I didn't do a test of removing a symbol, but it looks like you've sorted that. was that to test it didn't fall back to the library when using the tbd stub?
in my "rebuild the world" testing it seems like git and gcc are both a bit flaky. git sometimes hangs on a credential helper and gcc sometimes fails a self test with "file not found"
LnL has quit [Quit: exit 1]
this may be a dumb question, but would generating our own tabi stubs reduce our dependence on Apple's open-sourcing cycle? (and make supporting code that depends on recent features more feasible?)
don't think so. they're equivalent to having .dylib files. we'd still have to get headers from somewhere