Repository navigation
Open
Description
What were you doing when it happened?
Diagnostics: please drag crash-2026-10-31_12-58-28.zip (the launcher showed it in Finder) onto this issue: it has the full session log and macOS's crash reports.
Bloodborne for Mac 0.5.12-42-g78c368d-dirty: crash report
- Mac: Mac17,7 (Apple M5 Max), 36 GB, macOS 27.0.1 (Build 26A434)
- Settings: output 2560x1440, upscaler off, cap 120 FPS, variable timing, windowed
- Session: started 2026-10-31 12:58:28 -05:00, ran 55 min 55 s
- Exit status: 76
- Cause: the game stopped responding (no new frame for 30 seconds, 237299 frames before; BB_HANG_SECONDS) (at 13:54:23.576)
- Log: game-2026-10-31_12-58-28.log, 2137 lines: 1 critical, 261 errors, 29 warnings
Warnings and errors before the cause (5 most recent)
13:15:03.947 WARN GPU [Render.Vulkan] <Warning> emit_spirv.cpp:459 operator(): Float32 denorm flushing is not supported by the GPU
13:16:41.691 WARN == Stall during rendering at flush 79825
13:17:52.373 WARN GPU [Render.Vulkan] <Warning> image_view.cpp:44 LogOnce: image view format cannot view the texture in Metal (block or depth reinterpretation): using the image's format
13:19:50.495 WARN == Stall during rendering at flush 92854
13:27:00.728 ERROR GPU [Render.Vulkan] <Error> liverpool_to_vk.cpp:801 SurfaceFormat: Unknown data_format=15 and num_format=4: using a stand-in
game-2026-10-31_12-58-28.log (12 lines around the cause)
13:54:23.576 INFO EzWorkPool_TypeB1 running; 24a41b6<2efb9c<21892<1f590bd<2032405<20340c9<263b246<host_start 0xa9<_pthread_start 0x73<thread_start 0xf
13:54:23.576 INFO EzWorkPool_TypeA3 blocked in condvar (0x71091007b0) for 286508 ms; __psynch_cvwait 0xa<cond_wait 0xcc<2083ecd<207e9e8<208bfe5<20340ab<263b246<host_start 0xa9<_pthread_start 0x73<thread_start 0xf
13:54:23.576 INFO EzWorkPool_TypeA2 blocked in condvar (0x7109100610) for 286505 ms; __psynch_cvwait 0xa<cond_wait 0xcc<2083ecd<207e9e8<208bfe5<20340ab<263b246<host_start 0xa9<_pthread_start 0x73<thread_start 0xf
13:54:23.576 INFO EzWorkPool_TypeA1 blocked in condvar (0x7109100470) for 286508 ms; __psynch_cvwait 0xa<cond_wait 0xcc<2083ecd<207e9e8<208bfe5<20340ab<263b246<host_start 0xa9<_pthread_start 0x73<thread_start 0xf
13:54:23.576 INFO EzWorkPool_TypeA0 blocked in condvar (0x71091002d0) for 286508 ms; __psynch_cvwait 0xa<cond_wait 0xcc<2083ecd<207e9e8<208bfe5<20340ab<263b246<host_start 0xa9<_pthread_start 0x73<thread_start 0xf
13:54:23.576 INFO FD4JobWorker:0 blocked in semaphore (0x13) for 3343834 ms; __psynch_cvwait 0xa<sem_wait 0x247<263ae6a<106747f<263b246<host_start 0xa9<_pthread_start 0x73<thread_start 0xf
13:54:23.576 INFO Core.Logging.BackgroundStrategy blocked in semaphore (0x12) for 3343849 ms; __psynch_cvwait 0xa<sem_wait 0x247<263ae6a<26471c5<263b246<host_start 0xa9<_pthread_start 0x73<thread_start 0xf
13:54:23.576 INFO main blocked in condvar (0x71093dacf0) for 30468 ms; __psynch_cvwait 0xa<cond_wait 0xcc<2083ecd<207e9e8<1f5c2ae<1ed87c4<1d92108<103d438<103d2c1<1067579
13:54:23.576 CRIT Crash: the game stopped responding (no new frame for 30 seconds, 237299 frames before; BB_HANG_SECONDS)
13:54:23.576 INFO bbgpu-host: exiting: SIGTERM
13:54:23.576 INFO Exit: pipeline archive flushed in 0 ms
13:54:23.900 ERROR launcher: the game stopped with status 76 after 55 min 55 s
Metadata
Metadata
Assignees
Labels
No labels
Activity
noelmiller commentedon Oct 31, 2026
What I was doing: about 56 minutes of Hunter's Dream co-op (three sessions as host or guest, the last with two other players). The freeze came 7 seconds after the last session ended normally, while the game was reloading my world.
Timeline from the log:
One thread is stuck in a loop. In the hang report, the only game thread not waiting is
EzWorkPool_TypeB1:In the 1.09 image,
0x24a41b6is the end of a linked-list walk:It exits only when it gets back to the list head in
r12. Still running after 30 s means the nodes most likely form a loop that doesn't pass through the head, i.e. the list is corrupted. B2 waits on a mutex, called from the same function as B1 (0x1f58f30vs B1's0x1f590bd), so B1 most likely holds it. Main and the render threads wait on the workers.Ruled out for this session:
Render guard: absent shader ...lines, so none of the Crash: the game crashed (signal 11 at guest offset 0x2169edc) #99/Crash: the game crashed (signal 11 at guest offset 0x216a08c) #113 guards skipped anything.I can't see from the log what corrupted the list. I only have this one sample, and the report only takes one stack per thread. It comes at the same moment as #99 (the reload after a co-op session). That may be a coincidence; I'm only noting it.
The diagnostics zip is attached. Build: 0.5.12 plus my local online-play branch (PR #78 merged on top). "dirty" is only the
fsr-vulkansubmodule.Written with Claude Code.
Merge issue Wooloomooloo2#118 hang diagnostics into dev
evill33t commentedon Oct 31, 2026
Opened diagnostics draft #119. The local 1.09 executable confirms the reported list walk. The surrounding code appears to be the UI's ActionScript collector, but one stack sample cannot prove a persistent cycle or identify the write that damaged the list. A shared cause with #99/#113 remains unproven.
For the next reproduction, please:
Hang report: snapshotblocks, worker registers and anyList #118:lines, not only the short crash excerpt. If the original archive is still available, please attach that too; no archive is currently visible here.No extra sampling flag is required. Leave
BB_HANG_SECONDSenabled at its normal threshold. The added diagnostics only read memory during hang reporting; they record up to 64 list nodes and distinguish an observed cycle, reaching the head, unreadable memory and an inconclusive node limit. This is evidence gathering; the underlying hang remains unresolved.Wooloomooloo2 commentedon Nov 5, 2026
@noelmiller @evill33t a possible lead for this one: in the #127 hangs the game's 256 KiB per-frame heap slowly fills over 1.5 to 2 hours of play, because the game's occlusion queries never get a result on the port (FINDINGS 2026-11-05). Once it overflows, records go missing and the game stops. If you hit this again, check whether the log has "Frame pool: frame heap … full" lines in the minutes before the freeze. Co-op might just be where your session happened to cross that line.
Capture bounded worker-list evidence for issue Wooloomooloo2#118 hangs