Segmentation fault in `scheduler.run(.wait)`

[edman]

Hello 👋

I'm getting a segmentation fault in `scheduler.run(.wait).

Here's the stacktrace:

Segmentation fault at address 0x7f1d0fc5fe20
.../.cache/zig/p/aio-0.0.0-776t3qVVBQCS5e6tiQ2qs6ds8N3NaEvbxqLg6womBPI_/src/coro/Scheduler.zig:85:24: 0x11f067f in tick (zigproject)
            next = node.next;
                       ^
.../.cache/zig/p/aio-0.0.0-776t3qVVBQCS5e6tiQ2qs6ds8N3NaEvbxqLg6womBPI_/src/coro/Scheduler.zig:111:30: 0x11e52b1 in run (zigproject)
            if (try self.tick(.blocking) == 0) break;
                             ^
.../zigproject/src/main.zig:282:22: 0x11e4a02 in main (zigproject)
    try scheduler.run(.wait);
                     ^
/nix/store/hmipmal12akswjhv9wmlkk13hsm3nf7w-zig-0.14.0/lib/zig/std/start.zig:656:37: 0x11cf5ea in posixCallMainAndExit (zigproject)
            const result = root.main() catch |err| {
                                    ^
/nix/store/hmipmal12akswjhv9wmlkk13hsm3nf7w-zig-0.14.0/lib/zig/std/start.zig:271:5: 0x11cf19d in _start (zigproject)
    asm volatile (switch (native_arch) {
    ^
???:?:?: 0x0 in ??? (???)

It looks like it's running the POSIX backend (if it matters):

warning(aio_io_uring): unsupported OP: BIND
warning(aio_linux): io_uring unsupported, using the posix backend

Could it be this is caused by nested waits in the same scheduler? I think that's when the problem started to happen.

In my code I scheduler.spawn a task and then scheduler.run(.wait) for it. That task takes the scheduler as an argument and spawns new subtasks internally, then waits for them one by one with subtask.complete(.wait) to produce a final result.

Once all subtasks end, the original task ends, and that's when the segfault happens.

[Cloudef]

Hello can you provide the code that produces the problem?
The BIND io_uring op requires fairly recent kernel, it's possible to override https://github.com/Cloudef/zig-aio/blob/master/src/aio.zig#L26 if you don't use Bind op in your code.