1.21.0 is crashing Ruby

[travisbell]

Hey @ioquatix ,

I bumped io-event this morning and it doesn't seem like a very happy camper. We're using io_uring in prod which seems to be the trigger. Everything is fine here in dev (macOS) but on prod something is causing it to segfault.

This is Ruby 4.0.6, with latest async-* everything. I am not sure if I can provide you with something else, or if you're able to reproduce it.

For now, I've locked io-event to 1.19.5 which seems stable.

[tavianator]

Can you test a candidate fix? I will have a PR up shortly

[travisbell]

Yup, no problem.

[ioquatix]

Sorry about this... working on it.

[ioquatix]

Reproduced the segfault:

-- C level backtrace information -------------------------------------------
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(rb_print_backtrace+0x11) [0x55c787fb5ce7] /home/samuel/src/ruby-4.0.5/vm_dump.c:1105
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(rb_vm_bugreport) /home/samuel/src/ruby-4.0.5/vm_dump.c:1450
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(rb_bug_for_fatal_signal+0xfd) [0x55c787f67b2d] /home/samuel/src/ruby-4.0.5/error.c:1131
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(sigsegv+0x42) [0x55c787b9c9e2] /home/samuel/src/ruby-4.0.5/signal.c:948
/usr/lib/libc.so.6(0x7f4ade43e6f0) [0x7f4ade43e6f0]
/home/samuel/Developer/socketry/io-event/ext/IO_Event.so(IO_Event_List_pop+0x7) [0x7f4ac29c39b4] ./io/event/selector/../list.h:61
/home/samuel/Developer/socketry/io-event/ext/IO_Event.so(IO_Event_Selector_URing_Completion_acquire) ./io/event/selector/uring.c:201
/home/samuel/Developer/socketry/io-event/ext/IO_Event.so(IO_Event_Selector_URing_io_wait) ./io/event/selector/uring.c:859
/home/samuel/Developer/socketry/io-event/ext/IO_Event.so(IO_Event_Selector_URing_io_wait) (null):0
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(vm_call_cfunc_with_frame_+0xfc) [0x55c787c3a8ec] /home/samuel/src/ruby-4.0.5/vm_insnhelper.c:3901
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(vm_sendish+0xb5) [0x55c787c5f7ea] /home/samuel/src/ruby-4.0.5/vm_insnhelper.c:6123
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(vm_exec_core) /home/samuel/src/ruby-4.0.5/insns.def:904
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(vm_exec_loop+0xa) [0x55c787c4ee8a] /home/samuel/src/ruby-4.0.5/vm.c:2825
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(rb_vm_exec) /home/samuel/src/ruby-4.0.5/vm.c:2801
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(rb_vm_invoke_proc+0x65) [0x55c787c52985] /home/samuel/src/ruby-4.0.5/vm.c:1965
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(rb_fiber_start+0x1a6) [0x55c787f39196] /home/samuel/src/ruby-4.0.5/cont.c:2595
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(ec_switch+0x0) [0x55c787f394e2] /home/samuel/src/ruby-4.0.5/cont.c:935
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(fiber_restore_thread) /home/samuel/src/ruby-4.0.5/cont.c:901
/home/samuel/.rubies/ruby-4.0.5/bin/ruby(fiber_entry) /home/samuel/src/ruby-4.0.5/cont.c:933

cc @tavianator

[ioquatix]

Potential fix: #219

[ioquatix]

@travisbell are you able to try v1.21.1 and report back? We fixed at least 3 different issues, one of them was a reproducible crash ☝🏼

[travisbell]

I just deployed this to our staging environment and I am not seeing the crash anymore.

Thanks for the quick fix and turnaround guys!

[tavianator]

No problem, thanks for testing! (Also thanks for TMDB, I use it quite a lot)