search

ValveSoftware / Proton

Compatibility tool for Steam Play based on Wine and additional components
Other
25.82k stars 1.13k forks source link

Marvel Rivals (2767030) #8300

Open EnnuiL opened 3 months ago

EnnuiL commented 3 months ago

Compatibility Report

System Information

I confirm:

Proton log: steam-2767030.log

Symptoms

When attempting to launch Marvel Rivals through its launcher, the game will spit this dialog out and then it will return to the launcher: An error dialog titled "Message", with the following text: "DirectX 12 is not supported by your system. Try running without -dx12 or -d3d12 command line argument.". This text is followed by an entire line of solid black.

Reproduction

1) Have a system with an Intel Arc GPU (?) 2) Start the launcher on Steam 3) Dismiss the outdated OS and outdated GPU drivers warning 4) Click on "Launch"

Jonathing commented 3 months ago

I am having this same issue on my Framework Laptop 16 with AMD Radeon 7700S, a card that supports Vulkan, DX12, etc and has been more than capable of running these games (including Overwatch) in the past.

I've tried using Protontricks to include DXVK manually and the 2017 and 2022 VC++ redists, but none have helped so far. I've tried on Proton 9.0, Proton Experimental, and GE-Proton9-20, all with the same result.

FuzzyQuils commented 3 months ago

I was able to get in-game but halfway through a match, I got a GPU hang.

amdgpu 0000:09:00.0: amdgpu: ring gfx_0.0.0 timeout, signaled seq=9668550, emitted seq=9668552
amdgpu 0000:09:00.0: amdgpu: Process information: process GameThread pid 20633 thread vkd3d_queue pid 20733"
<GPU reset snip>
[drm:amdgpu_cs_ioctl [amdgpu]] *ERROR* Failed to initialize parser -125!

Proton Experimental Bleeding-Edge. Radeon RX 5700XT on Mesa 24.3.1 Kernel 6.11, Arch Linux, KDE Wayland.

This is with the game's default settings it picks for a 5700XT on first startup. I'd briefly tested frame generation then turned it off.

I was also using Discord Canary's new streaming features at the same time, however the GPU hang report clearly points to the game causing the hang.

nnyyxxxx commented 3 months ago

this is not an issue with proton

image ![image](https://github.com/user-attachments/assets/0d42656b-9599-4b18-a271-83c3880ae006)
kisak-valve commented 3 months ago

Hello @EnnuiL, try setting the game's launch options to force_vk_vendor=-1 %command% (from https://gitlab.freedesktop.org/mesa/mesa/-/issues/12280).

Note: This is a mesa/ANV specific driver workaround.

EnnuiL commented 3 months ago

Hello @EnnuiL, try setting the game's launch options to force_vk_vendor=-1 %command% (from https://gitlab.freedesktop.org/mesa/mesa/-/issues/12280).

Note: This is a mesa/ANV specific driver workaround.

I tried it just now (and did try the WINE_HIDE_INTEL_GPU=1 environment variable while I was trying to make it run), but I still get the same "DirectX 12 is not supported on your system" error

Jonathing commented 3 months ago

I've resolved this issue by using Winetricks/Protontricks to install vcrun2022 and vkd3d. It might be prudent on either Valve or NetEase to set the proton configuration for this game to use those DLLs.

AMD GPU btw, has nothing to do with the Intel Arc this issue is about, but I figured I'd mention since I also couldn't boot the game with the same issue.

EnnuiL commented 3 months ago

I've resolved this issue by using Winetricks/Protontricks to install vcrun2022 and vkd3d. It might be prudent on either Valve or NetEase to set the proton configuration for this game to use those DLLs.

AMD GPU btw, has nothing to do with the Intel Arc this issue is about, but I figured I'd mention since I also couldn't boot the game with the same issue.

I tried that now and well? Still didn't work, I have the same issue as before; I did also try Proton GE 9-20 as well as Proton 9.0-3 (other than Proton Experimental), and yeah, no dice

WMan22 commented 3 months ago

So I had a similar issue, and I fixed it by adding force_vk_vendor=-1 %command% to the launch options. Apparently, Marvel Rivals does not like intel GPUs, even if it's your iGPU and not the dGPU that actually runs the game. I sent a bug report to mesa and apparently it might be getting resolved.

Edit: Oh, my bad, kisak-valve already mentioned it, sorry.

SpiritWinD86 commented 3 months ago

for players with same error on an old gpu's such as radeon r9 290x\390 - game will run with VKD3D_FEATURE_LEVEL=12_0 %command% in the launch options, proton experimental. (just skip pop-up notices about outdated os and driver -> confirm-continue and launch).

EnnuiL commented 3 months ago

Huh, I have dug further into this; I have finally found a ProtonDB entry from someone else having this issue (and with an Arc card) while on the vkd3d-proton repo, I found https://github.com/HansKristian-Work/vkd3d-proton/issues/2042 as well as mesa#11469, which points out that all Unreal Engine 5 titles are apparently broken on Intel Arc cards on Linux; That's not good!

EnnuiL commented 3 months ago

Attention Intel Arc card users! (or Intel iGPUs, so really, anything that uses the ANV driver and is post-Xe) I have found the fix to this: it turns out ANV didn't support a specific Vulkan extension that was needed in order to emulate Unreal Engine 5 on DX12; by compiling mesa#32525 and running Steam on its devenv, it works flawlessly! There's only one issue: the Mesa workaround needs to be used if you use the compiled Mesa, but once the patch's actually merged? It won't be a problem anymore since the commit that links the workaround to Marvel Rivals is already on upstream

Marvel Rivals' main menu, working on an Intel Arc A380 + Linux system

But yeah, the only thing left is to wait for the Mesa patch to be reviewed and merged (and hopefully released soon alongside the force_vk_vendor=-1 workaround) As a sidenote, the videos are particularly laggy, but so far, it's not a big problem

Eris7090 commented 3 months ago

Hello @EnnuiL, try setting the game's launch options to force_vk_vendor=-1 %command% (from https://gitlab.freedesktop.org/mesa/mesa/-/issues/12280).

Note: This is a mesa/ANV specific driver workaround.

Tried this but sadly didn't work. I'm on fedora 41 with an A750

rejedai commented 3 months ago

Have no media in game, all cutscenes black on rx6700xt

Hondawyyy commented 3 months ago

Hello @EnnuiL, try setting the game's launch options to force_vk_vendor=-1 %command% (from https://gitlab.freedesktop.org/mesa/mesa/-/issues/12280).

Note: This is a mesa/ANV specific driver workaround.

how can i use that? like when i type "force_vk_vendor=-1 %command%" in launch option it says cannot be found

hjpaul7 commented 3 months ago

Tried multiple versions of proton, but I get this error and crash after "Press any button to start":

marvelrivals

ryao commented 3 months ago

Sometimes, when alt-tabbing under Wayland, the game stops registering mouse clicks. Alt-tabbing out of the game and back into the game restores mouse click recognition. I am running Proton Experimental's bleeding_edge branch.

I have not tested it under the Xorg X11 server, although I had issues like this in other games in the past when I used it, so I assume it is a Wine bug.

WMan22 commented 3 months ago

Sometimes, when alt-tabbing under Wayland, the game stops registering mouse clicks. Alt-tabbing out of the game and back into the game restores mouse click recognition. I am running Proton Experimental's bleeding_edge branch.

I have not tested it under the Xorg X11 server, although I had issues like this in other games in the past when I used it, so I assume it is a Wine bug.

Are you on KDE Plasma? I have this problem with Warframe. It usually happens when I click over to another screen then back, usually I can fix it by opening up the steam overlay with shift tab, clicking on it, then closing steam overlay with shift tab again.

gquinche commented 3 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2533433893

Thanks for the report and digging a non trivial solution! Can you clarify for which graphic cards this fix will work? so maybe I don't get my hopes up before time.

I see a mention of 12.5 GFX in the merge requests you linked (but not on this one which seems to be the most probable merge to happen https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32676 ) and also I see you mention post-XE iGPUs, however in wikipedia I don't find any iGPUs after the 12 generation. I suspect you refer to raptor lake gpus which in theory seem are not XE and probably shouldn't be in the 12 gen section, and that both tiger lake and alder lake are not supported. Thanks!

EDIT: From the last comments on the merge I see they mention trying to keep Iris possible in the future with a change in the code (I think this equivalent to XE) that suggest that this currently isn't possible but might be in the future, feel free to correct me if I'm wrong below.

gab462 commented 3 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2533433893

I can confirm that compiling this MR and setting VK_ICD_FILENAMES accordingly fixes the game on my Intel Arc A750. Thanks for the investigative work!

beh-10257 commented 3 months ago

I have a gtx 1650 ti and I am lost on why this doesn't work I am on nixos but I am using flatpak version so IDK here is the log file steam-2767030.log I tried everything from proton 9 to experimental to bleeding edge to ge and tkg so yeah help

PadTrick commented 3 months ago

Replying to #8300 (comment)

I can confirm that compiling this MR and setting VK_ICD_FILENAMES accordingly fixes the game on my Intel Arc A750. Thanks for the investigative work!

archlinux bumped to 24.3.2 but i cant get it to work. still getting this error with all UE5 dx12 games. can you check if it works and maybe provide a guide ?

yofukashino commented 3 months ago

But yeah, the only thing left is to wait for the Mesa patch to be reviewed and merged (and hopefully released soon alongside the force_vk_vendor=-1 workaround) As a sidenote, the videos are particularly laggy, but so far, it's not a big problem

Can confirm as an intel Arc user this fixes dx12 errors but for me, the game still crashes while compiling shaders. Everytime at 40/42 %

EnnuiL commented 3 months ago

Hm, I feel like an elephant on the room that I notice is pretty recurring should be mentioned on this issue: MRivals is really unstable, and within 1 or 2 rounds (or even shortly after joining a match!), it will hard crash;

I have suspected it was the fault of https://github.com/ValveSoftware/steam-for-linux/issues/11446, but by renaming gameoverlayrenderer.so, the overlay is 100% disabled (confirmed by devenv Mesa's logs) and yet it crashes anyway; I then suspected that Fossilize was being confused by the launcher, so I disabled libVkLayer_steam_fossilize.so and as a result, the Fossilize logs have disappeared from Mesa's logs; Still, it still crashes?

I know it isn't an Intel Arc-only issue because well, I am using Mesa with the PR applied as well as force_vk_vendor=-1 being automatically applied by it (I've switched to the one mentioned here!), and yeah, ProtonDB complains about a mid-round crash a lot, with Nvidia and AMD GPUs being involved;

I'm willing to share relevant logs for this, I'll just need to know what would be useful

EDIT: oh god I forgot to mention actually useful information: the crashes that happen are more like a freeze followed by the UE crash reporter (I tried to grab logs from there, but I failed at that); Also this freeze-crash can be reproduced easily (and safely) by creating a custom game with the Spider-Islands map, fill it with hard bots (optional step?), and then play as many rounds as necessary; so far? It quickly freezes during the first minute or two;

A custom game is necessary because if you try this with online matches, the moderation system will consider that you have left in the middle of a round and will add a strike towards you; I did stop playing online due to that!

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2558519502

So, I suspect that once the PR is merged (it's still being reviewed), it would be part of Mesa 25.0.0, so it will take a while until it releases; I could maybe put up a guide in order to build the indev Mesa on Arch and use it on Steam, but I am worried about that, since indev Mesa is indev and can explode on different ways (and a commit did explode here! it got fixed though), plus having to always open Steam through a command in order to avoid the DX12 error prompt will be pretty annoying;

If it's fine anyway to post said guide, I'll put together one and post here

beh-10257 commented 3 months ago

I have a gtx 1650 ti and I am lost on why this doesn't work I am on nixos but I am using flatpak version so IDK here is the log file steam-2767030.log I tried everything from proton 9 to experimental to bleeding edge to ge and tkg so yeah help

I still can't launch the game in the first place IDK if this is a direct x12 problem or what but can anyone look the logs and say if they are even right

ezequielramos commented 3 months ago

Tried multiple versions of proton, but I get this error and crash after "Press any button to start":

marvelrivals

did you find any solution to this? I'm facing the same problem.

beh-10257 commented 3 months ago

image

no was never able

PadTrick commented 3 months ago

image

no was never able

try a different proton version, have similar issue with a game, tried 9.0.4 and was getting this error, went back to proton 8 and game was working.

beh-10257 commented 3 months ago

it doesn't even launch for me in proton 8 I would really appreciate if some developer actually looked at the logs and is there a better way to get support ??

FuzzyQuils commented 3 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2560203329

What maps was this occurring on?

I am currently running an RTX 3060 Ti in my machine but on both this GPU and my previous RX 5700XT, the game had a chance of getting a random GPU hang if the GI settings were set to any of the Lumen variants. (SSGI doesn't seem affected)

In addition both Tokyo 2099 maps were pretty consistent triggers for this crash. (Happens inside the Lumen GI shaders from what myself and one other person troubleshooting was able to figure out)

If this is strictly an Intel hardware thing, there's also a chance ANV isn't agreeing with a shader Marvel Rivals is trying to render with. (Perhaps the same one in this case; GPU hangs reset significantly more reliably without really taking out the entire desktop on Intel Arc GPUs in my experience)

EnnuiL commented 3 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2561249522

It can happen with any map if you play for long enough (I got a freeze at one of the Yggsgard maps for example), but yeah, in my own desperate testing? I have come to the conclusion that both Shin-Shibuya and Spider-Islands are crash-prone, with Spider-Islands in particular being able to cause a crash within the first minute pretty easily (flukes are possible though, and this means it is possible to play a round or two without crashes; it'll inevitably crash though)

I do find interesting that the apparent fix for you was switching out of Lumen, because my GI settings were already on SSGI (Low Quality), and yet the hang keeps happening; I did try the Custom Game + Spider-Islands test once more, and I could confirm that SSGI on High Quality was still affected

I guess a Proton log of the issue would be useful, so here's mine: steam-2767030.log (Proton Experimental, Lowest Settings, AMD FSR 3 enabled (?); Proton GE 9-21 is affected too though)

(oh also yes, the hang only affected the game itself; the desktop did struggle deeply during it, but you can still (barely) interact with it despite the massive lag)

EDIT: Useful information: being on the Defense team on Spider-Islands will consistently crash earlier than being on the Attack team; also I found out that going fullscreen and lowering the resolution to 1280x720 will reduce the crash frequency to the point you can squeeze 20-30 seconds of stable gameplay on the test before It Actually Happens

beh-10257 commented 3 months ago

I can't get this game to run neither in nvk nor in propriatary nvidia drivers in my gtx 1650ti

EnnuiL commented 3 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2561317265

I came to the thrilling conclusion that it's not a Linux issue, but rather one that affects Windows too! Yeah.. looking at some places, it appears that yeah, the crash is something that affects a random portion of everyone, it doesn't discriminate by any specific hardware (a friend of mine had a RX 7900 XTX (Linux) be affected by this issue), it's just the game having a serious issue itself;

I guess this is the part where I give up debugging this any further; Here's the rest of my findings:

yofukashino commented 3 months ago
  • EDIT: Merging mesa#32305 into mesa#32676 will not fix this game! In fact, it will ensure that the ingame shader compilation crashes at 40% consistently (mesa#32676 alone is fine though)

As my game is crashing at 40%, I Tried the mesa#32676 but it still does crash. Error Mentioned Below

l\Binaries\Win64\Marvel-Win64-Shipping.exe: ../src/intel/compiler/brw_reg_type.h:164: brw_reg_type brw_type_larger_of(brw_reg_type, brw_reg_type): Assertion `(a & ~BRW_TYPE_SIZE_MASK) == (b & ~BRW_TYPE_SIZE_MASK)' failed.

Without it, just the DX12 Error

EnnuiL commented 2 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2563238026

Oh, I see the problem: it turns out mesa#32305 isn't the issue, it's a Mesa bug at the point where mesa#32676 is based; rebasing into upstream Mesa's main branch (or just merging it) fixes the 40% crash and makes the game kinda playable, at least until The Dreaded Freeze happens;

After testing mesa#32305 + mesa#32676 + merge of upstream main, I can conclude that mesa#32305 is irrelevant here and that Stalker 2 fixes are not Marvel Rivals fixes despite both using UE5; The only useful PR here that I've seen is mesa#32676, which fixes the "DX12 not supported" prompt UE5 crash that happens on start (DX12 prompt is fixed by indev Mesa 24.3.2+ automatically applying force_vk_vendor=-1)

EnnuiL commented 2 months ago

Welp! It turns out it is very likely that I have reached to a wrong conclusion; I have determined that the Spider-Islands (and friends) crash is one exclusive to Intel GPUs, or at least to the Intel Arc A380; From what I've seen, there are reports of crashes and general instability and they are quite bad, but this specific one with the reproduction steps detailed in https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2561317265? There is a fix:

By using DXVK_FILTER_DEVICE_NAME="AMD" as an env var and using the Ryzen 7900's iGPU, I have successfully played like, 10 rounds without a crash! The performance is quite bad though despite the extreme measures I've used (Gamescope on 600p, minimum settings, etc.), but it's good enough to be playable, and most importantly? It doesn't crash

If someone else has an Intel GPU and can run the patched Mesa, please let me know if a Custom Game with Spider-Islands selected ends crashing or not; I'm considering reporting this to Mesa itself if this is indeed a driver issue

TotalTaxAmount commented 2 months ago

@beh-10257 Did you ever get it to work? I am having the same issue with NVIDIA on NixOS

FuzzyQuils commented 2 months ago

I have a gtx 1650 ti and I am lost on why this doesn't work I am on nixos but I am using flatpak version so IDK here is the log file steam-2767030.log I tried everything from proton 9 to experimental to bleeding edge to ge and tkg so yeah help

From looking at your log, try disabling/uninstalling MangoHUD, it's been acting funky for a lot of people lately.

beh-10257 commented 2 months ago

I have a gtx 1650 ti and I am lost on why this doesn't work I am on nixos but I am using flatpak version so IDK here is the log file steam-2767030.log I tried everything from proton 9 to experimental to bleeding edge to ge and tkg so yeah help

From looking at your log, try disabling/uninstalling MangoHUD, it's been acting funky for a lot of people lately.

didn't help

beh-10257 commented 2 months ago

@beh-10257 Did you ever get it to work? I am having the same issue with NVIDIA on NixOS

no nothing worked I'm hoping a nixpkgs update or an nvidia driver update fixes this tbh it sucked not being able to play with friends but whatever

TotalTaxAmount commented 2 months ago

Switching to GE-Proton8-20 worked for me

NorinB commented 2 months ago

@beh-10257 I had the same issue as you. I also got this Unreal Engine crash after clicking Launch from inside the Launcher. Actually setting the launch option DXVK_FILTER_DEVICE_NAME="AMD" worked for me (thank you for the hint @EnnuiL ♥️). Using GE Proton 9.22 btw.

I'm currently on an Intel 10500K CPU and an AMD RX Vega 64 GPU. Although not being an expert, I believe Proton (or dxvk for that matter) had trouble selecting the correct GPU and tried to use the integrated GPU of my CPU instead of my dedicated one. I guess setting this launch option set Proton (or dxvk) up to use my dedicated GPU instead.

Summing up, I believe you need to find a way to tell Proton (or dxvk) to use your dedicated GPU, right? So maybe, in your case, try using the launch option DXVK_FILTER_DEVICE_NAME="NVIDIA" (not sure, if "NVIDIA" is the correct string for filtering, but if it isn't, it should be something similar.)

You can find out your GPU device name by using vulkaninfo and running it using vulkaninfo | grep deviceName. This prints out all of your available GPUs.

Also found another reference on Reddit, where somelese also mentioned, that you need to explicitly specify the GPU for dxvk in some cases: https://www.reddit.com/r/linux_gaming/comments/v0nbot/warning_for_multi_gpu_users_you_have_to_specify/

I hope this provides you with a new starting point for troubleshooting or maybe even resolves your issue.

kb-1000 commented 2 months ago

I haven't been having any notable issues with crashes - but I HAVE noticed that saving highlights does not seem to work. The progress bar goes through and the GPU is under load, but no file ends up being saved. Ennui tried this as well and could reproduce this issue. I've tried digging into this a bit - and to me it sounds like this issue is caused by Wine's Media Foundation implementation missing an AAC encoder: image I've tried installing mfAACEnc.dll from Windows 7 SP2, but that only made it freeze at 90%. If you try installing the entire Media Foundation alongside it using <workaround removed by moderator>, even the intro video doesn't play... seems like the only way to solve this would be implementing that AAC encoder in Wine.

kisak-valve commented 2 months ago

Hello @kb-1000, the workaround you mentioned is legally problematic and has been removed.

beh-10257 commented 2 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2564607574

ohhh it worked so proton can't do that automatically thanks a lot anyway thanks a lot

ryao commented 2 months ago

Sometimes, when alt-tabbing under Wayland, the game stops registering mouse clicks. Alt-tabbing out of the game and back into the game restores mouse click recognition. I am running Proton Experimental's bleeding_edge branch. I have not tested it under the Xorg X11 server, although I had issues like this in other games in the past when I used it, so I assume it is a Wine bug.

Are you on KDE Plasma? I have this problem with Warframe. It usually happens when I click over to another screen then back, usually I can fix it by opening up the steam overlay with shift tab, clicking on it, then closing steam overlay with shift tab again.

Yes. Thanks for the tip. The next time I have it happen, I will try it.

ryao commented 2 months ago

When in the turret as Punisher (via E), after beginning to fire, the game will often keep playing the firing sound, even if you stop firing.

The game will often freeze during the intro movies when nvidia reflex is enabled.

NVRM: nvAssertFailedNoLog: Assertion failed: pEventNotificationList->pendingEventNotifyCount == 0 @ event_notification.c:289 spams dmesg. The dmesg spam is a race condition according to NVIDIA/open-gpu-kernel-modules#716. A fix is queued for Nvidia has a fix for the 570 driver.

Also, the game sometimes freezes. I suspect it is related to nvidia reflex.

FuzzyQuils commented 2 months ago

Replying to https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2567187742

I was able to get a Proton Log when reproducing the video playback hang with Reflex turned on; the trigger seems to be having the very last frame of the video decoded and displayed, as it always hangs at the end of the video playing in question. (Tutorial videos for character abilities in the menus also make the game hang with Reflex turned on)

The hang seems to be caused by vkLatencySleepNV crashing inside winevulkan/the driver itself:

1140.829:0648:064c:warn:seh:handle_syscall_fault backtrace: --- Exception 0xc0000005 at 0x7f6bc54b4740: /usr/lib/pressure-vessel/overrides/lib/x86_64-linux-gnu/libnvidia-glcore.so.565.77 + 0x2b4740.
1140.830:0648:064c:trace:unwind:dwarf_virtual_unwind function 7f6bc54b4740 base 0x7f6bc54b4736 cie 0x7f6bc6ed6aa8 len 14 id 0 version 1 aug 'zR' code_align 1 data_align -8 retaddr %rip
1140.830:0648:064c:warn:seh:dwarf_virtual_unwind backtrace: 0x7f6bc54b4740: /usr/lib/pressure-vessel/overrides/lib/x86_64-linux-gnu/libnvidia-glcore.so.565.77 + 0x2b4740.
1140.830:0648:064c:trace:unwind:execute_cfa_instructions 7f6bc54b4736: DW_CFA_def_cfa %rsp, 8
1140.830:0648:064c:trace:unwind:execute_cfa_instructions 7f6bc54b4736: DW_CFA_offset %rip, -8
1140.830:0648:064c:trace:unwind:dwarf_virtual_unwind fde 0x7f6bc70b5b70 len 10 personality (nil) lsda (nil) code 7f6bc54b4736-7f6bc54b4740
1140.830:0648:064c:trace:unwind:dwarf_virtual_unwind next function rip=0000000000000000
1140.830:0648:064c:trace:unwind:dwarf_virtual_unwind   rax=0000000000000000 rbx=00007f6bc72cc860 rcx=00007f6be1fa6ced rdx=0000000000000001
1140.830:0648:064c:trace:unwind:dwarf_virtual_unwind   rsi=00000000c020462a rdi=00000000000000f8 rbp=000055556fafb9c0 rsp=00000001000ffb38
1140.830:0648:064c:trace:unwind:dwarf_virtual_unwind    r8=0000000000000000  r9=00000001000feacc r10=1814294de8bf1120 r11=0000000000000246
1140.830:0648:064c:trace:unwind:dwarf_virtual_unwind   r12=0000000000000422 r13=0000555573305780 r14=000055556e1268d0 r15=00005555733057d0
1140.830:0648:064c:warn:seh:dump_syscall_fault backtrace: returning to user mode ip=00006fffeeecf94a ret=c0000005
1140.830:0648:064c:err:msvcrt:_wassert (L"!status && \"vkLatencySleepNV\"",L"../src-wine/dlls/winevulkan/loader_thunks.c",5772)

I also noticed the firing sound thing but the last time I played Punisher was a while ago and may have been while I was still using Reflex. Take it with a grain of salt but if this guy's telling the truth on the Steam Forums, this is potentially a game bug and not easily fixable Proton-side. https://steamcommunity.com/app/2767030/discussions/0/596260472619048276/?ctp=3#c596260472619081188

ryao commented 2 months ago

I was able to get a Proton Log when reproducing the video playback hang with Reflex turned on; the trigger seems to be having the very last frame of the video decoded and displayed, as it always hangs at the end of the video playing in question. (Tutorial videos for character abilities in the menus also make the game hang with Reflex turned on)

That is where it appeared to freeze for me too.

The hang seems to be caused by vkLatencySleepNV crashing inside winevulkan/the driver itself:

Awesome work!

I also noticed the firing sound thing but the last time I played Punisher was a while ago and may have been while I was still using Reflex.

I have had it happen both with and without reflex. I have also once seen it not happen without reflex.

ryao commented 2 months ago

Pressing "VIEW AND SHARE HIGHLIGHTS" and clicking confirm fails to save the recording. It will say: "Unable to Save: Unknown Error".

FuzzyQuils commented 2 months ago

Pressing "VIEW AND SHARE HIGHLIGHTS" and clicking confirm fails to save the recording. It will say: "Unable to Save: Unknown Error".

https://github.com/ValveSoftware/Proton/issues/8300#issuecomment-2564620864 Someone tried to debug this a while ago and came to the conclusion that Wine's Media Foundation implementation is missing an AAC encoder.