Some gaps occur on stretchy characters. There hit particlarly Linux, or newish high resolution screens on Windows. I will attach a patch to mitigate these pixel/roundoff issues. It does the trick for me on Windows on Linux.

Comment on attachment 194978 [details] [diff] [review] patch @@ -1602,7 +1603,9 @@ // if we are not a largeop in display mode, return if size fits - if (targetSize <= 0 || (!largeop && IsSizeOK(charSize, targetSize, aStretchHint))) { + if ((targetSize <= 0) || + (!largeop && ((isVertical && charSize >= targetSize) || + IsSizeOK(charSize, targetSize, aStretchHint)))) { Ops... that bit from bug 306004 crept in here. It is not relevant here.

You might want to add a comment that this is a bit of a hack and in the new Gfx world maybe we can fix this "for real".

It appeared from testing that some of the pieces been glued together have curved end points (with the tip of the curves possibly rounded off). It is likely that we will still have to jam these pieces to account for such curves. That's why I used the non-fixed rulethickness (instead of a fixed amount of pixels) to jam enough for the curves that are scaling.

okay. yuck.

Checked in.

Comment on attachment 194978 [details] [diff] [review] patch A similar beautification as bug 306543, but this one is specific to stretchy MathML characters.

Checked in the 1.8 branch.

This checkin caused bug 321994

Apparently not. I have track that bug down to --enable-pango (unless this checkin awoke something else in pango. you never know...).