[PATCH] D51571: [Xray] Darwin providing defined value for weak symbols to fix linkage issue

[Dean Michael Berris via Phabricator via llvm-commits]

dberris added a subscriber: aizatsky.
dberris added a comment.

I understand that there’s a problem here, but the question is whether is
this the right solution.

First, do the tests even run for macOS/Darwin in compiler-rt? I don’t think
they do as we’re still running them on Linux hosts (see lit configuration
for the test cases).

Second, these symbols ought to be generated by the linker if the right
sections are in the MachO binary (they are array extents). Does the LLVM
backend even generate the instrumentation map for Darwin yet? Maybe that
change has to happen first?

Lastly, is there a way to reproduce these? Is there a buildbot failing? Can
you share a link to the failure? When did it start and can we revert the
change that cause it while we investigate what the proper solutions are to
making XRay work on Darwin?


Repository:
  rCRT Compiler Runtime

https://reviews.llvm.org/D51571