Bug 16664 – translated image invalidates wrong repaint rect

• Component: SVG
• Status: RESOLVED FIXED

[Terrance Davis]

When a PNG is loaded and immediately translated along the x-axis the result is transparency where the image moves away from it's initial location. There appears to be a box of transparency to the right of the initial location that does not extend far down the window.

I have put together a page to show the behavior:

http://genedavis.com/od&d/js/dojo-release-1.0.2/dojox/gfx/tests/test_image1_modified.html

The logo that is loaded and translated is cut off, but still there, just transparent. The behavior is corrected if user interaction with the window causes an new translation or manipulation of the image.  

my system:

Safari 3
Mac OS X 10.4
WebKit r29022
Dojo Toolkit 1.0.2 (using silverlight?)

[David Kilzer (:ddkilzer)]

Confirmed with a local debug build of WebKit r29032 with Safari 3.0.4 (523.12.2) on Mac OS X 10.4.11 (8S165).

Looks like a repaint issue as using "Force Repaint" from the Debug menu will make the rest (right half) of the cut-off image appear.  (Cmd-A, or select-all, does the same thing.)

[Eric Seidel (no email)]

Yeah, we simply invalidate the wrong rect.  We'll need to make a simpler test case.  I expect this will just end up being a dup of bug 16015

[Eric Seidel (no email)]

To see the wrong rect being invalidated, I used Quartz Debug.  Unfortunately that's slightly painful with an example this large (since you watch a bunch of other painting first)

[Eric Seidel (no email)]

Real websites are hitting this.  Adding to SVGHitlist

[Cameron McCormack (:heycam)]

A smaller test case: http://mcc.id.au/temp/2008/image-translate.html

[Dirk Schulze]

(In reply to comment #5)
> A smaller test case: http://mcc.id.au/temp/2008/image-translate.html

I see the same result in Firefox. Is this bug fixed? We might need another test case.

[Simon Fraser (smfr)]

Created attachment 54304 [details]
Testcase

[Dirk Schulze]

(In reply to comment #7)
> Created an attachment (id=54304) [details]
> Testcase

Sadly the image is not reacheable, but I tested it localy. I could see a gray rect with a stroke, and the image is displayed on the right position.

Doesn't the initial bug discribe a problem with translation? This test doesn't translate the image.

[Dirk Schulze]

It works with trunk and we did a lot of rewrite on the repaintRect-code. I bet this fixed for a long time now. Closing the bug. If someone finds another example, feel free to reopen the bug.