[HarfBuzz] harfbuzz: Branch 'master'
behdad at kemper.freedesktop.org
Sun Mar 5 21:51:21 UTC 2017
src/hb-ot-layout-gsubgpos-private.hh | 24 +++++++++++++++++++++++-
1 file changed, 23 insertions(+), 1 deletion(-)
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Sun Mar 5 13:51:01 2017 -0800
Add TODO item
diff --git a/src/hb-ot-layout-gsubgpos-private.hh b/src/hb-ot-layout-gsubgpos-private.hh
index 244c395..005ed21 100644
@@ -996,7 +996,29 @@ static inline bool apply_lookup (hb_apply_context_t *c,
- /* Recursed lookup changed buffer len. Adjust. */
+ /* Recursed lookup changed buffer len. Adjust.
+ * TODO:
+ * Right now, if buffer length increased by n, we assume n new glyphs
+ * were added right after the current position, and if buffer length
+ * was decreased by n, we assume n match positions after the current
+ * one where removed. The former (buffer length increased) case is
+ * fine, but the decrease case can be improved in at least two ways,
+ * both of which are significant:
+ * - If recursed-to lookup is MultipleSubst and buffer length
+ * decreased, then it's current match position that was deleted,
+ * NOT the one after it.
+ * - If buffer length was decreased by n, it does not necessarily
+ * mean that n match positions where removed, as there might
+ * have been marks and default-ignorables in the sequence. We
+ * should instead drop match positions between current-position
+ * and current-position + n instead.
+ * It should be possible to construct tests for both of these cases.
end += delta;
if (end < int (match_positions[idx]))
More information about the HarfBuzz