I discovered that our logic to prevent splitting harfbuzz clusters during wrapping only worked if the run in consideration was at the beginning of the shaped text. A Gio fuzz test for rune accounting was failing as a result. I've fixed the logic and added a new test case to exercise the non-zero-offset cluster splitting case.
I discovered that our logic to prevent splitting harfbuzz clusters during wrapping only worked if the run in consideration was at the beginning of the shaped text. A Gio fuzz test for rune accounting was failing as a result. I've fixed the logic and added a new test case to exercise the non-zero-offset cluster splitting case.
Signed-off-by: Chris Waldon christopher.waldon.dev@gmail.com