GNU bug report logs - #36609
27.0.50; Possible race-condition in threading implementation

Previous Next

Package: emacs;

Reported by: Andreas Politz <politza <at> hochschule-trier.de>

Date: Thu, 11 Jul 2019 20:52:02 UTC

Severity: normal

Tags: fixed

Found in version 27.0.50

Fixed in version 28.1

Done: dick <dick.r.chiang <at> gmail.com>

Bug is archived. No further changes may be made.

Full log


Message #168 received at 36609 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: dick.r.chiang <at> gmail.com
Cc: 36609 <at> debbugs.gnu.org
Subject: Re: bug#36609: 27.0.50; Possible race-condition in threading
 implementation
Date: Sun, 20 Jun 2021 18:53:40 +0300
> From: dick.r.chiang <at> gmail.com
> Cc: 36609 <at> debbugs.gnu.org
> Date: Sun, 20 Jun 2021 10:01:37 -0400
> 
> Yes!  If you just add a negation to the __atomic_sub_fetch call

You mean, use the patch below instead?

> this passes all the MREs.

Thanks, will install soon.

> It's not clear how you want to deal with the #else GNUC_PREREQ (4, 7, 0).

By hoping no one uses this and expects threads to be stable enough
under GTK.

diff --git a/src/xgselect.c b/src/xgselect.c
index 0d91d55bad..92b118b955 100644
--- a/src/xgselect.c
+++ b/src/xgselect.c
@@ -34,12 +34,27 @@ static GMainContext *glib_main_context;
 
 void release_select_lock (void)
 {
+#if GNUC_PREREQ (4, 7, 0)
+  if (__atomic_sub_fetch (&threads_holding_glib_lock, 1, __ATOMIC_ACQ_REL) == 0)
+    g_main_context_release (glib_main_context);
+#else
   if (--threads_holding_glib_lock == 0)
     g_main_context_release (glib_main_context);
+#endif
 }
 
 static void acquire_select_lock (GMainContext *context)
 {
+#if GNUC_PREREQ (4, 7, 0)
+  if (__atomic_fetch_add (&threads_holding_glib_lock, 1, __ATOMIC_ACQ_REL) == 0)
+    {
+      glib_main_context = context;
+      while (!g_main_context_acquire (context))
+	{
+	  /* Spin. */
+	}
+    }
+#else
   if (threads_holding_glib_lock++ == 0)
     {
       glib_main_context = context;
@@ -48,6 +63,7 @@ static void acquire_select_lock (GMainContext *context)
 	  /* Spin. */
 	}
     }
+#endif
 }
 
 /* `xg_select' is a `pselect' replacement.  Why do we need a separate function?




This bug report was last modified 4 years and 27 days ago.

Previous Next


GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.