Bug 6375 – SourceAlpha do not display properly until window is redrawn (resized)

• Status: RESOLVED
• Resolution: FIXED
• Component: SVG

[Nefaur Khandker]

Although I'm not sure which part of the document fails to display properly (it
may be that certain types of nodes of the document), upon first loading
firefox-logo.svg in either Safari or DrawTest (using the latest version of
WebKit), only a portion of the document is initially drawn. The portion that is
drawn is consistently the same portion of the document.

I have noticed several actions that then cause the document to be displayed in
its entirety:
1) Resizing the Safari or DrawTest window
2) Minimizing the Safari or DrawTest window, and then reactivating the window.
(Please note that while the window is in the dock, the SVG document is still not
displayed correctly, and it that it is only displayed properly when the window
is reactivated.)
3) In Safari, clicking the Back button, and then the Forward button (although
clicking the Reload button causes the document to again display incorrectly)
4) In DrawTest, clicking on a toolbar button

[Nefaur Khandker]

Created attachment 5486 [details]
Firefox logo in SVG format

[Nefaur Khandker]

Created attachment 5487 [details]
A screenshot of Safari incorrectly displaying the SVG document

[Nefaur Khandker]

Created attachment 5488 [details]
A screenshot of Safari correctly displaying the SVG document

[Eric Seidel (no email)]

Something is not being invalidated properly.  Possibly related to the gradients.  We definitely used to 
render this perfectly.

[Eric Seidel (no email)]

This has to do with gradients failing to render the first pass through.  Not sure why yet.  This affects lots of SVGs.

[Joost de Valk (AlthA)]

Adding Regression keyword.

[Joost de Valk (AlthA)]

Upping to P1 because this is a regression.

[Eric Seidel (no email)]

This is a fundemental rendering issue.  Without a fix for this we would not ship an SVG enabled WebKit.  Adding SVGHitList keyword and changing to p2.

[Eric Seidel (no email)]

This is a regression from previous versions of WebKit+SVG, but not from a previous shipping version of Safari, hence why I've removed the regression keyword and set only to p2.

[Eric Seidel (no email)]

*** Bug 6874 has been marked as a duplicate of this bug. ***

[Alexander Kellett]

Created attachment 6037 [details]
testcase showing an issue with sourcealpha and feMerge

fixing the attached testcase also appears to have fixed this issue

[Alexander Kellett]

Created attachment 6040 [details]
layouttests patch

[Alexander Kellett]

Created attachment 6041 [details]
patch fixing this issue

[Eric Seidel (no email)]

Thanks for another great fix.

[Eric Seidel (no email)]

Comment on attachment 6040 [details]
layouttests patch

Looks good.  I tweaked a bit when landing.

[Eric Seidel (no email)]

Comment on attachment 6041 [details]
patch fixing this issue

Looks great.