Bug 28661 – [REGRESSION] Flickr thumbnails aren't clickable, and sometimes don't load completely

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[Alin S]

Created attachment 38440 [details]
Flickr failure to load

In flickr.com when you immediately log in you are taken to your personal homepage, in current nightlies none of the images are clickable directly unless you click the very bottom edge, this is not normal function. Some images fail to load except the bottom clickable edge alltogether. A web archive of the problem resolves the problem, although when flickr.com is viewed in real time it fails to load images sometimes as seen in the attached screenshot along with the only bottom edge clickable problem as stated above. I am not a developer but this function is not correct and has something to do with rendering / layout. This limits the usability of the browser for me severely and I am sure anyone else with this bug is impacted severely by it.

[Alin S]

in addition requested by bdash in IRC I am using build #47677 I first formally noticed this about 2-3 nights ago, somewhere around build #47546 or the build before it, but no more than 5 days ago. I have 1password installed, I tried this on my fresh install on a different computer and there is no difference. still bugged.

[Alin S]

continues to exist in r47686

[Alin S]

still present 47880

[Alin S]

WORKS 47388.

WORKS 47409

BROKEN AS OF 47546

[Mark Rowe (bdash)]

<rdar://problem/7181507>

[Alin S]

r47883 still present.

[Alin S]

still here 48355

[Adele Peterson]

More useful information would be to know exactly what revision caused this problem.

[Alin S]

WORKS 47388.

WORKS 47409

BROKEN AS OF 47546

(In reply to comment #8)
> More useful information would be to know exactly what revision caused this
> problem.

[Adele Peterson]

That's a range.

[Eric Seidel (no email)]

http://trac.webkit.org/changeset/47546 does not seem related.  Maybe I'm not understanding your comment.  Do you know the exact revision which regressed, or just the nightly range?

[Alin S]

(In reply to comment #10)
> That's a range.

it is broken in 47546 + builds. any build below that are not affected. all builds above it are, is that eliminate the "range"?

[Alin S]

(In reply to comment #11)
> http://trac.webkit.org/changeset/47546 does not seem related.  Maybe I'm not
> understanding your comment.  Do you know the exact revision which regressed, or
> just the nightly range?

i do not know exact revision, i know the nightly build number that i tested with in a sequential test of a bunch of builds and clicked "about safari"

it had an r before the number

[Alin S]

Excuse the flood of changes. I had revised it using an old title and old priority tag. i also changed it up to 10.6 since that is what i am using...

[Simon Fraser (smfr)]

I can't reproduce this with a nightly or TOT build. I am signed in to Flickr, and thumbnails look fine everywhere.

[Simon Fraser (smfr)]

I do see the issue where thumbnails are not clickable.

[Simon Fraser (smfr)]

Created attachment 39702 [details]
Reduction

The "position: relative" on photo-container seems to be the issue.

[Simon Fraser (smfr)]

This regressed with http://trac.webkit.org/changeset/47440

[Alin S]

(In reply to comment #17)
> Created an attachment (id=39702) [details]
> Reduction
> 
> The "position: relative" on photo-container seems to be the issue.

i think thats it, either way only the bottom edge is clickable on all of my macs, i even tried it at the apple store, guy was nice enough to let me throw webkit on a iMac there and it was busted there too on a brand new 23" imac

[Simon Fraser (smfr)]

Created attachment 39706 [details]
Simpler reduction

[Dave Hyatt]

I suspect this is more trouble from the RenderBox for the image not being positioned yet.  Although the y-position on the InlineBox is right, it's out of sync with the corresponding RenderBox.

I think the fix is to start moving RenderBoxes during the horizontal and vertical placement of the InlineBox (and not put it off to a separate pass).

[Alin S]

anyone want to verify that this seems to be working properly in r49550

[Alin S]

 Anyone want to close this bug, it works for me.