The newly built gnatmake tool crashes. It looks like this is because it is being passed the wrong length for a variable length array.
Note that there are a variety of known x86-64 abi issues. If llvm-gcc code is trying to call into gcc code, this could be a problem.
Is this fixed now? x86-64 abi should be mostly happy.
Currently the build fails because of PR2556 (at -O2 and -O0). I haven't been able to improve the Ada x86-64 situation because of this. Previously I was able to build at -O0. Looking at Ada test suite failures then indicated that indeed there were a bunch of ABI problems, eg: wrong handling of functions returning arrays (which I guess got forgotten about since C-like languages don't do that). I have local patches fixing some of them but even so it wasn't enough to get Ada building at -O2 at that time.
is this fixed or even worth keeping open if not?
I'm planning on fixing this kind of issue in dragonegg, but not in llvm-gcc.