[SCM] WebKit Debian packaging branch, webkit-1.2, updated. upstream/1.1.90-6072-g9a69373
dimich at chromium.org
dimich at chromium.org
Thu Apr 8 02:19:08 UTC 2010
The following commit has been merged in the webkit-1.2 branch:
commit 606df2f517494a2cc91b5e86a189955271ae905e
Author: dimich at chromium.org <dimich at chromium.org@268f45cc-cd09-0410-ab3c-d52691b4dbfc>
Date: Thu Mar 11 00:08:18 2010 +0000
Make Document::postTask to use a single queue of tasks, to fire them in order
https://bugs.webkit.org/show_bug.cgi?id=35943
Reviewed by Darin Adler.
JavaScriptCore:
The patch uses CFRunLoopTimer to schedule execution of tasks instead of performSelectorOnMainThread which apparently can starve other event sources.
The timer is used when the schedule request is coming on the main thread itself. This happens when the task is posted on the main thread or
when too many tasks are posted and the queue does 'stop and re-schedule' to make sure run loop has a chance to execute other events.
* wtf/mac/MainThreadMac.mm:
(WTF::timerFired):
(WTF::postTimer):
(WTF::scheduleDispatchFunctionsOnMainThread): Use timer posted to the current RunLoop if scheduling the task execution while on the main thread.
WebCore:
Test: existing worker-cloneport.html which was broken by initial patch in http://trac.webkit.org/changeset/55593.
Additional test which indirectly verifies the order of execution will come as part of https://bugs.webkit.org/show_bug.cgi?id=34726
* dom/Document.cpp:
(WebCore::Document::postTask): Always use the same task queue, independent of what thread is posting the task.
LayoutTests:
Updated the test since the order of events coming from independent task sources has changed.
This is a behavior change, but is consistent with the spec (which specifically says these events may be fired
in any order) and the last version of FF (3.5.8) which was completing the test successfully.
* http/tests/appcache/top-frame-2-expected.txt:
* http/tests/appcache/top-frame-2.html:
git-svn-id: http://svn.webkit.org/repository/webkit/trunk@55816 268f45cc-cd09-0410-ab3c-d52691b4dbfc
diff --git a/JavaScriptCore/ChangeLog b/JavaScriptCore/ChangeLog
index d198bf2..6e198ec 100644
--- a/JavaScriptCore/ChangeLog
+++ b/JavaScriptCore/ChangeLog
@@ -1,3 +1,19 @@
+2010-03-10 Dmitry Titov <dimich at chromium.org>
+
+ Reviewed by Darin Adler.
+
+ Make Document::postTask to use a single queue of tasks, to fire them in order
+ https://bugs.webkit.org/show_bug.cgi?id=35943
+
+ The patch uses CFRunLoopTimer to schedule execution of tasks instead of performSelectorOnMainThread which apparently can starve other event sources.
+ The timer is used when the schedule request is coming on the main thread itself. This happens when the task is posted on the main thread or
+ when too many tasks are posted and the queue does 'stop and re-schedule' to make sure run loop has a chance to execute other events.
+
+ * wtf/mac/MainThreadMac.mm:
+ (WTF::timerFired):
+ (WTF::postTimer):
+ (WTF::scheduleDispatchFunctionsOnMainThread): Use timer posted to the current RunLoop if scheduling the task execution while on the main thread.
+
2010-03-10 Geoffrey Garen <ggaren at apple.com>
Windows build fix: added new symbol.
diff --git a/JavaScriptCore/wtf/mac/MainThreadMac.mm b/JavaScriptCore/wtf/mac/MainThreadMac.mm
index 0ddd5f6..957431e 100644
--- a/JavaScriptCore/wtf/mac/MainThreadMac.mm
+++ b/JavaScriptCore/wtf/mac/MainThreadMac.mm
@@ -29,8 +29,10 @@
#import "config.h"
#import "MainThread.h"
+#import <CoreFoundation/CoreFoundation.h>
#import <Foundation/NSThread.h>
#import <wtf/Assertions.h>
+#import <wtf/Threading.h>
@interface WTFMainThreadCaller : NSObject {
}
@@ -63,9 +65,36 @@ void initializeMainThreadPlatform()
#endif
}
+static bool isTimerPosted; // This is only accessed on the 'main' thread.
+
+static void timerFired(CFRunLoopTimerRef timer, void*)
+{
+ CFRelease(timer);
+ isTimerPosted = false;
+ WTF::dispatchFunctionsFromMainThread();
+}
+
+void postTimer()
+{
+ ASSERT(isMainThread());
+
+ if (isTimerPosted)
+ return;
+
+ isTimerPosted = true;
+ CFRunLoopAddTimer(CFRunLoopGetCurrent(), CFRunLoopTimerCreate(0, 0, 0, 0, 0, timerFired, 0), kCFRunLoopCommonModes);
+}
+
+
void scheduleDispatchFunctionsOnMainThread()
{
ASSERT(staticMainThreadCaller);
+
+ if (isMainThread()) {
+ postTimer();
+ return;
+ }
+
#if USE(WEB_THREAD)
[staticMainThreadCaller performSelector:@selector(call) onThread:webThread withObject:nil waitUntilDone:NO];
#else
diff --git a/LayoutTests/ChangeLog b/LayoutTests/ChangeLog
index 542ec06..8dda756 100644
--- a/LayoutTests/ChangeLog
+++ b/LayoutTests/ChangeLog
@@ -1,3 +1,17 @@
+2010-03-10 Dmitry Titov <dimich at chromium.org>
+
+ Reviewed by Darin Adler.
+
+ Make Document::postTask to use a single queue of tasks, to fire them in order
+ https://bugs.webkit.org/show_bug.cgi?id=35943
+
+ Updated the test since the order of events coming from independent task sources has changed.
+ This is a behavior change, but is consistent with the spec (which specifically says these events may be fired
+ in any order) and the last version of FF (3.5.8) which was completing the test successfully.
+
+ * http/tests/appcache/top-frame-2-expected.txt:
+ * http/tests/appcache/top-frame-2.html:
+
2010-03-10 Antonio Gomes <tonikitoo at webkit.org>
The test was introduced in r55796, it has been failing in Gtk bots since then.
diff --git a/LayoutTests/http/tests/appcache/top-frame-2-expected.txt b/LayoutTests/http/tests/appcache/top-frame-2-expected.txt
index 6e99bb6..41a3fe0 100644
--- a/LayoutTests/http/tests/appcache/top-frame-2-expected.txt
+++ b/LayoutTests/http/tests/appcache/top-frame-2-expected.txt
@@ -2,6 +2,5 @@ Test that a subframe without manifest gets picked by a relevant application cach
Should say SUCCESS:
-checking
SUCCESS
diff --git a/LayoutTests/http/tests/appcache/top-frame-2.html b/LayoutTests/http/tests/appcache/top-frame-2.html
index c239d08..edde8c2 100644
--- a/LayoutTests/http/tests/appcache/top-frame-2.html
+++ b/LayoutTests/http/tests/appcache/top-frame-2.html
@@ -16,26 +16,47 @@ function log(message)
document.getElementById("result").innerHTML += message + "<br>";
}
+function debug(message)
+{
+ // If running manually in the browser, print the sequence of events.
+ if (!window.layoutTestController)
+ log(message);
+}
+
+var receivedExpectedMessage = false;
+var receivedCheckingEvent = false;
+var receivedNoupdateEvent = false;
+
function test()
{
applicationCache.onnoupdate = null;
applicationCache.oncached = null;
// When a new main resource is associated with the cache, an update should be started.
- applicationCache.onchecking = function() { log("checking") }
+ applicationCache.onchecking = function() { debug("checking"); receivedCheckingEvent = true; checkDone(); }
+ applicationCache.onnoupdate = function() { debug("noupdate"); receivedNoupdateEvent = true; checkDone(); }
var ifr = document.createElement("iframe");
ifr.setAttribute("src", "resources/subframe-2.html");
document.body.appendChild(ifr);
}
+function checkDone()
+{
+ if (receivedExpectedMessage && receivedCheckingEvent && receivedNoupdateEvent) {
+ log("SUCCESS");
+ if (window.layoutTestController)
+ layoutTestController.notifyDone();
+ }
+}
+
applicationCache.onnoupdate = function() { test() }
applicationCache.oncached = function() { test() }
applicationCache.onupdateready = function() { log("FAIL: received unexpected updateready event") }
applicationCache.onerror = function() { log("FAIL: received unexpected error event") }
-window.addEventListener("message", function() { log("SUCCESS"); if (window.layoutTestController) layoutTestController.notifyDone() }, false);
+window.addEventListener("message", function() { debug("message"); receivedExpectedMessage = true; checkDone(); }, false);
</script>
</body>
diff --git a/WebCore/ChangeLog b/WebCore/ChangeLog
index a018878..a9614fd 100644
--- a/WebCore/ChangeLog
+++ b/WebCore/ChangeLog
@@ -1,3 +1,16 @@
+2010-03-10 Dmitry Titov <dimich at chromium.org>
+
+ Reviewed by Darin Adler.
+
+ Make Document::postTask to use a single queue of tasks, to fire them in order
+ https://bugs.webkit.org/show_bug.cgi?id=35943
+
+ Test: existing worker-cloneport.html which was broken by initial patch in http://trac.webkit.org/changeset/55593.
+ Additional test which indirectly verifies the order of execution will come as part of https://bugs.webkit.org/show_bug.cgi?id=34726
+
+ * dom/Document.cpp:
+ (WebCore::Document::postTask): Always use the same task queue, independent of what thread is posting the task.
+
2010-03-10 Sanjeev Radhakrishnan <sanjeevr at chromium.org>
Reviewed by Darin Fisher.
diff --git a/WebCore/dom/Document.cpp b/WebCore/dom/Document.cpp
index 7a78fb3..243f83c 100644
--- a/WebCore/dom/Document.cpp
+++ b/WebCore/dom/Document.cpp
@@ -4685,25 +4685,6 @@ void Document::scriptImported(unsigned long identifier, const String& sourceStri
#endif
}
-class ScriptExecutionContextTaskTimer : public TimerBase {
-public:
- ScriptExecutionContextTaskTimer(PassRefPtr<Document> context, PassOwnPtr<ScriptExecutionContext::Task> task)
- : m_context(context)
- , m_task(task)
- {
- }
-
-private:
- virtual void fired()
- {
- m_task->performTask(m_context.get());
- delete this;
- }
-
- RefPtr<Document> m_context;
- OwnPtr<ScriptExecutionContext::Task> m_task;
-};
-
struct PerformTaskContext : Noncopyable {
PerformTaskContext(PassRefPtr<DocumentWeakReference> documentReference, PassOwnPtr<ScriptExecutionContext::Task> task)
: documentReference(documentReference)
@@ -4730,12 +4711,7 @@ static void performTask(void* ctx)
void Document::postTask(PassOwnPtr<Task> task)
{
- if (isMainThread()) {
- ScriptExecutionContextTaskTimer* timer = new ScriptExecutionContextTaskTimer(static_cast<Document*>(this), task);
- timer->startOneShot(0);
- } else {
- callOnMainThread(performTask, new PerformTaskContext(m_weakReference, task));
- }
+ callOnMainThread(performTask, new PerformTaskContext(m_weakReference, task));
}
Element* Document::findAnchor(const String& name)
--
WebKit Debian packaging
More information about the Pkg-webkit-commits
mailing list