User-Agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.9a1) Gecko/20051002 Firefox/1.6a1 Build Identifier: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.9a1) Gecko/20051002 Firefox/1.6a1 Deer Park a2 on MacOSX has some trouble with stretchy font rendering. Larger root gryphes are not displayed in the "Torture Test" #13 and so are the over/underbraces ("glue" part in particular) of #19 and #22. Zooming in makes these rendered correctly. This problem was not encountered in Firefox 1.5b (with fontEncoding.properties copied, ref bug #228804 comment #7). I put the screenshots at http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/screenshots.html Further, I edited mathfontMath2.properties and mathfontMath4.properties to change the "glue" part of these gryphes to "not relevant to the char" (\uFFFD) like \u221A = \uF8FF\uFFFD\uF8FF\uFFFD\u221A\uF8FF\uF8FF\uF8FF # Sqrt, radic (the forth of the right hand side is changed) and got preferable rendering, see the screenshot page. I have no idea what the point in dissociating the glue parts which *become* rendered quite fine. Is this specific to Mac? Should this tweak be incorporated as the fix for this problem, or should we fix the code to avoid similar paradoxes in, say, STIX migration? Reproducible: Always Steps to Reproduce: 1. Download the nightly build of Firefox 2. Go to http://www.mozilla.org/projects/mathml/demo/texvsmml.xhtml 3. See the tests #13, #19 and #22 Actual Results: Larger root gryphes disappears, overbraces and underbraces are shattered (no "glue" parts) Expected Results: Correct rendering like the sample images in the left column

Firefox 1.5b2 is with this problem (and the same procedure works for it) too. > I have no idea what the point in dissociating... Having that said, I now suspect the canges made to mozilla/layout/mathml/base/src/nsMathMLChar.cpp (between v1.109 and v1.110 "Prevent gaps in stretchy chars") might have someting to do with this. THe glue gryphes from Mathematica fonts look very small in the encoding table, they might be cut down to zero-size and causing troubles... So that zooming in makes this problem dissapear. I don't know what fix to do so can anybody check this?

This is a regression from bug 307157. I should have been more careful to avoid turning the glue into nothingness. This patch makes sure to clamp the amount by which the glue is pretended to be smaller than it is. The glue in Mathematica fonts can really be tiny: http://www.mozilla.org/projects/mathml/fonts/encoding/math4.gif

Comment on attachment 200059 [details] [diff] [review] patch please don't request approval until you've got sufficient reviews, landed and been verified on the trunk. Thanks.

Checked in the trunk.

Comment on attachment 200059 [details] [diff] [review] patch too late for non-critical bugs in this release.

Comment on attachment 200059 [details] [diff] [review] patch drivers, I am re-requesting approval because it is a _critical_ fix. Without it, many of the achievements that have gone into getting MathML on the Mac are annihilated. As you might recollect from my earlier requests, MathML is all about appearance. Without this fix, there are ugly "holes" in strecthy MathML characters, leading to formulas that don't make sense -- we get meaningless renderings such as these: http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/16fntEncdCopied.gif http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/16fntEncdCopied2.gif instead of these correct ones: http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/16fntEncdCopiedZoom.gif http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/16fntEncdCopiedZoom2.gif

Checked in the 1.8 branch.