bug#29347: 27.0.50; C-g doesn't quit minibuffer

[Eli Zaretskii]

> Date: Fri, 24 Nov 2017 18:17:58 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: 29370@debbugs.gnu.org, 29347@debbugs.gnu.org, rms@gnu.org
> 
> I think I fixed the problem which caused Emacs to hang in this
> particular scenario, but I'm not sure Richard is seeing the same
> problem.  The change I installed on the emacs-26 branch is below;
> Richard, please try applying it, and see if your problems go away.

Sorry, wrong patch.  Please use the one below instead.

diff --git a/src/thread.c b/src/thread.c
index c03cdda..1ded8f5 100644
--- a/src/thread.c
+++ b/src/thread.c
@@ -97,7 +97,12 @@ post_acquire_global_lock (struct thread_state *self)
 static void
 acquire_global_lock (struct thread_state *self)
 {
-  sys_mutex_lock (&global_lock);
+  /* If some Lisp was interrupted by C-g while inside pselect, the
+     signal handler could have called maybe_reacquire_global_lock, in
+     which case we are already holding the lock and shouldn't try
+     taking it again, or else we will hang forever.  */
+  if (!(self && !self->not_holding_lock))
+    sys_mutex_lock (&global_lock);
   post_acquire_global_lock (self);
 }