[120] in Pthreads mailing list archive
small signals patch for 1_60-beta3
daemon@ATHENA.MIT.EDU (William S. Gribble)
Wed Sep 13 18:32:26 1995
From: "William S. Gribble" <grib@cs.utexas.edu>
Date: Wed, 13 Sep 1995 16:54:16 -0500
To: pthreads@MIT.EDU
Cc: pthreads-bugs@MIT.EDU
I was running into a problem where a thread that had received a signal
via pthread_kill() would never call pthread_sig_process() to handle it.
The thread's sigcount and sigpending were correct, but none of the
four places that call pthread_sig_process were ever being reached.
Sorry I can't give example code; it would be hard to whittle the code
down enough. An effective patch was to add another check for
pending signals in the sig_handler_fake(), which ensures that a thread
with a pending signal won't lose that signal if it is the only thread and
is continuously calling pthread_yield().
I have applied the signal handling patches from the list that were
posted a week or two ago, so your line numbers may vary.
(the patch is to pthreads/pthreads/signal.c)
--- signal.c~ Tue Sep 12 09:33:55 1995
+++ signal.c Wed Sep 13 16:22:06 1995
@@ -420,6 +420,12 @@
}
pthread_kernel_lock++;
sig_handler(sig);
+
+ /* Handle any signals the current thread might have just gotten */
+ if (pthread_run && pthread_run->sigcount) {
+ pthread_sig_process();
+ }
+
while (!(--pthread_kernel_lock)) {
if (SIG_ANY(sig_to_process)) {
pthread_kernel_lock++;
