- Revision
- 169243
- Author
- [email protected]
- Date
- 2014-05-22 20:59:11 -0700 (Thu, 22 May 2014)
Log Message
http/tests/security/xss-DENIED-xsl-document-redirect.xml fails with NetworkProcess
https://bugs.webkit.org/show_bug.cgi?id=132523
Patch by Myles C. Maxfield <[email protected]> on 2014-05-22
Reviewed by Alexey Proskuryakov.
Source/WebCore:
We should not manufacture a response in the event of an error during a
synchronous XHR. In addition, this test removes two places that are
sensitive to such a manufactured response.
Updates test expectations.
* loader/DocumentThreadableLoader.cpp:
(WebCore::DocumentThreadableLoader::loadRequest): Don't inspect a
loader response if there is an error
* platform/network/cf/ResourceHandleCFNet.cpp:
(WebCore::ResourceHandle::platformLoadResourceSynchronously): Do not
manufacture a response
* platform/network/mac/ResourceHandleMac.mm:
(WebCore::ResourceHandle::platformLoadResourceSynchronously): Do not
manufacture a response
* xml/XSLTProcessorLibxslt.cpp:
(WebCore::docLoaderFunc): Only use the response's URL if no error
occurred.
LayoutTests:
Two tests have been rebaselined, and re-enable test that now passes. Because of
r23889, synchronous XHRs for file: URLs purposely don't invoke error handlers.
* fast/xmlhttprequest/xmlhttprequest-nonexistent-file-expected.txt: Updated
incorrect test description
* fast/xmlhttprequest/resources/xmlhttprequest-nonexistent-file-real.html: Ditto
* platform/mac-wk2/TestExpectations: Re-enable test that now passes.
Modified Paths
Diff
Modified: trunk/LayoutTests/ChangeLog (169242 => 169243)
--- trunk/LayoutTests/ChangeLog 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/LayoutTests/ChangeLog 2014-05-23 03:59:11 UTC (rev 169243)
@@ -1,3 +1,18 @@
+2014-05-22 Myles C. Maxfield <[email protected]>
+
+ http/tests/security/xss-DENIED-xsl-document-redirect.xml fails with NetworkProcess
+ https://bugs.webkit.org/show_bug.cgi?id=132523
+
+ Reviewed by Alexey Proskuryakov.
+
+ Two tests have been rebaselined, and re-enable test that now passes. Because of
+ r23889, synchronous XHRs for file: URLs purposely don't invoke error handlers.
+
+ * fast/xmlhttprequest/xmlhttprequest-nonexistent-file-expected.txt: Updated
+ incorrect test description
+ * fast/xmlhttprequest/resources/xmlhttprequest-nonexistent-file-real.html: Ditto
+ * platform/mac-wk2/TestExpectations: Re-enable test that now passes.
+
2014-05-22 Simon Fraser <[email protected]>
Fix the compositing/transitions/transform-on-large-layer.html test
Modified: trunk/LayoutTests/fast/xmlhttprequest/resources/xmlhttprequest-nonexistent-file-real.html (169242 => 169243)
--- trunk/LayoutTests/fast/xmlhttprequest/resources/xmlhttprequest-nonexistent-file-real.html 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/LayoutTests/fast/xmlhttprequest/resources/xmlhttprequest-nonexistent-file-real.html 2014-05-23 03:59:11 UTC (rev 169243)
@@ -71,7 +71,7 @@
</head>
<body _onload_="testXHRNonExistentFile()">
<p> Bug <a href="" REGRESSION: Async XMLHttpRequest never finishes on nonexistent files anymore </p>
- <p> In both cases, readyState 4 should be reached, and error handler should be invoked. </p>
+ <p> In both cases, readyState 4 should be reached. Because one of the requests is synchronous and for a file: URL, only a single error handler should be invoked. </p>
<div id="console"/>
</body>
</html>
Modified: trunk/LayoutTests/fast/xmlhttprequest/xmlhttprequest-nonexistent-file-expected.txt (169242 => 169243)
--- trunk/LayoutTests/fast/xmlhttprequest/xmlhttprequest-nonexistent-file-expected.txt 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/LayoutTests/fast/xmlhttprequest/xmlhttprequest-nonexistent-file-expected.txt 2014-05-23 03:59:11 UTC (rev 169243)
@@ -1,7 +1,7 @@
Bug 22475: REGRESSION: Async XMLHttpRequest never finishes on nonexistent files anymore
-In both cases, readyState 4 should be reached, and error handler should be invoked.
+In both cases, readyState 4 should be reached. Because one of the requests is synchronous and for a file: URL, only a single error handler should be invoked.
Doing an XHR to a nonexistent file.
ReadyState handler: readyState = 1
Modified: trunk/LayoutTests/platform/mac-wk2/TestExpectations (169242 => 169243)
--- trunk/LayoutTests/platform/mac-wk2/TestExpectations 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/LayoutTests/platform/mac-wk2/TestExpectations 2014-05-23 03:59:11 UTC (rev 169243)
@@ -407,9 +407,6 @@
webkit.org/b/123431 [ Mavericks ] http/tests/local/link-stylesheet-load-order-preload.html [ Failure ]
webkit.org/b/123431 [ Mavericks ] http/tests/local/link-stylesheet-load-order.html [ Failure ]
-# passes but has different logging
-[ Mavericks ] http/tests/security/xss-DENIED-xsl-document-redirect.xml [ Failure ]
-
# test results in DRT don't match in browser behavior, test seems broken
fast/dom/Window/mozilla-focus-blur.html [ Failure ]
Modified: trunk/Source/WebCore/ChangeLog (169242 => 169243)
--- trunk/Source/WebCore/ChangeLog 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/Source/WebCore/ChangeLog 2014-05-23 03:59:11 UTC (rev 169243)
@@ -1,3 +1,29 @@
+2014-05-22 Myles C. Maxfield <[email protected]>
+
+ http/tests/security/xss-DENIED-xsl-document-redirect.xml fails with NetworkProcess
+ https://bugs.webkit.org/show_bug.cgi?id=132523
+
+ Reviewed by Alexey Proskuryakov.
+
+ We should not manufacture a response in the event of an error during a
+ synchronous XHR. In addition, this test removes two places that are
+ sensitive to such a manufactured response.
+
+ Updates test expectations.
+
+ * loader/DocumentThreadableLoader.cpp:
+ (WebCore::DocumentThreadableLoader::loadRequest): Don't inspect a
+ loader response if there is an error
+ * platform/network/cf/ResourceHandleCFNet.cpp:
+ (WebCore::ResourceHandle::platformLoadResourceSynchronously): Do not
+ manufacture a response
+ * platform/network/mac/ResourceHandleMac.mm:
+ (WebCore::ResourceHandle::platformLoadResourceSynchronously): Do not
+ manufacture a response
+ * xml/XSLTProcessorLibxslt.cpp:
+ (WebCore::docLoaderFunc): Only use the response's URL if no error
+ occurred.
+
2014-05-22 Myles C. Maxfield <[email protected]>
Unreviewed typo fix.
Modified: trunk/Source/WebCore/loader/DocumentThreadableLoader.cpp (169242 => 169243)
--- trunk/Source/WebCore/loader/DocumentThreadableLoader.cpp 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/Source/WebCore/loader/DocumentThreadableLoader.cpp 2014-05-23 03:59:11 UTC (rev 169243)
@@ -410,9 +410,14 @@
InspectorInstrumentation::documentThreadableLoaderStartedLoadingForClient(&m_document, identifier, m_client);
- // No exception for file:/// resources, see <rdar://problem/4962298>.
- // Also, if we have an HTTP response, then it wasn't a network error in fact.
- if (!error.isNull() && !requestURL.isLocalFile() && response.httpStatusCode() <= 0) {
+ if (!error.isNull() && response.httpStatusCode() <= 0) {
+ if (requestURL.isLocalFile()) {
+ // We don't want XMLHttpRequest to raise an exception for file:// resources, see <rdar://problem/4962298>.
+ // FIXME: XMLHttpRequest quirks should be in XMLHttpRequest code, not in DocumentThreadableLoader.cpp.
+ didReceiveResponse(identifier, response);
+ didFinishLoading(identifier, 0.0);
+ return;
+ }
m_client->didFail(error);
return;
}
Modified: trunk/Source/WebCore/platform/network/cf/ResourceHandleCFNet.cpp (169242 => 169243)
--- trunk/Source/WebCore/platform/network/cf/ResourceHandleCFNet.cpp 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/Source/WebCore/platform/network/cf/ResourceHandleCFNet.cpp 2014-05-23 03:59:11 UTC (rev 169243)
@@ -512,14 +512,6 @@
if (error.isNull())
response = client->response();
- else {
- response = ResourceResponse(request.url(), String(), 0, String(), String());
- // FIXME: ResourceHandleMac also handles authentication errors by setting code to 401. CFNet version should probably do the same.
- if (error.domain() == String(kCFErrorDomainCFNetwork))
- response.setHTTPStatusCode(error.errorCode());
- else
- response.setHTTPStatusCode(404);
- }
data.swap(client->mutableData());
}
Modified: trunk/Source/WebCore/platform/network/mac/ResourceHandleMac.mm (169242 => 169243)
--- trunk/Source/WebCore/platform/network/mac/ResourceHandleMac.mm 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/Source/WebCore/platform/network/mac/ResourceHandleMac.mm 2014-05-23 03:59:11 UTC (rev 169243)
@@ -426,25 +426,6 @@
if (error.isNull())
response = client->response();
- else {
- // FIXME: We might not ever need to manufacture a response: This might all be dead code.
- // When exploring removal of this code, we should substitute appropriate ASSERTs.
- response = ResourceResponse(request.url(), String(), 0, String(), String());
- if (error.domain() == String(NSURLErrorDomain))
- switch (error.errorCode()) {
- case NSURLErrorUserCancelledAuthentication:
- // FIXME: We don't need to manufacture a 401 response if we say continueWithoutCredentialForAuthenticationChallenge:
- // in which case we'll get the real failure response. A reading of SynchronousLoaderClient.mm suggests we already do this.
- response.setHTTPStatusCode(401);
- break;
- default:
- response.setHTTPStatusCode(error.errorCode());
- }
- else {
- // FIXME: This is wrong. We shouldn't need to ever make up a 404.
- response.setHTTPStatusCode(404);
- }
- }
data.swap(client->mutableData());
}
Modified: trunk/Source/WebCore/xml/XSLTProcessorLibxslt.cpp (169242 => 169243)
--- trunk/Source/WebCore/xml/XSLTProcessorLibxslt.cpp 2014-05-23 03:31:57 UTC (rev 169242)
+++ trunk/Source/WebCore/xml/XSLTProcessorLibxslt.cpp 2014-05-23 03:59:11 UTC (rev 169243)
@@ -128,7 +128,10 @@
bool requestAllowed = globalCachedResourceLoader->frame() && globalCachedResourceLoader->document()->securityOrigin()->canRequest(url);
if (requestAllowed) {
globalCachedResourceLoader->frame()->loader().loadResourceSynchronously(url, AllowStoredCredentials, DoNotAskClientForCrossOriginCredentials, error, response, data);
- requestAllowed = globalCachedResourceLoader->document()->securityOrigin()->canRequest(response.url());
+ if (error.isNull())
+ requestAllowed = globalCachedResourceLoader->document()->securityOrigin()->canRequest(response.url());
+ else
+ data.clear();
}
if (!requestAllowed) {
data.clear();
