worker_loop starts by masking signals, since if a signal occurs in this thread it will crash the process:
However, there is a slight chance that a signal will occur before this. Here's a test case:
let pid = Unix.getpid ()
let some_worker_job () =
Lwt_unix.gethostbyname "www.google.com"
let _ =
Sys.(set_signal sigusr1) @@ Signal_handle (fun _ -> print_endline "Got USR1");
Lwt_main.run begin
let job = some_worker_job () in
print_endline "Mask USR1 in main thread to make sure worker gets it";
ignore (Thread.(sigmask SIG_BLOCK) [Sys.sigusr1] : _ list);
print_endline "Sending USR1";
assert (Unix.system ("kill -USR1 " ^ string_of_int pid) = Unix.WEXITED 0);
job
end
If I add a small delay then it reliably segfaults:
static void *worker_loop(void *data) {
lwt_unix_job job = (lwt_unix_job)data;
for (int64_t i = 1; i < 10000000; i++) {
__asm__ __volatile__("");
}
...
Starting new thread
Mask USR1 in main thread to make sure worker gets it
Sending USR1
fish: Job 1, './_build/default/test/unix/sign…' terminated by signal SIGSEGV (Address boundary error)
I spotted this while trying to track down some segfaults in my program. I'm not sure if this is the cause of those, as the race looks hard to hit without the delay.
Using pthread_attr_setsigmask_np to set an initial mask might fix it.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4