APPROVE COMMIT 21.5 This is a version of the earlier patch with cleaner comments. diff -r 27b87c12a3ab -r 4dcbe828fa9c src/ChangeLog --- a/src/ChangeLog Fri Dec 29 10:50:19 2017 +0000 +++ b/src/ChangeLog Fri Dec 29 20:45:41 2017 +0900 ＠＠ -1,3 +1,8 ＠＠ +2017-12-29 Stephen J. Turnbull <stephen(a)xemacs.org&gt; + + * event-gtk.c (emacs_gtk_drain_queue): + Work around gdk_events_pending lossage. + 2017-12-29 Aidan Kehoe <kehoea(a)parhasard.net&gt; * doprnt.c (emacs_vsprintf_string): diff -r 27b87c12a3ab -r 4dcbe828fa9c src/event-gtk.c --- a/src/event-gtk.c Fri Dec 29 10:50:19 2017 +0000 +++ b/src/event-gtk.c Fri Dec 29 20:45:41 2017 +0900 ＠＠ -1699,12 +1699,21 ＠＠ it on the queue, then we go into Fnext_event(), which calls emacs_gtk_drain_queue(). But gtk_events_pending() will always return TRUE if there are file-descriptor (aka our process) events - pending. Using GDK_events_pending() only shows us windowing - system events. + pending. + GDK_events_pending() documentation says it only shows us + windowing system events, but in GTK3 on Mac OS X "High Sierra" at least + it returns true even though gdk_event_peek returns NULL. So we use the + the latter. */ if (gdk_display_get_default ()) - while (gdk_events_pending ()) - gtk_main_iteration_do (FALSE); + { + GdkEvent *e = gdk_event_peek(); + while (e) { + gtk_main_iteration_do (FALSE); + gdk_event_free(e); + e = gdk_event_peek(); + } + } #ifdef HAVE_TTY drain_tty_devices ();