I can confirm this issue exists with non-flatpak Steam as well.
I can confirm that building a newer release with 142d72f17d9e15c7d63e2f38c1f200dc77992fd0 reverted resolves the issue for me.
I haven't been able to replicate it when using Steam to launch the game (it seems like the EAC runtime resolves the issue) but running a standalone EAC game (like Star Citizen) outside of Steam triggers the issue.
Another data point: Star Citizen (4.10.2, EAC/EOS) hits the same thing, and the logs show the parent lookup directly. This confirms what kagadar reported, with process-level evidence.
Setup: Fedora 44, kernel 7.2.10, Mesa 26.2.4, RX 7900 XTX. Non-Steam launch (RSI Launcher through Proton, StarCitizen_Launcher.exe is the parent of StarCitizen.exe). Renders through Vulkan, but the game process still loads vkd3d-proton's d3d12.dll/d3d12core.dll. Same prefix and game build for every run.
Results
| Build | Result |
|---|---|
| GE-Proton11-6 | works |
| proton-cachyos 20260703 | works |
| GE-Proton11-7 | fails (first bad build) |
GE-Proton11-9, with and without PROTON_NATIVE_AGS=1 |
fails |
| GE-Proton11-10 | fails |
| proton-cachyos 20261005 | fails |
| Proton Experimental (experimental-11.0-20261014) | fails |
On failing builds EAC reports Violation detected: 8 - Untrusted system file (…/StarCitizen_Launcher.exe) about 3 s after start, then error 30033.
Process log comparison (WINEDEBUG=err+all,+timestamp,+pid,+tid,+seh,+loaddll,+process,+server)
| GE-Proton11-6 (works) | GE-Proton11-9 (fails) | |
|---|---|---|
| Launcher | pid 094c | pid 0988 |
| Game | pid 0984 | pid 09b4 |
| Game loads d3d12 / d3d12core | yes / yes | yes / yes |
| Game opens its parent launcher | no | yes: open_process( pid=0988, access=0x1000 ) from the thread that loaded d3d12core |
EAC bootstrapper syscall workaround lines / sigsys_handler events |
2 / 53 | 2 / 53 |
| EAC result | Login completed |
Violation detected: 8, then 30033 |
So the game process does not touch its parent on the working build, and on the failing build it opens it with PROCESS_QUERY_LIMITED_INFORMATION. The EAC syscall handling and the AMD AGS path (PROTON_NATIVE_AGS=1) are not the difference.
Caveats: this is a correlation. I have not yet swapped the pre-142d72f vkd3d-proton DLLs into a failing build. WINEDLLOVERRIDES=d3d12=d does not disable vkd3d-proton here, so that test was inconclusive.
Why it matters here: non-Steam launchers (RSI Launcher, others) hand EAC a parent process it doesn't expect, whereas a Steam launch gives it Steam, which is probably why kagadar can't reproduce it through Steam. If the detection could skip the parent lookup for non-UE titles, make it lazy, or avoid opening the process, that would probably fix both Hunt and Star Citizen. Related reports: GloriousEggroll/proton-ge-custom#858 and mactan-sc/rsilauncher#25.
Happy to run any test build or patch here and supply the full +server logs on request (~290 MB each).
GE-Proton reverted 142d72f as a patch: https://github.com/GloriousEggroll/proton-ge-custom/commit/07925856e670bbd6ab09f6c4417d5a596ff69ca7. It fixes Star Citizen (non-Steam launch) in our tests once released. Reporting back after testing.
Update: GE-Proton11-11 includes a revert of 142d72f as a patch (https://github.com/GloriousEggroll/proton-ge-custom/commit/07925856e670bbd6ab09f6c4417d5a596ff69ca7), and I confirmed it fixes Star Citizen (non-Steam launch via RSI Launcher): no more Violation detected: 8 - Untrusted system file (StarCitizen_Launcher.exe), login completes.
Proton Experimental and proton-cachyos still carry 142d72f and still fail for me (experimental-11.0-20261014, proton-cachyos 20261005). Would it be possible to take the same change (don't read the parent process's image for the engine version check) or an alternative that avoids mapping the parent image?
Earlier analysis with logs: GloriousEggroll/proton-ge-custom#858.
This also affects EAC in Helldivers 2, that was added in the latest patch (20th October)
proton experimentalx4 2026-10ge-proton11-11x1 2026-10ge-proton11-10x1 2026-10ge-proton11-6x1 2026-10ge-proton11-7x1 2026-10ge-proton11-9x1 2026-10PROTON_NATIVE_AGS=1x1 2026-10PROTON_NATIVE_AGS=1`)x1 2026-10WINEDEBUG=err+all,+timestamp,+pid,+tid,+seh,+loaddll,+process,+server`)x1 2026-10WINEDLLOVERRIDES=d3d12=dx1 2026-10d3d12.dllx1 2026-10d3d12core.dllx1 2026-10
For full context, I originally reported this on Proton's issue tracker before going ham with bisects: https://github.com/ValveSoftware/Proton/issues/3898#issuecomment-5937897818
Software information
Hunt: Showdown 1896, with clean prefix and USER dir.
Running in Steam Flatpak on Fedora 44 with niri and ntfsplus used as NTFS driver. I know my config is freaky, but I swear I tested both with and without VKShade and extest both before and after the parent proc merge commit.
System information
NVIDIA-SMI 615.71.09 KMD Version: 615.71.09 CUDA UMD Version: 13.4Log files
Logs taken with official Proton Experimental, not the self-built version, with a run with no pre-existing prefix or USER dir.
steam-594650.log
Game.log