MagicBeanDev/Longshot

Watch

Fork 74

Game: nightly-80 crashes on startup (0xc0000005) #261

New issue

Open

opened 2026-11-13 18:14:04 +00:00 by gabrielsemke · 3 comments

gabrielsemke commented

2026-11-13 18:14:04 +00:00

Longshot version

nightly-80-7315957c

Channel

Nightly

What kind of problem

Crash, freeze or black screen

What happened

Launch Longshot from the launcher or run bin\longshot.exe directly. The game window opens and closes within 1–2 seconds, before the title screen appears.

The game log reports 0xc0000005 on sead::MainThread at longshot.exe+0x15b53346 while writing to 0x100000000000. Windows Error Reporting confirms the same exception code and offset. Nightly nightly-77-9f64464b also failed with the same code and offset.

Where in the game

During startup, before reaching the title screen.

How often

Every time

Graphics card and driver version

AMD Radeon RX 9070 XT, AMD driver 26.11.1 (LLPC), Vulkan 1.4

CPU, RAM and operating system

Intel Core i7-12700, 32 GB RAM, Windows 11 Pro (build 22631)

Changes from the defaults

The crash occurs with both the default native mesh shaders and RX_VK_MESH_SHADERS=compute in longshot_env.txt. It also reproduces with MSI Afterburner and RivaTuner closed. Windows Error Reporting confirmed that neither RTSS DLL was loaded in the attached runs.

A separate launch requesting the Intel UHD Graphics 770 with RX_VK_DEVICE=1 reproduced the same exception before the log confirmed which GPU was active.

longshot_log.txt

Screenshots or video

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-80-7315957c ### Channel Nightly ### What kind of problem Crash, freeze or black screen ### What happened Launch Longshot from the launcher or run `bin\longshot.exe` directly. The game window opens and closes within 1–2 seconds, before the title screen appears. The game log reports `0xc0000005` on `sead::MainThread` at `longshot.exe+0x15b53346` while writing to `0x100000000000`. Windows Error Reporting confirms the same exception code and offset. Nightly `nightly-77-9f64464b` also failed with the same code and offset. ### Where in the game During startup, before reaching the title screen. ### How often Every time ### Graphics card and driver version AMD Radeon RX 9070 XT, AMD driver 26.11.1 (LLPC), Vulkan 1.4 ### CPU, RAM and operating system Intel Core i7-12700, 32 GB RAM, Windows 11 Pro (build 22631) ### Changes from the defaults The crash occurs with both the default native mesh shaders and `RX_VK_MESH_SHADERS=compute` in `longshot_env.txt`. It also reproduces with MSI Afterburner and RivaTuner closed. Windows Error Reporting confirmed that neither RTSS DLL was loaded in the attached runs. A separate launch requesting the Intel UHD Graphics 770 with `RX_VK_DEVICE=1` reproduced the same exception before the log confirmed which GPU was active. ### longshot_log.txt . ### Screenshots or video . ### 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.

longshot-reporte-de-cierre.zip

4.4 KiB

gabrielsemke added the

kind

bug

label

2026-11-13 18:14:04 +00:00

monsterlag commented

2026-11-13 19:39:28 +00:00

Triage: S0 (repeatable startup crash), reported on nightly-77 and nightly-80; affected reach is not established. This is not yet evidence of a Vulkan device-lost or an AMD-driver fault.

One source-level clue in the issue text: the reported write address, 0x100000000000, is exactly Longshot's native guest-memory base (A64_HOST_BASE, third_party/a64-kit/rt/a64.h:61; the Windows reservation uses it in host/hos_main.c:147). If the faulting instruction is a lifted guest-memory access, that address would correspond to guest offset zero. That is a hypothesis consistent with a guest null-pointer write, not a confirmed root cause or a reason to commit a speculative fix.

The next useful evidence is symbolication of longshot.exe+0x15b53346 for the exact nightly-80 executable, plus the guest PC/registers immediately before the write. A maintainer who can read the attached diagnostic ZIP may already have those lines. Please keep binaries, generated code and dumps local; text diagnostics are enough.

I confirmed that the issue has an attached ZIP through the issue API, but my attachment-download route returned HTTP403, so I stopped it and have not read the archive. I have not reproduced this crash on the local NVIDIA setup. The Intel-requested launch also does not establish an Intel reproduction until the selected-device line is known.

Source reference:

#define A64_HOST_BASE 0x100000000000ULL /* 16 TB: far from anything Windows or glibc map by default */

Triage: S0 (repeatable startup crash), reported on nightly-77 and nightly-80; affected reach is not established. This is not yet evidence of a Vulkan device-lost or an AMD-driver fault. One source-level clue in the issue text: the reported write address, `0x100000000000`, is exactly Longshot's native guest-memory base (`A64_HOST_BASE`, `third_party/a64-kit/rt/a64.h:61`; the Windows reservation uses it in `host/hos_main.c:147`). If the faulting instruction is a lifted guest-memory access, that address would correspond to guest offset zero. That is a hypothesis consistent with a guest null-pointer write, not a confirmed root cause or a reason to commit a speculative fix. The next useful evidence is symbolication of `longshot.exe+0x15b53346` for the exact nightly-80 executable, plus the guest PC/registers immediately before the write. A maintainer who can read the attached diagnostic ZIP may already have those lines. Please keep binaries, generated code and dumps local; text diagnostics are enough. I confirmed that the issue has an attached ZIP through the issue API, but my attachment-download route returned HTTP403, so I stopped it and have not read the archive. I have not reproduced this crash on the local NVIDIA setup. The Intel-requested launch also does not establish an Intel reproduction until the selected-device line is known. Source reference: https://git.magicbean.dev/MagicBeanDev/Longshot/src/commit/d906c8f4276652b51c5e30dc5e938648c9e8d42d/third_party/a64-kit/rt/a64.h#L61

gabrielsemke commented

2026-11-13 19:50:17 +00:00

Author

Thank you for looking into this issue.

I've attached the longshot_log.txt file directly, since you mentioned that the ZIP download returned an HTTP 403 error.

I've also reattached the diagnostic ZIP.

Please let me know if you need any additional information or specific tests. I'm happy to help.

Thank you for looking into this issue. I've attached the longshot_log.txt file directly, since you mentioned that the ZIP download returned an HTTP 403 error. I've also reattached the diagnostic ZIP. Please let me know if you need any additional information or specific tests. I'm happy to help.

longshot_log.3.txt

6.2 KiB

longshot_log.4.txt

6.3 KiB

longshot_env.txt

26 B

longshot_log.1.txt

4.5 KiB

longshot_log.2.txt

4.5 KiB

gabrielsemke commented

2026-11-13 20:05:46 +00:00

Author

I managed to capture a Windows crash dump and extract the CPU context at the exact point of failure.

The dump confirms:

Exception: 0xC0000005 (write access violation)
Faulting module: longshot.exe
Faulting offset: 0x15B53346
Write address: 0x100000000000

CPU registers:

RIP = 0x7FF7713E3346
RDI = 0x0000000000000000
RSI = 0x0000100000000000
R8  = 0x0000000000000000

Faulting instruction:

mov byte ptr [rdi+rsi], r8b

This confirms that the instruction attempts to write to the guest-memory base address with a zero offset.

I understand this doesn't establish the underlying cause. Would these CPU registers help with the investigation, or do you also need the guest PC/registers from a special debug build?

I can provide additional diagnostic information if needed.

I managed to capture a Windows crash dump and extract the CPU context at the exact point of failure. The dump confirms: - Exception: 0xC0000005 (write access violation) - Faulting module: longshot.exe - Faulting offset: 0x15B53346 - Write address: 0x100000000000 CPU registers: ``` RIP = 0x7FF7713E3346 RDI = 0x0000000000000000 RSI = 0x0000100000000000 R8 = 0x0000000000000000 ``` Faulting instruction: ```asm mov byte ptr [rdi+rsi], r8b ``` This confirms that the instruction attempts to write to the guest-memory base address with a zero offset. I understand this doesn't establish the underlying cause. Would these CPU registers help with the investigation, or do you also need the guest PC/registers from a special debug build? I can provide additional diagnostic information if needed.