Closed acowley closed 7 years ago
Here is where the error I saw with the memory sanitizer seems to originate, which is why I did the OS conditional bit.
I can do the formatting changes, but the larger point here is the API expansion and I have no idea how to write a proof of a necessity for that.
EDIT: Here's an idea: would you be willing to expose the Internal
modules the way packages like text
and bytestring
do? Then more things could live in separate packages.
I have a few commits here, and I can break this into multiple PRs if that would be preferred. Each is pretty distinct and comes with a log message describing the change.
The last two commits involve changes to tests to account for varying OS support within LLVM itself, and a Haskell re-entrancy issue with GHC-8.0.1 on OS X that doesn't seem to impact @cocreature, so I have noted a guess that maybe this is another platform-dependent issue. I haven't been able to resolve that, but segfaulting as it does prevents the test suite from completing, so I am selectively disabling that test for the time being.