Incorrect table background rendering for some zoom levels

[automatedbugreportingfacility]

pdf.js version: 0.7.339

1. Open http://users.ece.gatech.edu/~sudha/academic/class/ece4100-6100/Lectures/Module13-MultiCore/multicore-introduction.pdf
2. The table on page 5 is unreadable.

[gigaherz]

Console log:

[08:48:35.926] PDF d92cfb8dd7ceab2b37579055eaeb515b [1.5 Acrobat Distiller 9.0.0 (Windows) / PScript5.dll Version 5.2.2] (PDF.js: 0.7.339)
[08:48:36.260] Warning: TODO: TilingType: 1
[08:48:37.487] Warning: TODO: TilingType: 1

[gigaherz]

I don't know enough about that warning to say if it's related to the actual issue, but comparing with Adobe Reader it appears to be a transparency issue.

[automatedbugreportingfacility]

There's some progress in version 0.8.229, it's mostly readable now, but it still looks wrong.

[SamyCookie]

The PDF file is not available anymore. Could you reupload it ?

[THausherr]

multicore-introduction.pdf
from
https://web.archive.org/web/20130822053356/users.ece.gatech.edu/~sudha/academic/class/ece4100-6100/Lectures/Module13-MultiCore/multicore-introduction.pdf

[SamyCookie]

Thanks @THausherr

No issue for me anymore with Firefox 66 and PDF.js 2.2.107

[timvandermeij]

Still doesn't work for me with Firefox 66 and PDF.js 2.2.107 on Arch Linux. The table on page 5 still contains cells with different colors and patterns.

[SamyCookie]

Oh so sorry, how I could miss that ? I confirm the bad rendering on my Windows 10.

[THausherr]

But that's more a scaling problem, it goes away when zooming. It looks bad at 100%, but gets better at 150%, and really nice at 200%. (I'm looking at the actual page 5 labelled "11", the "Fundamental Trends" table)

(I used Edge and Chrome, not FF)

[THausherr]

Now I tried with FF and there the tiles look identical at 220%. However FF runs with an "old" monitor, while Edge and Chrome runs with a 4K monitor.

[automatedbugreportingfacility]

Fixed by #16117, just 3 days short of this bug's 10-year anniversary! However, the fix works around a d2d bug and is scheduled to be removed once the bug is fixed, but here it is noted that this occurred on Arch Linux as well, so the comment in the code about it being a "Windows specific bug" is probably not the full story, and I'm not sure if a revert would be good. Also, it'd probably be a good idea to add this pdf as a regression test to avoid breaking it in the future. @calixteman

[calixteman]

@timvandermeij could you check that the patch in #16117 really fixed the issue for you ?

[timvandermeij]

@calixteman Interestingly, it did (I also didn't expect this).

The current rendering for me (Arch Linux, Firefox 113.0.1) is:

This matches the rendering of e.g. Okular and seems correct. However, if I do git revert 1617ee6c3f45aa061300482844ec2bd0526dc2e8 to revert the patch from #16117 the rendering is broken as I remembered it was originally:

It looks like the suggestion from @automatedbugreportingfacility is correct here: it doesn't seem to be Windows-specific, so we should change the comment there and therefore can't easily revert this later on, and adding page 5 of this PDF to the test suite might be a good idea to avoid a regression.