759421 - Reftest analyzer doesn't update images properly

Status: RESOLVED FIXED

(Testing :: Reftest, defect)

Description

[Kartikaya Gupta (email:[e-mail])]

1. Run the reftest analyzer on the android-R1 failure at https://tbpl.mozilla.org/?tree=Try&rev=1392fae7b86c
2. Click on the first failure (test-displayport-bg.html) and view the "Image 2". Observe how there is no image
3. Click on the second failure (test_bug641198.html)
4. Click on the first failure again (test-displayport-bg.html) and view the "Image 2". Observe how there is now an image, and it is the "Image 2" from the second failure.

For some reason the analyzer is unable to properly load the "Image 2" for the first failure, and just shows whatever was last shown when it is requested to show that image.

Comment 1

[Kartikaya Gupta (email:[e-mail])]

This probably happens because the base64 image data in the log is interrupted by random lines looking like this:

before 368640, after 368640, break 06ce5000
before 368640, after 368640, break 06ce5000
before 368640, after 368640, break 06ce5000
before 368640, after 368640, break 06ce5000

So the base64 image data parsed out of the log is not a valid image.

Comment 2

[Kartikaya Gupta (email:[e-mail])]

Attachment: Patch

Picking Cameron for review since he was the last to make significant changes to that file, I think. Feel free to tag somebody else for review.

Comment 3

[Kartikaya Gupta (email:[e-mail])]

Attachment: Patch

Oops, picked up a stray unrelated change in the last patch.

Comment 4

[Cameron McCormack (:heycam)]

Comment on attachment 628044 [details] [diff] [review]
Patch

I'm not a fan of the large embedded data URI here.  Can you add this image as a separate file and set the <image>'s xlink:href to reference it on error?  (Or would there be a problem with the getImageData call after drawing that http-referenced image on the canvas, which there wouldn't be with the data: URI one?)

Comment 5

[Kartikaya Gupta (email:[e-mail])]

I'm not really a fan of either but I wanted to make sure that the analyzer remained a standalone file. If I move it out into a separate file then anybody using the analyzer as a standalone will have to update to also include the image. Note that if the error image doesn't load it will go into an infinite loop because it will run the error handler again, so that is really an undesirable scenario.

Comment 6

[Cameron McCormack (:heycam)]

Comment on attachment 628044 [details] [diff] [review]
Patch

OK, fair enough.

Comment 7

[Kartikaya Gupta (email:[e-mail])]

https://hg.mozilla.org/integration/mozilla-inbound/rev/17d9f93a9ff7

Comment 8

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/17d9f93a9ff7