Bug 42374 – Add possibility to dumpAsText and generate a pixel test result

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Nikolas Zimmermann]

When using layoutTestController.setDumpAsText() most DRT implementations will not generate a pixel test result.
This breaks svg/dynamic-updates test, which don't want to dump render tree information, but pure text, and a pixel test result. Without pixel test results these tests are useless.
It's unfortunate, this hasn't been noticed for a long time - DRT has changed at some point to not generate pixel results when dumpAsText=true.

I propose introducing setDumpAsTextAndPixels logic - already tested, works fine and gives me what I want.

[Nikolas Zimmermann]

Created attachment 61661 [details]
Patch

Maciej, do you think that would be acceptable?

[WebKit Review Bot]

Attachment 61661 [details] did not pass style-queue:

Failed to run "['WebKitTools/Scripts/check-webkit-style']" exit_code: 1
WebKitTools/DumpRenderTree/wx/DumpRenderTreeWx.cpp:206:  Boolean expressions that span multiple lines should have their operators on the left side of the line instead of the right side.  [whitespace/operators] [4]
Total errors found: 1 in 11 files


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

[Nikolas Zimmermann]

The style failure is in existing code, I'll fix it before landing (if this patch gets a positive review).

[Nikolas Zimmermann]

Created attachment 61774 [details]
Updated patch

Updated patch after IRC discussion with Maciej.

[WebKit Review Bot]

Attachment 61774 [details] did not pass style-queue:

Failed to run "['WebKitTools/Scripts/check-webkit-style']" exit_code: 1
WebKitTools/DumpRenderTree/wx/DumpRenderTreeWx.cpp:206:  Boolean expressions that span multiple lines should have their operators on the left side of the line instead of the right side.  [whitespace/operators] [4]
Total errors found: 1 in 11 files


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

[Nikolas Zimmermann]

Oops, fixed style issue (not uploading a new patch to let EWS run)

[WebKit Review Bot]

Attachment 61774 [details] did not build on chromium:
Build output: http://webkit-commit-queue.appspot.com/results/3413423

[Nikolas Zimmermann]

Created attachment 61778 [details]
Updated patch v2

Fix style issue and chromium build.

[Nikolas Zimmermann]

Issue 1:
float* totals is passed from fastSetInteriorPixels, and the new helper function setDestinationPixels is called multiple times. That means you're doing the divisior division multiple times, accumulated!

for (int i = 0; i < numTotals; ++i)
    totals[i] = totals[i] / divisor + bias;

Issue 2:
for (int i = 0; i < 3; ++i)
    image->set(pixel++, clampRGBAValue(totals[0], maxAlpha));

You're always clamping totals[0], not totals[i].

Very evil!

[Nikolas Zimmermann]

Oops, ignore that commented on wrong br.

[Dirk Schulze]

Comment on attachment 61778 [details]
Updated patch v2

LGTM r=me

[Nikolas Zimmermann]

Discussed the concept with Maciej, he was fine with it.
Landed in r63536.