Original bug ID: 4979
Reporter: gerd
Assigned to: @xavierleroy
Status: closed (set by @xavierleroy on 2011-05-29T10:19:34Z)
Resolution: fixed
Priority: normal
Severity: minor
Version: 3.11.1
Fixed in version: 3.12.0+dev
Category: ~DO NOT USE (was: OCaml general)

Bug description

Just got an error Sys_error("Mutex.lock: error code 0") in a win32 program using threads heavily. I wondered how that was possible. Apparently, the true reason of the error is hidden, and the error code 0 is a lie. (It's usually an invalid descriptor.)

So I dug a bit deeper into the ocaml runtime, and found this (systhreads/win32.c):

CAMLprim value caml_mutex_lock(value mut)
{
int retcode;
Begin_root(mut) /* prevent deallocation of mutex */
enter_blocking_section();
retcode = WaitForSingleObject(Mutex_val(mut), INFINITE);
leave_blocking_section();
End_roots();
if (retcode == WAIT_FAILED) caml_wthread_error("Mutex.lock");
return Val_unit;
}

In caml_wthread_error, the win32 function GetLastError is used to obtain the error code. However, this code was already overwritten in leave_block_section (there is a WaitForSingleObject right at the beginning to get the master lock).

Suggested fix: Get the error code before leave_blocking_section, and pass it to caml_wthread_error.