Bug 54874 – Frame opener is not reset between tests.

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Vsevolod Vlasov]

Layout tests for XML viewer (see https://bugs.webkit.org/show_bug.cgi?id=13807) should be run in a frame with frame opener reset to null.

[Vsevolod Vlasov]

Created attachment 83319 [details]
Patch

[WebKit Review Bot]

Attachment 83319 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WebCore/ChangeLog', u'Source/WebCor..." exit_code: 1

Source/WebKit/win/WebFrame.cpp:752:  Extra space after ( in function call  [whitespace/parens] [4]
Source/WebKit/win/WebFrame.h:147:  The parameter name "frame" adds no information, so it should be removed.  [readability/parameter_name] [5]
Source/WebKit/win/WebFrame.h:149:  Extra space after ( in function call  [whitespace/parens] [4]
Total errors found: 3 in 28 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Build Bot]

Attachment 83319 [details] did not build on win:
Build output: http://queues.webkit.org/results/7943323

[Vsevolod Vlasov]

Created attachment 83322 [details]
Patch

[WebKit Review Bot]

Attachment 83322 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WebCore/ChangeLog', u'Source/WebCor..." exit_code: 1

Source/WebKit/win/WebFrame.cpp:752:  Extra space after ( in function call  [whitespace/parens] [4]
Source/WebKit/win/WebFrame.h:146:  Extra space after ( in function call  [whitespace/parens] [4]
Total errors found: 2 in 28 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Build Bot]

Attachment 83322 [details] did not build on win:
Build output: http://queues.webkit.org/results/7943332

[Alexey Proskuryakov]

Comment on attachment 83322 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=83322&action=review

> Source/WebCore/ChangeLog:8
> +        No new tests. (no code affected)

You could as well explain what has been done, not what hasn't. E.g.: "no behavior change, just exporting a method for DumpRenderTree use."

> Source/WebKit/chromium/public/WebFrame.h:173
> +    // Reset the frame that opened this frame to 0.
> +    virtual void resetOpener() = 0;

Here an in all other ports: it's arguable whether this method resets opener or not (resets to what?). I suggest naming it clearOpener() or forgetOpener().

> Source/WebKit/gtk/webkit/webkitwebframe.h:112
> +WEBKIT_API void
> +webkit_web_frame_reset_opener       (WebKitWebFrame       *frame);

This should probably be in webkitwebframeprivate.h - you could ask Gtk folks to be sure.

> Source/WebKit/mac/WebView/WebFrame.h:195
> +    @method resetOpener
> +    @discussion Resets mainFrame opener to 0
> +*/
> +- (void)resetOpener;

Please don't change anything in WebFrame.h - it's public Apple API, and any changes need to be vetted in an internal Apple process.

New methods can be added to WebFramePrivate.h.

> Source/WebKit/qt/Api/qwebframe.h:149
> +    void resetOpener();

Again, not sure if it's OK to add an API method. The private interface is in qwebframe_p.h.

>> Source/WebKit/win/WebFrame.cpp:752
>> +HRESULT STDMETHODCALLTYPE WebFrame::resetOpener( void)
> 
> Extra space after ( in function call  [whitespace/parens] [4]

Why does this need a (void) at all? That only sometimes makes a difference in plain C.

>> Source/WebKit/win/WebFrame.h:146
>> +    virtual HRESULT STDMETHODCALLTYPE resetOpener( void);
> 
> Extra space after ( in function call  [whitespace/parens] [4]

Ditto - probably doesn't need (void).

> Source/WebKit2/ChangeLog:9
> +        (WKBundleFrameCopyChildFrames):

Please add a ChangeLog comment explaining what you did.

> Source/WebKit2/ChangeLog:10
> +        (WKBundleFrameResetOpener):

WebKit2 isn't an API, so we are more relaxed about adding functions there at the moment. But WKBundleFrameResetOpener doesn't look like they would ever become APIs, so it's best to put them in WKBundleFramePrivate.h right away.

> Tools/ChangeLog:6
> +        DumpRenderTree should reset frame opener between tests.
> +        https://bugs.webkit.org/show_bug.cgi?id=54874

Please explain what issue you are fixing. You are saying that opener needs to be cleared, but what to and when is it set, in the first place?

> Tools/DumpRenderTree/win/DumpRenderTree.cpp:895
> +    if (FAILED(frame->resetOpener()))
> +        return;

Your implementation of resetOpener can never FAIL, as far as I can tell. Perhaps it's best to ASSERT the result - returning right before exiting the function makes no sense anyway.

[Vsevolod Vlasov]

Comment on attachment 83322 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=83322&action=review

>> Source/WebCore/ChangeLog:8
>> +        No new tests. (no code affected)
> 
> You could as well explain what has been done, not what hasn't. E.g.: "no behavior change, just exporting a method for DumpRenderTree use."

Done

>> Source/WebKit/chromium/public/WebFrame.h:173
>> +    virtual void resetOpener() = 0;
> 
> Here an in all other ports: it's arguable whether this method resets opener or not (resets to what?). I suggest naming it clearOpener() or forgetOpener().

Done (clearOpener).

>> Source/WebKit/gtk/webkit/webkitwebframe.h:112
>> +webkit_web_frame_reset_opener       (WebKitWebFrame       *frame);
> 
> This should probably be in webkitwebframeprivate.h - you could ask Gtk folks to be sure.

Done (DumpRenderTreeSupportGTK).

>> Source/WebKit/mac/WebView/WebFrame.h:195
>> +- (void)resetOpener;
> 
> Please don't change anything in WebFrame.h - it's public Apple API, and any changes need to be vetted in an internal Apple process.
> 
> New methods can be added to WebFramePrivate.h.

Done.

>> Source/WebKit/qt/Api/qwebframe.h:149
>> +    void resetOpener();
> 
> Again, not sure if it's OK to add an API method. The private interface is in qwebframe_p.h.

Done (DumpRenderTreeSupportQT).

>>> Source/WebKit/win/WebFrame.cpp:752
>>> +HRESULT STDMETHODCALLTYPE WebFrame::resetOpener( void)
>> 
>> Extra space after ( in function call  [whitespace/parens] [4]
> 
> Why does this need a (void) at all? That only sometimes makes a difference in plain C.

Some methods there were written in this weird form.
I changed it to clearOpener().

>>> Source/WebKit/win/WebFrame.h:146
>>> +    virtual HRESULT STDMETHODCALLTYPE resetOpener( void);
>> 
>> Extra space after ( in function call  [whitespace/parens] [4]
> 
> Ditto - probably doesn't need (void).

Done.

>> Source/WebKit2/ChangeLog:9
>> +        (WKBundleFrameCopyChildFrames):
> 
> Please add a ChangeLog comment explaining what you did.

Done.

>> Source/WebKit2/ChangeLog:10
>> +        (WKBundleFrameResetOpener):
> 
> WebKit2 isn't an API, so we are more relaxed about adding functions there at the moment. But WKBundleFrameResetOpener doesn't look like they would ever become APIs, so it's best to put them in WKBundleFramePrivate.h right away.

Done.

>> Tools/ChangeLog:6
>> +        https://bugs.webkit.org/show_bug.cgi?id=54874
> 
> Please explain what issue you are fixing. You are saying that opener needs to be cleared, but what to and when is it set, in the first place?

Done.

>> Tools/DumpRenderTree/win/DumpRenderTree.cpp:895
>> +        return;
> 
> Your implementation of resetOpener can never FAIL, as far as I can tell. Perhaps it's best to ASSERT the result - returning right before exiting the function makes no sense anyway.

Done.

[Vsevolod Vlasov]

Created attachment 83383 [details]
Patch with fixes

[Build Bot]

Attachment 83383 [details] did not build on win:
Build output: http://queues.webkit.org/results/7946377

[Alexey Proskuryakov]

Comment on attachment 83383 [details]
Patch with fixes

Looks fine. cq-, because this still breaks Windows.

[Vsevolod Vlasov]

Created attachment 83536 [details]
Patch with fixes for win build

The only change from previous patch is touching (adding new line) Source/WebKit/win/Interfaces/WebKit.idl file to trigger build scripts.
This change is also mentioned in ChangeLog

[WebKit Commit Bot]

The commit-queue encountered the following flaky tests while processing attachment 83536 [details]:

http/tests/security/cross-frame-access-port-explicit-domain.html bug 54668 (author: [e-mail])
The commit-queue is continuing to process your patch.

[WebKit Commit Bot]

Comment on attachment 83536 [details]
Patch with fixes for win build

Clearing flags on attachment: 83536

Committed r79570: <http://trac.webkit.org/changeset/79570>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Andrew Wilson]

Reverted r79570 for reason:

Breaks chromium build because glue/mocks/mock_web_frame.h/cc was not updated

Committed r79586: <http://trac.webkit.org/changeset/79586>

[WebKit Commit Bot]

Comment on attachment 83536 [details]
Patch with fixes for win build

Clearing flags on attachment: 83536

Committed r79764: <http://trac.webkit.org/changeset/79764>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Ryosuke Niwa]

Broke Chromium builds: http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Win%20Builder/builds/3981/steps/compile/logs/stdio

[Ryosuke Niwa]

I'm sorry but I don't know how to fix the build so I'm rolling out the patch for now since the author is using @chromium.org address.  You should probably add a concrete clearOpener on the chromium side first and then land this patch.

[Pavel Feldman]

Committed r79793: <http://trac.webkit.org/changeset/79793>