[morx] MORX-32: Insertion At Mark Without SetMark

[brawer]

When processing a morx table, CoreText and HarfBuzz currently behave differently when inserting a glyph sequence at the current mark position and no mark has actually been set. Admittedly it’s a corner case, but you still might want to fix it. Compare these:

https://rawgit.com/unicode-org/text-rendering-tests/master/reports/FreeStack.html#MORX-32
https://rawgit.com/unicode-org/text-rendering-tests/master/reports/CoreText.html#MORX-32

[behdad]

Just fixed that a few minutes ago: #1195

[behdad]

Actually, the fix for this broke MORX-31 for me. If, not setting mark is to be same as mark being set at index 0, why does one not get more insertions for the first test in MORX-31? Ie. XXAYYAZZ should produce INSXXINSAYYAZZ.

[behdad]

I see CoreText has more problems. But yeah, I think the correct expectation should be changed, now that we want (as per MORX-32) an unset mark to be considered at 0. One way or another. We cannot have both. @nedley let me know which way you want to go.

[behdad]

@brawer Can we change test to current HarfBuzz behavior at least, until we agree with Apple on what to do? I personally think assuming mark=0 is not useful, but I need to choose one way or another. Other lookup assumes mark not set skipping mark action. We can do that here, then MORX-33 needs to be changed.

[brawer]

Personally I’d agree that assuming a default of mark=0 is not useful. Changed MORX-32 in test suite.