Re-shape after breaking when needed

[conor-93]

HarfRust exposes UNSAFE_TO_BREAK and UNSAFE_TO_CONCAT flags on GlyphInfo, though Parley doesn't use these. Parley shapes first, then breaks, and doesn't do any subsequent shaping.

For scripts with contextual shaping (where presentation of glyphs may depend on their neighbours), common in cursive scripts like Arabic, this can result in incorrect shaping.

[xStrom]

Could this also lead to the line length being longer, past max advance, and thus needing a recursive line break index choice?

If so, we should get a concerete test case for that scenario to make sure we handle it properly.

[behdad]

Could this also lead to the line length being longer, past max advance, and thus needing a recursive line break index choice?

In Pango what we do, if the line is justified, we adjust the word spaces after the reshape to align. But the for unjustified text is that it fits after the reshape.

[nicoburns]

But the for unjustified text is that it fits after the reshape.

Are you saying that unjustified text will always fit on the line after reshaping? That is surprising-to-me but good news if true.
Parley doesn't justify text until after breaking all lines anyway, so I don't think justification will be affected by this.

[behdad]

Are you saying that unjustified text will always fit on the line after reshaping?

Sorry. Broken sentence in my previous comment. For unjustified text the assumption is that it fits, but there's no guarantee.