Closed Bug 379606 Opened 16 years ago Closed 16 years ago

rect-with-rx-or-ry-01.svg fails on OS X


(Core :: SVG, defect)

Not set





(Reporter: jwatt, Assigned: jwatt)





(1 file, 1 obsolete file)

The reftest rect-with-rx-or-ry-01.svg fails on the tinderbox "MacOSX Darwin 8.8.4 qm-xserve01 dep unit test". The 'data:' base64 encoded images it outputs are identical (luser got gimp to compare them) so it's a mystery to me why it would register as a failure. For the build/reftest log see:

I'm not sure who to ask for help on this.
Attached image rect-with-rx-or-ry-01.svg (obsolete) —
Attachment #263608 - Attachment is obsolete: true
> are identical (luser got gimp to compare them)

How, exactly?  I haven't found a 100% good way to do it...

In any case, the data: URIs are for sure different.  Is there antialiasing stuff going on here that matters?  Or subpixel positioning?
I don't know how luser did the comparison. CC'ing him so he can speak for himself. :-)
Assignee: general → jwatt
I fixed the reftest failure by removing the assumption that the antialiased edge of a square rect with rx and ry equal to width/2 is exactly the same as a circle with radius width/2.
Closed: 16 years ago
Resolution: --- → FIXED
I just pasted them into the same document as layers and set the top layer to "subtract".
Yes, but how did you make sure all pixels were black and not something like rgb(0,0,1) for 3 pixels and black for the rest?
I checked the histogram, everything was zero valued.  Not sure if the subtract operation is really what I wanted, or how that handles alpha though.
Yeah, that's what I've been using.  I hoped there was something better. The problem with the histogram is that for an 600x800 image (the real one is larger, actually), a 1-px difference in color from black to rgb(1,0,0) will give a mean red channel of about 0.000002.  And the histogram output is only to one decimal place...
I'm using GIMPshop on Windows here, and the Histogram window has a 'Count' field, so you can see the exact number of pixels.
You need to log in before you can comment on or make changes to this bug.