Behdad Esfahbod
Behdad Esfahbod
> it's just that the magic number feels a bit arbitrary and may come back biting us again in the future. If the font is so big that Python ends...
What you sketch is correct indeed. But I have in the past read "_By default, the first ItemVariationData structure (index 0) will be used. The vsindex operator can be used...
cc @PeterCon @drott @dfrg @rsheeter
Indeed we want this. It's just that it violates how `hb_shape_justify` is currently implemented. I'll see what I can do. Sorry about the wait.
Yeah this is a bit unfortunate. The HB guarantees go beyond just monotonicity. The indices are use to index into string memories and out-of-range numbers can do Bad Things™.
> it fails the shaping. It had not occurred to me. Maybe it's best to do that in trivially bad cases, and just fix the monotonity if needed. Also, maybe...
> We have received reports of covscan run on harfbuzz code. Below are some of the reports for which we need upstream help to analyze them and either fix it...
No more actionable items here I'm afraid. Closing.
@LaurenzV Sorry to ask here. I read a comment from you saying one of the layout optimizations in HB is producing different result from the non-optimization path. Can you please...
> The kerning lookup has several class kerning subtables, and the kerning for `V` is split across two of them. Other implementations apply the `V` kerning in the first of...