wenyongh
commented
11 months ago @g0djan I haven't seen the similar issue before, it seems like there is something wrong with the lock operations, maybe a mutex is unlocked without locking it before or without initializing it? Could you help dump the stacktrace? If it is not easy, maybe you can change the code to dump which file/line is calling os_mutex_unlock:
//platform_api_vmcore.h
int
os_mutex_unlock_internal(korp_mutex *mutex, char *file, int line);
#define os_mutex_unlock(mutex) os_mutex_unlock(mutex, __FILE__, __LINE__)
//c file, like posix_thread.c
//change os_mutex_unlock to
int
os_mutex_unlock_internal(korp_mutex *mutex, char *file, int line)
{
int ret;
os_printf("##%s, line: %d\n", file, line);
assert(mutex);
ret = pthread_mutex_unlock(mutex);
return ret == 0 ? BHT_OK : BHT_ERROR;
}
g0djan
It happens on different tests for me, but this stacktace I got running this test https://github.com/bytecodealliance/libc-test/blob/master/src/functional/strtof.c and this is the most desciptive stacktrace for this problem I've got so far.
Also Sig11 happened on https://github.com/bytecodealliance/libc-test/blob/master/src/functional/clock_gettime.c but I didn't get any stacktrace for it.
Compiled the test to wasm as part of wasi-libc tests following manual and later to AOT by running
This I got from running on Nvidia Shield TV P2897. Though it's hard to reproduce on Nvidia shield tv. I got similar SIG11 on a Roku 4200X device which reproduces very well but I didn't figure whether it's the same problem or not.
@wenyongh @yamt @abrown has anyone seen anything similar ever?