This change was accepted and landed. (Github link: https://github.com/llvm/llvm-project/commit/a0b65a7bcd6065688189b3d678c42ed6af9603db#diff-16c8c6eb85e05438f5d6c60ff9869072a3a3b1618aa1481ac7a0cb049f06f51d )

fix phrasing in commit message

fixed phrasing in comment

fix the headerguards and add a comment explaining the undef in FuchsiaTest.h

add headerguards to the new Test.h and FuchsiaTest.h

add the functions to the spec in the correct places,
and add them to the entrypoints list for aarch64

LGTM

fix the previous patch, the macro was placed on the wrong line.

Add the macro definition for fuchsia (or others) to hook into for extra attributes on the libc functions

submit comment

update the macro in common.h to force the decltypes to use the definition in the __llvm_libc namespace

update fmaf.cpp to be in line with the rest of the functions.

fix the alias name in common.h.def

accepted with nits

LGTM

final cleanup

submit comments

added more tests

remove references to targetos

after talking with Siva about it it looks like we're going to stick with this for the moment and not go with lit tests. The reason is that currently the wrappergen unit tests about as dependent on LLVM support as wrappergen itself, adding lit tests would make portability potentially harder.

add testapi.td so that the tests are no longer platform dependent and clean up the code a little.

submit comments

address most of the comments

I'm not quite sure why a lit test would be better in this situation, does it have better support for testing complete binaries?

remove commented out include

submit comments

This change finishes the work started in the previous one, all of the tests now pass.
Big thank you to abrachet for the comments.

submit comments

(NOT QUITE WORKING) get the rest of the llvm dependencies out StreamWrapper and FDReader

I guess I've got some more changes ahead of me. This new commit (linked here) is where I plan on putting the next set, and it currently has a few based on what you've said.

switch the tests to using the global headers to match the sources they're testing.

update: I was reminded of the other patch that I'm currently in the process of getting reviewed (linked here: https://reviews.llvm.org/D91394) that will make the string and math tests independent of headergen since they can use the system libc's header. This means if I can make the unit tests fully independent then you can bring up both the string and math parts of LLVM-libc without the whole project.

Do our tests depend on the generated headers? I thought they did, if so it will be very hard to remove our dependency on the rest of llvm.

reset the #includes for everything not in src/string and src/math

lgtm with nits

LGTM

submit messages (I really gotta remember to mark these before I run the arc command)

make error message more helpful

submit comments

add check for a void argument as well as other arguments in the same function.

submit comments

change continue to break

added header guard to llvmlibc_errno.h, and a comment on assert.h explaining the lack of header guard (thank you to abrachet).

fix missing newline at the end of a file.

move the errno macro to its own file to match assert.

update the assert files to also match

remove unnecessary comment

fix commit message locally

Fixed nits.

Minor formatting fixes.

Fix formatting and move data path to PublicAPICommand

Remove the redundant new parser