Behdad Esfahbod
Behdad Esfahbod
> One test fails though. Need to investigate. One of the failing tests is in Thai and we fail to zero mark advances, I suppose because there is no mark...
https://github.com/msys2/MINGW-packages/issues/11053
> [msys2/MINGW-packages#11053](https://github.com/msys2/MINGW-packages/issues/11053) Humm. But that's for ARM. No idea why our bots are failing.
@khaledhosny please advise if there's anything you think we should fix.
> turns out Uniscribe will zero the width of glyphs with class 0 GDEF glyph class when they are the result of applying `mset` feature but otherwise will not zero...
Oh. How about this: same issue I referenced was about similar problem with Hebrew. The solution we settled on is to ignore GPOS and do fallback positioning if GPOS doesn't...
This proof-of-concept fixes at least the first issue. Didn't try the second, but I guess it should fix both: ``` diff --git a/src/hb-ot-shape-complex-arabic.cc b/src/hb-ot-shape-complex-arabic.cc index 1f244f940..e0a5c19b0 100644 --- a/src/hb-ot-shape-complex-arabic.cc +++...
> Makes sense as a solution, I guess, if you can figure out a tidy way to do it that covers both scripts handled by the shaper. I can definitely...
> One thing I wish I had done different is to have a `cluster_offset` or `cluster_start` in all the `hb_buffer_add_*` APIs. They all start cluster at 0 at text start....
> Are multiple calls to `hb_buffer_add_*` without resetting the buffer possible right now? If no, then there is no practical change in behavior. It is possible.