In the following testcase the width of the 'svg' element is given a percentage value of 100%. When the window is resized the width of the 'svg' element should resize with it. Load the testcase with a narrow window, then widen it before moving the mouse around. You'll see the width stays the same as it was at the time the document loaded.

I just realised the testcase needs my fix for bug 272630 to work properly. You can get the idea though if you place the left edge of the browser close to the left edge of the screen.

When we call IsPercentageAwareChild on the outer svg frame, we have the following code: 1146 const nsStylePosition* pos = aFrame->GetStylePosition(); 1147 1148 if (eStyleUnit_Percent == pos->mWidth.GetUnit() and gdb says the following: (gdb) p pos->mWidth $23 = {mUnit = eStyleUnit_Auto, mValue = {mInt = 0, mFloat = 0}} So the frame is claiming to not have a percentage width, which means the block code feels no need to reflow it when the available width changes. Sounds to me like SVG should be mapping the width attribute into style...

Although the testcase fails with firefox 1.5, it does seem to work as expected with a current trunk build.

Hmm.... Yes, it does. I wonder why. When did that change?

It was fixed on trunk between 1.9a1_2006021306 and 1.9a1_2006021319. http://bonsai.mozilla.org/cvsquery.cgi?module=PhoenixTinderbox&date=explicit&mindate=2006-02-13+05%3A00&maxdate=2006-02-13+20%3A00

Could be bug 326725.... who knows. Worksforme, I guess, though someone should file a followup bug on fixing the code to actually be correct.

Seems to be broken on current trunk.

Adding width/height to the nsSVGSVGElement attribute map does fix this problem, but causes gecko to complain about the testcase in bug 369850.

This bug was just fixed by the fix for bug 294086. The reftest layout/reftests/svg/sizing/dynamic--inline-resize-window-width.xhtml (and others) checked in for bug 294086 cover this bug too.