild hyprland, and (hopefully) your stack trace should contain symbols
The upgrade throws an error. I checked myself the build log, the error is provoked because there is no /usr/share/protocols/hyprland-global-shortcuts-v1.xml file. This is quite similar to a bug I previously filled (https://bugs.gentoo.org/891799). I would like to create a PR and fix this. Could you guide me so that in the future I can be more useful to the distro? (I have zero experience writing patches) The error is in the line 35 of protocols/mason.build (https://github.com/hyprwm/Hyprland/blob/main/protocols/meson.build#L35). It should be ['/usr/share', 'protocols/hyprland-global-shortcuts-v1.xml'] By working over the patch for the version 0.21.0 (https://github.com/gentoo-mirror/guru/blob/master/gui-wm/hyprland/files/0.21.0-meson.patch), I came up with this: diff --git a/protocols/meson.build b/protocols/meson.build index 9d47e0a..d9fc8a6 100644 --- a/protocols/meson.build +++ b/protocols/meson.build @@ -29,7 +29,7 @@ protocols = [ ['pointer-constraints-unstable-v1.xml'], ['tablet-unstable-v2.xml'], ['idle.xml'], - [hl_protocol_dir, 'protocols/hyprland-toplevel-export-v1.xml'] + ['/usr/share', 'protocols/hyprland-toplevel-export-v1.xml'] # - [hl_protocol_dir, 'protocols/hyprland-global-shortcuts-v1.xml'] # + ['/usr/share', 'protocols/hyprland-global-shortcuts-v1.xml'] ] wl_protos_src = [] wl_protos_headers = [] By adding the two lines that begin with #, would it be enough? Or I should change the numbers after @@ and the index too?
I did a little bit of homework this time ahahah :D
Nevermind, I found the manual entry (https://wiki.gentoo.org/wiki/Creating_a_patch). This is the patch diff --git a/protocols/meson.build b/protocols/meson.build index 458de86..294d9ee 100644 --- a/protocols/meson.build +++ b/protocols/meson.build @@ -31,8 +31,8 @@ protocols = [ ['pointer-constraints-unstable-v1.xml'], ['tablet-unstable-v2.xml'], ['idle.xml'], - [hl_protocol_dir, 'protocols/hyprland-toplevel-export-v1.xml'], - [hl_protocol_dir, 'protocols/hyprland-global-shortcuts-v1.xml'] + ['/usr/share', 'protocols/hyprland-toplevel-export-v1.xml'], + ['/usr/share', 'protocols/hyprland-global-shortcuts-v1.xml'] ] wl_protos_src = [] wl_protos_headers = [] I'm going to test it and if it works I'm going to open a PR.
Ah yes, I remember that bug quite well, it was annoying to fix :D It seems upstream added a new file that behaves similarly to the previous one, and the bug filed upstream is still not resolved (https://github.com/hyprwm/Hyprland/issues/1207) Unfortunately I did not catch this change, and I suspect it means many more people will have issues with the upgrade The patch you are proposing would work. The proper way to make a patch file is to clone the Hyprland repository, checkout the tag for version in question (v0.24.1), modify the meson.build file as you propose, and use the `git diff` command to generate a patch, eg. `git diff > meson.patch` However, there is another modification that needs to be done in the ebuild (see line 84, it needs to be done also for the hyprland-global-shortcuts-v1.xml) Regarding the PR, just to be sure you are talking about GURU and not Hyprland? Because this patch is only needed for Gentoo, so it doesn't need to be sent upstream. Also, this is out of scope for this bug, so feel free to open a new one.
Alright, here is the bug link https://bugs.gentoo.org/904299. Let's fix that and then we'll come back here.
I'm back! To recap (a terminal snippet is worth 1000 words): sicro@sicro ~ $ cat /etc/portage/env/hyprland.conf CFLAGS="${CFLAGS} -ggdb3" CXXFLAGS="${CXXFLAGS} -ggdb3" LDFLAGS="${LDFLAGS} -rdynamic" FEATURES="-splitdebug -compressdebug nostrip" sicro@sicro ~ $ cat /etc/portage/package.env/hyprland gui-wm/hyprland hyprland.conf sicro@sicro ~ $ emerge -1 gui-wm/hyprland [ebuild R ] gui-wm/hyprland-0.24.1-r1 sicro@sicro ~ $ emerge -p dev-util/debugedit sys-devel/gdb [ebuild R ] sys-devel/gdb-13.1.90_p20230429 [ebuild R ] dev-util/debugedit-5.0-r1 sicro@sicro ~ $ cat .config/hyprland/hyprland.conf monitor = eDP-1, preferred, auto, 1 monitor = HDMI-A-1, preferred, auto, 1, mirror, eDP-1 I inserted only useful information from the outputs. I rebooted and after inserting the HDMI cable to another monitor the crash occurred. As shown here the crash report got its symbols stripped: sicro@sicro ~ $ cat .cache/hyprland/hyprlandCrashReport9738.txt Backtrace: #0 | Hyprland(_ZN13CrashReporter18createAndSaveCrashEi+0xa6f) [0x5562596183df] ?? ??:0 #1 | Hyprland(_Z25handleUnrecoverableSignali+0x3c) [0x55625953f02c] ?? ??:0 #2 | /lib64/libc.so.6(+0x38ba0) [0x7f2236710ba0] ?? ??:0 #3 | Hyprland(_ZN11CCompositor16getWorkspaceByIDERKi+0x13) [0x556259541fb3] ?? ??:0 #4 | Hyprland(_ZN17CHyprMasterLayout18recalculateMonitorERKi+0x40) [0x556259631580] ?? ??:0 #5 | Hyprland(_ZN13CHyprRenderer13renderMonitorEP8CMonitor+0x769) [0x5562595cf8a9] ?? ??:0 #6 | Hyprland(_Z13handleWrappedP11wl_listenerPv+0x3c) [0x556259584b9c] ?? ??:0 #7 | /usr/lib64/libwayland-server.so.0(wl_signal_emit_mutable+0x7c) [0x7f2237186b1c] ?? ??:0 #8 | Hyprland(+0x1e17a8) [0x55625965b7a8] ?? ??:0 #9 | /usr/lib64/libdrm.so.2(drmHandleEvent+0xf7) [0x7f2237162c17] ?? ??:0 #10 | Hyprland(handle_drm_event+0x4c) [0x55625965dc2c] ?? ??:0 #11 | /usr/lib64/libwayland-server.so.0(wl_event_loop_dispatch+0xd2) [0x7f2237188b22] ?? ??:0 #12 | /usr/lib64/libwayland-server.so.0(wl_display_run+0x25) [0x7f2237186555] ?? ??:0 #13 | Hyprland(main+0x428) [0x55625952ca28] ?? ??:0 #14 | /lib64/libc.so.6(+0x238ca) [0x7f22366fb8ca] ?? ??:0 #15 | /lib64/libc.so.6(__libc_start_main+0x85) [0x7f22366fb985] ?? ??:0 #16 | Hyprland(_start+0x21) [0x5562595328c1] ?? ??:0 Does anybody know if there is something I am missing here?
(In reply to ignamartinoli from comment #22) > I'm back! > > To recap (a terminal snippet is worth 1000 words): > > > sicro@sicro ~ $ cat /etc/portage/env/hyprland.conf > CFLAGS="${CFLAGS} -ggdb3" > CXXFLAGS="${CXXFLAGS} -ggdb3" > LDFLAGS="${LDFLAGS} -rdynamic" > FEATURES="-splitdebug -compressdebug nostrip" > > sicro@sicro ~ $ cat /etc/portage/package.env/hyprland > gui-wm/hyprland hyprland.conf > > sicro@sicro ~ $ emerge -1 gui-wm/hyprland > > [ebuild R ] gui-wm/hyprland-0.24.1-r1 > > sicro@sicro ~ $ emerge -p dev-util/debugedit sys-devel/gdb > > [ebuild R ] sys-devel/gdb-13.1.90_p20230429 > [ebuild R ] dev-util/debugedit-5.0-r1 > > sicro@sicro ~ $ cat .config/hyprland/hyprland.conf > monitor = eDP-1, preferred, auto, 1 > monitor = HDMI-A-1, preferred, auto, 1, mirror, eDP-1 > > > I inserted only useful information from the outputs. > > I rebooted and after inserting the HDMI cable to another monitor the crash > occurred. > > As shown here the crash report got its symbols stripped: > > > sicro@sicro ~ $ cat .cache/hyprland/hyprlandCrashReport9738.txt > Backtrace: > #0 | Hyprland(_ZN13CrashReporter18createAndSaveCrashEi+0xa6f) > [0x5562596183df] > ?? > ??:0 > #1 | Hyprland(_Z25handleUnrecoverableSignali+0x3c) [0x55625953f02c] > ?? > ??:0 > #2 | /lib64/libc.so.6(+0x38ba0) [0x7f2236710ba0] > ?? > ??:0 > #3 | Hyprland(_ZN11CCompositor16getWorkspaceByIDERKi+0x13) [0x556259541fb3] > ?? > ??:0 > #4 | Hyprland(_ZN17CHyprMasterLayout18recalculateMonitorERKi+0x40) > [0x556259631580] > ?? > ??:0 > #5 | Hyprland(_ZN13CHyprRenderer13renderMonitorEP8CMonitor+0x769) > [0x5562595cf8a9] > ?? > ??:0 > #6 | Hyprland(_Z13handleWrappedP11wl_listenerPv+0x3c) [0x556259584b9c] > ?? > ??:0 > #7 | /usr/lib64/libwayland-server.so.0(wl_signal_emit_mutable+0x7c) > [0x7f2237186b1c] > ?? > ??:0 > #8 | Hyprland(+0x1e17a8) [0x55625965b7a8] > ?? > ??:0 > #9 | /usr/lib64/libdrm.so.2(drmHandleEvent+0xf7) [0x7f2237162c17] > ?? > ??:0 > #10 | Hyprland(handle_drm_event+0x4c) [0x55625965dc2c] > ?? > ??:0 > #11 | /usr/lib64/libwayland-server.so.0(wl_event_loop_dispatch+0xd2) > [0x7f2237188b22] > ?? > ??:0 > #12 | /usr/lib64/libwayland-server.so.0(wl_display_run+0x25) > [0x7f2237186555] > ?? > ??:0 > #13 | Hyprland(main+0x428) [0x55625952ca28] > ?? > ??:0 > #14 | /lib64/libc.so.6(+0x238ca) [0x7f22366fb8ca] > ?? > ??:0 > #15 | /lib64/libc.so.6(__libc_start_main+0x85) [0x7f22366fb985] > ?? > ??:0 > #16 | Hyprland(_start+0x21) [0x5562595328c1] > ?? > ??:0 > > > Does anybody know if there is something I am missing here? After looking at this for a bit I think the above output is correct. The symbols are included, just mangled and offset. The question marks are from addr2line where Gentoo's default pie comes in. If you disable pie the crash reporter can resolve the addr2line bits as well. And here is an example unstripped, LDFLAGS="-no-pie" and debugging info included so CXXFLAGS="-gdwarf". Backtrace: #0 | ./build/src/Hyprland(_ZN13CrashReporter18createAndSaveCrashEi+0x75e) [0x4ed54e] _ZN13CrashReporter18createAndSaveCrashEi /home/ask/sources/Hyprland/build/../src/debug/CrashReporter.cpp:86 #1 | ./build/src/Hyprland(_Z25handleUnrecoverableSignali+0x3c) [0x5b4f1c] _Z25handleUnrecoverableSignali /home/ask/sources/Hyprland/build/../src/Compositor.cpp:35 #2 | /lib64/libc.so.6(+0x38fa0) [0x7fa4a7b82fa0] ?? ??:0 #3 | /lib64/libc.so.6(+0x87f9c) [0x7fa4a7bd1f9c] ?? ??:0 #4 | /lib64/libc.so.6(gsignal+0x12) [0x7fa4a7b82ef2] ?? ??:0 #5 | /lib64/libc.so.6(abort+0xd3) [0x7fa4a7b6c4ad] ?? ??:0 #6 | /usr/lib/gcc/x86_64-pc-linux-gnu/13/libstdc++.so.6(+0xa7ca3) [0x7fa4a7ea7ca3] ?? ??:0 #7 | /usr/lib/gcc/x86_64-pc-linux-gnu/13/libstdc++.so.6(+0xdd1b6) [0x7fa4a7edd1b6] ?? ??:0 #8 | /usr/lib/gcc/x86_64-pc-linux-gnu/13/libstdc++.so.6(+0xdd221) [0x7fa4a7edd221] ?? ??:0 #9 | /usr/lib/gcc/x86_64-pc-linux-gnu/13/libstdc++.so.6(+0xdd483) [0x7fa4a7edd483] ?? ??:0 #10 | ./build/src/Hyprland() [0x473842] _ZN11CCompositor10initServerEv /home/ask/sources/Hyprland/build/../src/Compositor.cpp:125 (discriminator 2) #11 | ./build/src/Hyprland(main+0xa70) [0x48e9b0] main /home/ask/sources/Hyprland/build/../src/main.cpp:100 #12 | /lib64/libc.so.6(+0x2398a) [0x7fa4a7b6d98a] ?? ??:0 #13 | /lib64/libc.so.6(__libc_start_main+0x85) [0x7fa4a7b6da45] ?? ??:0 #14 | ./build/src/Hyprland(_start+0x21) [0x491261] _start ??:? Disabling pie isn't optimal so i've played around with it a bit more. Internally in the code they use the following. $ addr2line -e build/src/Hyprland -f 0x564c73fb258e ?? ??:0 This doesn't mean that the information is unobtainable with pie enabled, just execinfo doesn't give the information in a convenient way. $ addr2line -e build/src/Hyprland -f _ZN13CrashReporter18createAndSaveCrashEi+0x75e _ZN13CrashReporter18createAndSaveCrashEi /home/ask/sources/Hyprland/build/../src/debug/CrashReporter.cpp:86 Im still looking at this if this could be improved upstream, I can see that NixOS bug reports have the same issues.
we do not support this cersion of the package anymore. Does this still happen? If not, can we close this?
No, this issue is still present on recent Hyprland versions
Honestly, I'm not really sure this is a bug per se. It's not great but the builtin crash reporters for _anything_ tend to be subpar compared to a coredump with gdb/lldb or perhaps elfutils.
Ok, then we'll close this as CANTFIX
