If a font is used only for vertical metrics/layout, the FT2 backend fails to ensure mFUnitsPerDevUnit is set appropriately, and this can lead to totally broken shaping -- glyph advances return zero, for example, so everything overprints. I can reproduce the brokenness with a trunk build on Linux by loading a page such as http://mongolfont.com/test/firefox/div.html and then pressing Ctrl-+ to zoom in. The initial rendering of the page looks fine, but after it zooms, the Chinese rendering is borked, although the Japanese looks fine (I think this has something to do with the fact that the Chinese text ends up using several font faces on my configuration). Also, in a debug build, the assertion "mFUnitsConvFactor not valid" fires (lots).

I think this fixes things here; will flag for review once I've tested a bit more.

Comment on attachment 8545517 [details] [diff] [review] Ensure mFUnitsConvFactor is set up properly for vertical fonts in the FT2 backend This also fixes a couple of weird reftest failures I've been seeing on B2G tests, which were presumably due to the messed-up metrics.

Comment on attachment 8545517 [details] [diff] [review] Ensure mFUnitsConvFactor is set up properly for vertical fonts in the FT2 backend Review of attachment 8545517 [details] [diff] [review]: ----------------------------------------------------------------- (In reply to Jonathan Kew (:jfkthame) from comment #0)> > I can reproduce the brokenness with a trunk build on Linux by loading a page > such as http://mongolfont.com/test/firefox/div.html and then pressing Ctrl-+ > to zoom in. The initial rendering of the page looks fine, but after it > zooms, the Chinese rendering is borked, although the Japanese looks fine (I > think this has something to do with the fact that the Chinese text ends up > using several font faces on my configuration). Also, in a debug build, the > assertion "mFUnitsConvFactor not valid" fires (lots). FTR, on my system both the Chinese and the Japanese are borked before the patch and fine after it.