Game: Longshot closed unexpectedly while you were playing #1239

Closed
opened 2026-11-16 21:33:37 +00:00 by jsc0 · 6 comments

Longshot version

nightly-189-aa95d35f (source 719431ac)

Channel

Nightly

What kind of problem

Crash, freeze or black screen

What happened

Longshot closed unexpectedly while you were playing. The graphics driver failed. Update your graphics driver, and turn off overlays and tools hooked into the game. (The report from the crash window is pasted below.)

Where in the game

kokiri forest walking up to the girl on top of the building in the middle of the village

How often

Once so far

Graphics card and driver version

AMD Radeon RX 6800 XT, AMD proprietary driver 26.10.1 (AMD proprietary shader compiler)

CPU, RAM and operating system

AMD Ryzen 7 5800X3D 8-Core Processor (16 threads), 32 GB, Windows 11 (build 26300)

Changes from the defaults

No response

longshot_log.txt

log attached

Screenshots or video

No response

Before you send it

  • The problem is still there on the newest release or nightly.
  • I searched the open issues for it.
  • I'm not attaching game files, keys, the game's executable, its .bin caches or saves.
### Longshot version nightly-189-aa95d35f (source 719431ac) ### Channel Nightly ### What kind of problem Crash, freeze or black screen ### What happened Longshot closed unexpectedly while you were playing. The graphics driver failed. Update your graphics driver, and turn off overlays and tools hooked into the game. (The report from the crash window is pasted below.) ### Where in the game kokiri forest walking up to the girl on top of the building in the middle of the village ### How often Once so far ### Graphics card and driver version AMD Radeon RX 6800 XT, AMD proprietary driver 26.10.1 (AMD proprietary shader compiler) ### CPU, RAM and operating system AMD Ryzen 7 5800X3D 8-Core Processor (16 threads), 32 GB, Windows 11 (build 26300) ### Changes from the defaults _No response_ ### longshot_log.txt log attached ### Screenshots or video _No response_ ### Before you send it - [x] The problem is still there on the newest release or nightly. - [x] I searched the open issues for it. - [x] I'm not attaching game files, keys, the game's executable, its .bin caches or saves.
Member

Possibly related, for context:

  • #1040 AMD Radeon 860M (APU): Kokiri Forest VK_ERROR_DEVICE_LOST without RX_VK_ASTC_CPU=1 & 4096 MB budget thrashing with it — Related Kokiri Forest crash on AMD Windows, but different GPU and identified device-loss/ASTC behavior.
  • #365 [RX 6600] Game crashes with GPU timeout — Related AMD GPU crash during gameplay; its reported GPU-timeout symptom and location differ.
Possibly related, for context: - #1040 AMD Radeon 860M (APU): Kokiri Forest VK_ERROR_DEVICE_LOST without RX_VK_ASTC_CPU=1 & 4096 MB budget thrashing with it — Related Kokiri Forest crash on AMD Windows, but different GPU and identified device-loss/ASTC behavior. - #365 [RX 6600] Game crashes with GPU timeout — Related AMD GPU crash during gameplay; its reported GPU-timeout symptom and location differ.
Contributor

I'm taking this one; a fix PR will follow. Same crash reproduced on an RX 6800 XT (Navi 21) on Windows, rc.7, twice: Kokiri Forest, with the camera on Saria on her stump, a few minutes in.

The cause, narrowed down one run at a time:

  • RX_VK_CHECKPOINTS=1: the GPU hangs in the draw of program 89ca750f9852b7ca. It has a mesh stage and no task stage.
  • The RX 6000 rule on Windows (task_mesh_fallback_default in gpu/vk/mesh_mode.h) only sends task + mesh programs to the compute fallback (uses_mesh_fallback, gpu/vk/vk_internal.h). So this mesh-only program still runs on the driver's native mesh path, the one Hookshot issue 996 found hanging on RDNA 2.
  • RX_VK_SKIP_PROG=89ca750f9852b7ca: no hang in 6.5 minutes there, but the foliage's leaf cards are missing.
  • RX_VK_MESH_SHADERS=compute (every mesh program on the fallback): 20 minutes at Saria, no crash, foliage drawn.

Workaround until the fix lands: RX_VK_MESH_SHADERS=compute.

The fix: on Navi 21-23 on Windows, mesh-only programs take the fallback too; RX_VK_TASK_MESH_FALLBACK=0 still turns the whole rule off. Probably also #1237 (RX 6750 XT, Navi 22, Kokiri).

I'm taking this one; a fix PR will follow. Same crash reproduced on an RX 6800 XT (Navi 21) on Windows, rc.7, twice: Kokiri Forest, with the camera on Saria on her stump, a few minutes in. The cause, narrowed down one run at a time: - `RX_VK_CHECKPOINTS=1`: the GPU hangs in the draw of program `89ca750f9852b7ca`. It has a mesh stage and no task stage. - The RX 6000 rule on Windows (`task_mesh_fallback_default` in gpu/vk/mesh_mode.h) only sends task + mesh programs to the compute fallback (`uses_mesh_fallback`, gpu/vk/vk_internal.h). So this mesh-only program still runs on the driver's native mesh path, the one Hookshot issue 996 found hanging on RDNA 2. - `RX_VK_SKIP_PROG=89ca750f9852b7ca`: no hang in 6.5 minutes there, but the foliage's leaf cards are missing. - `RX_VK_MESH_SHADERS=compute` (every mesh program on the fallback): 20 minutes at Saria, no crash, foliage drawn. Workaround until the fix lands: `RX_VK_MESH_SHADERS=compute`. The fix: on Navi 21-23 on Windows, mesh-only programs take the fallback too; `RX_VK_TASK_MESH_FALLBACK=0` still turns the whole rule off. Probably also #1237 (RX 6750 XT, Navi 22, Kokiri).
Member

@GrimVisage Do you have a branch or PR for the RX 6800 XT mesh-only fallback fix that we can test? We'd like to help with an independent check before the stable cut. We can review the fallback and cache paths and help coordinate a Windows RX 6800 XT retest on the exact patched build, using default mesh settings at Saria's stump in Kokiri Forest. If you've already run that retest, could you link the build and result here?

@GrimVisage Do you have a branch or PR for the RX 6800 XT mesh-only fallback fix that we can test? We'd like to help with an independent check before the stable cut. We can review the fallback and cache paths and help coordinate a Windows RX 6800 XT retest on the exact patched build, using default mesh settings at Saria's stump in Kokiri Forest. If you've already run that retest, could you link the build and result here?
Contributor

@futondestroyer Thanks, and thanks for the credit in #1650. I had the same change on a local branch (mesh-only programs added to the Windows Navi 21–23 rule, with a routing test that fails on main), but I hadn't pushed it before #1650 landed, so there's nothing separate to review. #1650 covers it.

The retest on the patched build hasn't been run yet. The 20 minutes at Saria with the foliage intact was on rc.7 with RX_VK_MESH_SHADERS=compute forced, not on a build with the fix.

I'm running it on the RX 6800 XT (Windows 11, AMD proprietary driver) on v3.0.0 from the launcher, with no RX_ switches and default mesh settings:

  • 20 minutes facing Saria on her stump from a fresh program cache: device loss yes/no, foliage drawn, and the startup log's fallback line (vk: mesh-only and task + mesh programs use the compute fallback ...)
  • 5 minutes at the same spot with a warm cache: fps, frame time and VRAM from the log

I'll post the result and log lines here and on #1650 when it's done.

@futondestroyer Thanks, and thanks for the credit in #1650. I had the same change on a local branch (mesh-only programs added to the Windows Navi 21–23 rule, with a routing test that fails on main), but I hadn't pushed it before #1650 landed, so there's nothing separate to review. #1650 covers it. The retest on the patched build hasn't been run yet. The 20 minutes at Saria with the foliage intact was on rc.7 with `RX_VK_MESH_SHADERS=compute` forced, not on a build with the fix. I'm running it on the RX 6800 XT (Windows 11, AMD proprietary driver) on v3.0.0 from the launcher, with no `RX_` switches and default mesh settings: - 20 minutes facing Saria on her stump from a fresh program cache: device loss yes/no, foliage drawn, and the startup log's fallback line (`vk: mesh-only and task + mesh programs use the compute fallback ...`) - 5 minutes at the same spot with a warm cache: fps, frame time and VRAM from the log I'll post the result and log lines here and on #1650 when it's done.
Contributor

Retested on v3.0.0, which has the fix (#1650). On an RX 6800 XT, AMD driver 26.10.1, Windows 11 build 26300, with default mesh settings and no RX_ switches: two runs of about 10 minutes each, facing Saria on her stump. No device loss and no hang, and the foliage drew. The startup log shows vk: mesh-only and task + mesh programs use the compute fallback. Details, numbers and logs: #1650 (comment).

That PC isn't the one in this report (it has a Ryzen 5 5600X and 16 GB). @jsc0, could you check whether it still happens for you on v3.0.0? Closing is the team's call.

Retested on v3.0.0, which has the fix (#1650). On an RX 6800 XT, AMD driver 26.10.1, Windows 11 build 26300, with default mesh settings and no `RX_` switches: two runs of about 10 minutes each, facing Saria on her stump. No device loss and no hang, and the foliage drew. The startup log shows `vk: mesh-only and task + mesh programs use the compute fallback`. Details, numbers and logs: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/1650#issuecomment-52658. That PC isn't the one in this report (it has a Ryzen 5 5600X and 16 GB). @jsc0, could you check whether it still happens for you on v3.0.0? Closing is the team's call.
Owner

Closing this as fixed: #1650 is merged and addresses it. If that's wrong, or it still happens on a current build, feel free to re-open it.

Closing this as fixed: [#1650](https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/1650) is merged and addresses it. If that's wrong, or it still happens on a current build, feel free to re-open it.
Sign in to join this conversation.
No milestone
No project
No assignees
5 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot#1239
No description provided.