Behdad Esfahbod

> makes sense. Did you or someone else verify that this workaround still works in CoreText and doesn't break anything else? No I haven't.

What's the conclusion here?

Or just do the right thing and collect all referenced user nameIDs and subset to those.

As Erik suggested, varLib and MutatorMath use the same algorithm, it's just that MutatorMath is picking a different base master here. varLib is designed to produce OpenType-compatible deltas. We can...

> Isn't it just the choice of default that causes the problem? Yeah but in a font we want the Regular to stay the default.

You are correct that sparse GPOS merging is not in the design of the merger. That would be a much harder problem.

> If varLib had absolutely _no_ provisions for sparseness then this might fall into one big "probably won't work" bucket, even with no warning. > > But that isn't really...

My problem is not in contrib. Though, trying now, I get this different error: Linking CXX executable glsl_compiler libglsl_optimizer.a(glsl_parser_extras.cpp.o): In function `do_common_optimization(exec_list*, bool, bool, unsigned int, gl_shader_compiler_options const*)': /home/behdad/src/github/glsl-optimizer/src/glsl/glsl_parser_extras.cpp:1523: undefined...

Here's a sample backtrace from glsl_test: behdad:glsl-optimizer 130 (master)$ gdb --args ./glsl_test tests/ GNU gdb (GDB) 7.6-gg23 Copyright (C) 2013 Free Software Foundation, Inc. License GPLv3+: GNU GPL version 3...

> HarfBuzz has merged support for shaping these fonts. Now we just need to figure out what to do (assuming we even need to do anything) with the cmaps on...