зеркало из https://github.com/mozilla/pjs.git
Bug 550118: Try to not process XPCOM events when a plugin process spins a nested glib event loop. r=karlt
This commit is contained in:
Родитель
ac65c720e6
Коммит
407c6ae670
|
@ -36,6 +36,10 @@
|
|||
*
|
||||
* ***** END LICENSE BLOCK ***** */
|
||||
|
||||
#ifdef MOZ_WIDGET_GTK2
|
||||
#include <glib.h>
|
||||
#endif
|
||||
|
||||
#include "base/process_util.h"
|
||||
|
||||
#include "mozilla/ipc/SyncChannel.h"
|
||||
|
@ -45,7 +49,6 @@
|
|||
#include "nsContentUtils.h"
|
||||
#include "nsCRT.h"
|
||||
#include "nsNPAPIPlugin.h"
|
||||
#include "nsThreadUtils.h"
|
||||
|
||||
using base::KillProcess;
|
||||
|
||||
|
@ -875,17 +878,12 @@ PluginModuleParent::AnswerProcessSomeEvents()
|
|||
{
|
||||
PLUGIN_LOG_DEBUG(("Spinning mini nested loop ..."));
|
||||
|
||||
// XXX it would seem sensical to make the condition be
|
||||
// |NS_HasPendingEvents() && i < kMaxEventsToProcess|. The
|
||||
// problem is, the native appshell is just an observer of our
|
||||
// nsThread, and processes native events as a side effect of
|
||||
// nsThread::ProcessNextEvent(). Since native events are the ones
|
||||
// we really care about here, we need to go straight to
|
||||
// NS_ProcessNextEvent().
|
||||
for (int i = 0; i < kMaxChancesToProcessEvents; ++i)
|
||||
NS_ProcessNextEvent(nsnull, PR_FALSE);
|
||||
int i = 0;
|
||||
for (; i < kMaxChancesToProcessEvents; ++i)
|
||||
if (!g_main_context_iteration(NULL, FALSE))
|
||||
break;
|
||||
|
||||
PLUGIN_LOG_DEBUG(("... quitting mini nested loop"));
|
||||
PLUGIN_LOG_DEBUG(("... quitting mini nested loop; processed %i tasks", i));
|
||||
|
||||
return true;
|
||||
}
|
||||
|
|
Загрузка…
Ссылка в новой задаче