Native linux autoupdate appears to be broken on recent Canary

[EzioisAwesome56]

Recently the native fixes for linux auto updater appears to have stopped working correctly. I'm running Void linux glibc that is reasonably up to date with all its packages.

When canary puts out an update which would require downloading a new tarball, it will download the update and try extracting it. then seemingly something GPU related crashes and the entire thing ends with a bus error. the order of the GPU crashes vary sometimes but it usually ends up looking something like this, which was logged with an upgrade from 0.0.754 to 0.0.755, but has been happening nearly daily for a couple of weeks or so:

splashScreen.webContentsSend: SPLASH_UPDATE_STATE SPLASH_UPDATE_STATE [
  {
    status: 'downloading-updates',
    current: 1,
    total: 1,
    progress: 100
  }
]
14:27:40.891 › DiscordSplash.onStateUpdate: {"status":"downloading-updates","current":1,"total":1,"progress":100}
14:27:40.892 › Splash.onStateUpdate: {"status":"downloading-updates","current":1,"total":1,"progress":100}
14:27:41.471 › Splash.updateCountdownSeconds: undefined
[INFO] [nativeFixes/host] Extracting DiscordCanary
14:27:42.525 › Splash.updateCountdownSeconds: undefined
[INFO] [nativeFixes/host] Extracting libvk_swiftshader.so
[INFO] [nativeFixes/host] Extracting chrome_100_percent.pak
[INFO] [nativeFixes/host] Extracting discord-canary.desktop
[INFO] [nativeFixes/host] Extracting libGLESv2.so
[INFO] [nativeFixes/host] Extracting vk_swiftshader_icd.json
[INFO] [nativeFixes/host] Extracting discord.png
[INFO] [nativeFixes/host] Extracting chrome_200_percent.pak
[INFO] [nativeFixes/host] Extracting resources/bootstrap/manifest.json
[INFO] [nativeFixes/host] Extracting resources/build_info.json
[INFO] [nativeFixes/host] Extracting resources/app.asar
[11341:0911/142742.727748:ERROR:gpu_process_host.cc(953)] GPU process exited unexpectedly: exit_code=11
blackbox: 9/11/2025, 2:27:42 PM EDT 5 ❌ child-process-gone { type: 'GPU', reason: 'crashed', exitCode: 11, serviceName: 'GPU' }
[INFO] [nativeFixes/host] Extracting locales/ar.pak
[INFO] [nativeFixes/host] Extracting locales/en-GB.pak
[INFO] [nativeFixes/host] Extracting locales/hr.pak
[INFO] [nativeFixes/host] Extracting locales/sl.pak
[INFO] [nativeFixes/host] Extracting locales/gu.pak
[INFO] [nativeFixes/host] Extracting locales/af.pak
[INFO] [nativeFixes/host] Extracting locales/tr.pak
[INFO] [nativeFixes/host] Extracting locales/pt-BR.pak
[INFO] [nativeFixes/host] Extracting locales/et.pak
[INFO] [nativeFixes/host] Extracting locales/cs.pak
[INFO] [nativeFixes/host] Extracting locales/hi.pak
[INFO] [nativeFixes/host] Extracting locales/am.pak
[INFO] [nativeFixes/host] Extracting locales/fil.pak
[INFO] [nativeFixes/host] Extracting locales/zh-CN.pak
[INFO] [nativeFixes/host] Extracting locales/he.pak
[INFO] [nativeFixes/host] Extracting locales/it.pak
[INFO] [nativeFixes/host] Extracting locales/da.pak
[INFO] [nativeFixes/host] Extracting locales/en-US.pak
[INFO] [nativeFixes/host] Extracting locales/zh-TW.pak
[INFO] [nativeFixes/host] Extracting locales/th.pak
[INFO] [nativeFixes/host] Extracting locales/hu.pak
[INFO] [nativeFixes/host] Extracting locales/ro.pak
[INFO] [nativeFixes/host] Extracting locales/pl.pak
[INFO] [nativeFixes/host] Extracting locales/bn.pak
[INFO] [nativeFixes/host] Extracting locales/el.pak
[INFO] [nativeFixes/host] Extracting locales/sk.pak
[INFO] [nativeFixes/host] Extracting locales/kn.pak
[INFO] [nativeFixes/host] Extracting locales/ru.pak
[INFO] [nativeFixes/host] Extracting locales/nb.pak
[INFO] [nativeFixes/host] Extracting locales/nl.pak
[INFO] [nativeFixes/host] Extracting locales/es-419.pak
[INFO] [nativeFixes/host] Extracting locales/mr.pak
[INFO] [nativeFixes/host] Extracting locales/ml.pak
[INFO] [nativeFixes/host] Extracting locales/fr.pak
[INFO] [nativeFixes/host] Extracting locales/sr.pak
[INFO] [nativeFixes/host] Extracting locales/ca.pak
[INFO] [nativeFixes/host] Extracting locales/lv.pak
[INFO] [nativeFixes/host] Extracting locales/lt.pak
[INFO] [nativeFixes/host] Extracting locales/ko.pak
[INFO] [nativeFixes/host] Extracting locales/sw.pak
[INFO] [nativeFixes/host] Extracting locales/sv.pak
[INFO] [nativeFixes/host] Extracting locales/pt-PT.pak
[INFO] [nativeFixes/host] Extracting locales/ms.pak
[INFO] [nativeFixes/host] Extracting locales/bg.pak
[INFO] [nativeFixes/host] Extracting locales/uk.pak
[INFO] [nativeFixes/host] Extracting locales/vi.pak
[INFO] [nativeFixes/host] Extracting locales/fa.pak
[INFO] [nativeFixes/host] Extracting locales/es.pak
[INFO] [nativeFixes/host] Extracting locales/te.pak
[INFO] [nativeFixes/host] Extracting locales/ja.pak
[INFO] [nativeFixes/host] Extracting locales/ur.pak
[INFO] [nativeFixes/host] Extracting locales/fi.pak
[INFO] [nativeFixes/host] Extracting locales/ta.pak
[INFO] [nativeFixes/host] Extracting locales/id.pak
[INFO] [nativeFixes/host] Extracting locales/de.pak
[INFO] [nativeFixes/host] Extracting chrome-sandbox
[INFO] [nativeFixes/host] Extracting libffmpeg.so
[INFO] [nativeFixes/host] Extracting libEGL.so
[INFO] [nativeFixes/host] Extracting libvulkan.so.1
[INFO] [nativeFixes/host] Extracting icudtl.dat
[11341:0911/142742.805989:ERROR:gpu_process_host.cc(953)] GPU process exited unexpectedly: exit_code=7
Bus error

trying to start it afterwards only produces errors, but they are different for some reason, such as the following:

[0911/143403.366926:ERROR:icu_util.cc(229)] Couldn't mmap icu data file
Trace/breakpoint trap

another example i got after reinstalling 0.0.754 and then patching & updating:

[ERROR] [injector] Failed to inject: RangeError: Internal error. Icu error.
    at String.localeCompare (<anonymous>)
    at /home/alyssa/.config/moonlight-mod/dist/injector.js:1:9050
    at Array.sort (<anonymous>)
    at re (/home/alyssa/.config/moonlight-mod/dist/injector.js:1:9033)
    at async Object.Ye (/home/alyssa/.config/moonlight-mod/dist/injector.js:1:15189)
DiscordCanary 0.0.755
(electron) 'session.getPreloads' is deprecated and will be removed. Please use 'session.getPreloadScripts' instead.
(electron) 'session.setPreloads' is deprecated and will be removed. Please use 'session.registerPreloadScript' instead.
Starting app.
Removing old directory  0.0.754
Starting updater.

<--- Last few GCs --->

[15914:0x201400130000]      221 ms: Scavenge 12.6 (17.9) -> 10.0 (26.7) MB, pooled: 0 MB, 0.97 / 0.00 ms  (average mu = 1.000, current mu = 1.000) allocation failure; 
[15914:0x201400130000]      319 ms: Mark-Compact 19.1 (29.4) -> 12.4 (30.7) MB, pooled: 0 MB, 2.71 / 0.01 ms  (+ 0.1 ms in 0 steps since start of marking, biggest step 0.0 ms, walltime since start of marking 25 ms) (average mu = 0.991, current mu = 0.991)

<--- JS stacktrace --->

[15914:0911/143531.753593:ERROR:node_bindings.cc(180)] OOM error in V8: DateTimePatternGeneratorCache::CreateGenerator Allocation failed - process out of memory
Trace/breakpoint trap

even unpatching the updated copy will not fix it, it always results in a completely broken discord install that needs a complete reload using the provided tarballs

these logs where with moonlight stable 1.3.28, installed via moonlight-cli

[Cynosphere]

Considering there's already an underlying issue on Linux that started cropping up within the 30 versions of Electron where relaunching always causes a segfault, I doubt there's anything we can do to work around this on our end.

[Cynosphere]

The GPU part I can only speculate as possibly being caused by Vulkan or VAAPI options in Native Fixes, but there's a lack of information on if those are even enabled

[EzioisAwesome56]

sorry, I dunno why I didn't think of mentioning that

"Enable VAAPI features on Linux" is enabled, but the vulkan renderer is not
HEVC support on linux is also enabled

[Cynosphere]

It does make me wonder if there is something we can do to make extraction not hang the main thread somehow to where the splash video is causing the GPU process to timeout or whatever its doing.

I know that persistence (repatching on update) also hangs the main thread on Windows, causing the splash video to also stop playing, but it doesn't cause any issues with the GPU process in that case (surprisingly).

[EzioisAwesome56]

Just as an aside, I turned off the two options I mentioned in my previous post, and the auto updater still crashed, with similar sounding GPU errors, as shown below (tested again with 0.0.754 -> 0.0.755)

[INFO] [nativeFixes/host] Extracting chrome_200_percent.pak
[INFO] [nativeFixes/host] Extracting resources/bootstrap/manifest.json
[INFO] [nativeFixes/host] Extracting resources/build_info.json
[INFO] [nativeFixes/host] Extracting resources/app.asar
[16818:0911/151219.400253:ERROR:gpu_process_host.cc(953)] GPU process exited unexpectedly: exit_code=7
blackbox: 9/11/2025, 3:12:19 PM EDT 5 ❌ child-process-gone { type: 'GPU', reason: 'crashed', exitCode: 7, serviceName: 'GPU' }
<snip>
[INFO] [nativeFixes/host] Extracting icudtl.dat
[16818:0911/151219.487168:ERROR:gpu_process_host.cc(953)] GPU process exited unexpectedly: exit_code=11
Bus error                  ./DiscordCanary

which results in

[alyssa@alyssa-pc DiscordCanary]$ ./DiscordCanary 
Segmentation fault         ./DiscordCanary

[EzioisAwesome56]

just wanted to drop in and say as of recently discord did...something and now the auto updater works again without creating corrupted installs

edit: nevermind