Segfault when starting Factorio after some time

[gerion0]

I tested Moonshine with Factorio (playing well locally). It always crashes after some time with a segmentation fault.

It experienced two scenarios:

• The game loads and crashes exactly when trying to switch to the game menu.
• The game loads, reaches the game menu but crashes some time (20-30s) within the game menu. The picture is streamed well, the mouse is much to slow.

I tried with debug logging activated. This seems to be not helpful:

2026-08-18T11:36:20.302135Z DEBUG pixelforge::converter: ColorConverter::convert() took 949.788µs
2026-08-18T11:36:20.302147Z DEBUG pixelforge::encoder::h265::record: h265 submit frame 4868: idr=false, num_refs=1, cur_slot=0
2026-08-18T11:36:20.311647Z DEBUG pixelforge::encoder::resources: Encoded frame: offset=0, size=178503
2026-08-18T11:36:20.318833Z DEBUG pixelforge::converter: ColorConverter::convert() took 977.28µs
2026-08-18T11:36:20.318850Z DEBUG pixelforge::encoder::h265::record: h265 submit frame 4869: idr=false, num_refs=1, cur_slot=1
2026-08-18T11:36:20.327946Z DEBUG pixelforge::encoder::resources: Encoded frame: offset=0, size=176742
2026-08-18T11:36:20.335508Z DEBUG pixelforge::converter: ColorConverter::convert() took 983.272µs
2026-08-18T11:36:20.335522Z DEBUG pixelforge::encoder::h265::record: h265 submit frame 4870: idr=false, num_refs=1, cur_slot=0
2026-08-18T11:36:20.344853Z DEBUG pixelforge::encoder::resources: Encoded frame: offset=0, size=167004
2026-08-18T11:36:20.352199Z DEBUG pixelforge::converter: ColorConverter::convert() took 1.003289ms
2026-08-18T11:36:20.352216Z DEBUG pixelforge::encoder::h265::record: h265 submit frame 4871: idr=false, num_refs=1, cur_slot=1
2026-08-18T11:36:20.356987Z DEBUG pixelforge::encoder::resources: Encoded frame: offset=0, size=183271
2026-08-18T11:36:20.418798Z DEBUG pixelforge::converter: ColorConverter::convert() took 940.18µs
2026-08-18T11:36:20.418818Z DEBUG pixelforge::encoder::h265::record: h265 submit frame 4872: idr=false, num_refs=1, cur_slot=0
2026-08-18T11:36:20.435904Z DEBUG pixelforge::encoder::resources: Encoded frame: offset=0, size=5969517
zsh: segmentation fault (core dumped)  MOONSHINE_LOG=debug moonshine

Running via gdb gives this backtrace (Factorio):

#0  0x00007ffff7a329a0 in __memmove_avx_unaligned_erms () from /usr/lib64/libc.so.6
#1  0x000055555647e385 in pixelforge::encoder::resources::wait_and_read_bitstream ()
#2  0x000055555646d8d4 in pixelforge::encoder::pipeline::run_completion_thread ()
#3  0x000055555646fc7a in std::sys::backtrace::__rust_begin_short_backtrace ()
#4  0x0000555556466cf8 in core::ops::function::FnOnce::call_once{{vtable.shim}} ()
#5  0x00005555568e7d5f in alloc::boxed::{impl#31}::call_once<(), (dyn core::ops::function::FnOnce<(), Output=()> + core::marker::Send), alloc::alloc::Global> () at library/alloc/src/boxed.rs:2249
#6  std::sys::thread::unix::{impl#2}::new::thread_start () at library/std/src/sys/thread/unix.rs:118
#7  0x00007ffff796355f in start_thread () from /usr/lib64/libc.so.6
#8  0x00007ffff79e25ac in __clone3 () from /usr/lib64/libc.so.6

I'm using version 0.15.0, compiled by myself (Rust version 1.96.1).

[hgaiser]

Could you check if c2da458 fixes your issue?

1. Buffer overflows in the encoder were not detected, so on large frames it could result in a segmentation fault.
2. Scaled the max allowed buffer based on the resolution, similar to what was done in AV1 encoding. Not a huge fan of this, since frames this large shouldn't even happen in the first place. I'll look into this later maybe 🤷

[gerion0]

Could you check if c2da458 fixes your issue?

Thanks for the fix! Yes and no. I compiled the recent HEAD and the app wasn't even loading anymore at first (but moonshine remained alive). I traced it back to this error:

Thread 29 "PulseMainloop" received signal SIGSEGV, Segmentation fault.

So, when switching back to the PipeWire driver within the game (provoking #177 again), it loads now fine.
In my testing regarding sound with pulseaudio yesterday, the app was crashing moonshine after reaching the game menu (but with the sound playing fine), with this fix, it never reaches even the loading window, so this fix seems to cause the pulse issue.

I also noticed that the mouse pointer (the game is controlled by mouse+keyboard) is much slower in the streaming window than on other windows or the desktop of the client (I'm using moonlight-qt on a Linux PC as a client). Is this expected?

1. Buffer overflows in the encoder were not detected, so on large frames it could result in a segmentation fault.
2. Scaled the max allowed buffer based on the resolution, similar to what was done in AV1 encoding. Not a huge fan of this, since frames this large shouldn't even happen in the first place. I'll look into this later maybe 🤷

So the buffer is a multitude of the resolution? I was testing with multiple resolutions (e.g. 4k) and moonshine crashed with all.

[hgaiser]

Did you every find out what the issue was?