tree a78c38c25c011f0aad6e4a69cd32b5143c110d86
parent e5455c00d153aa0cb995eceea1f8eb9b77f3464f
author ap@apple.com <ap@apple.com@268f45cc-cd09-0410-ab3c-d52691b4dbfc> 1383238203 +0000
committer ap@apple.com <ap@apple.com@268f45cc-cd09-0410-ab3c-d52691b4dbfc> 1383238203 +0000

REGRESSION(r158333): http/tests/xmlhttprequest/response-encoding.html and xmlhttprequest-overridemimetype-content-type-header.html are failing
https://bugs.webkit.org/show_bug.cgi?id=123548

Reviewed by Brady Eidson.

Source/WebCore:

We had code that made sure that cached 200 responses weren't used for conditional
requests. But it didn't work the other way - cached 304 responses got reused for
subsequent unconditional requests!

Adding the test uncovered this bug.

* loader/cache/CachedRawResource.cpp: (WebCore::shouldIgnoreHeaderForCacheReuse):
Should never ignore conditional headers. Code in determineRevalidationPolicy
was already undoing this for conditional requests, but we also shouldn't use
WebCore cache if it holds a 304 response to conditional request.

* loader/cache/CachedResourceLoader.cpp: (WebCore::CachedResourceLoader::determineRevalidationPolicy):
Even though the changed code is only for raw resources, I think that we can never
get a conditional request here any more.

LayoutTests:

* TestExpectations: Unskip tests that used to be affected by response-empty-arraybuffer.html

* http/tests/xmlhttprequest/response-empty-arraybuffer-expected.txt:
* http/tests/xmlhttprequest/response-empty-arraybuffer.html:
Fix a stupid typo. This test actually fully passes.


git-svn-id: http://svn.webkit.org/repository/webkit/trunk@158362 268f45cc-cd09-0410-ab3c-d52691b4dbfc