Anyone have any remaining thoughts?

In D65609#1613721, @EricWF wrote:

wow. This is a fantastic amount of work. Thank you. No longer will libc++ accidentally break the pretty printers!

• Run a pep8 formatter.
• Address other comments from code review.

From the failure mode I was seeing, any value at all for VERSION will be fine, as long as it is a valid preprocessor string.

The revert is at https://reviews.llvm.org/D64687, which I will commit shortly if there are no objections.

This revision breaks python 2.7.16 builds, which are still supported by upstream python for a few more months. I'm preparing a revert.

After this revision lib/Transforms/Scalar/IndVarSimplify.cpp:2630 defines "Latch" which is only used inside an assert. This makes builds with -DNDEBUG fail with unused variable warnings.

Committed as r365713.

Remove other forward declaration also in the added header.

This revision violates some rules about the build structure.

I am absolutely not the right person to review the math. I think we need to trust that the author has it correct. The logic looks reasonable and the tests look basically thorough. I am inclined to accept it.

The number is the argument count plus one for return value. So this function should be named

I can't speak to how useful this would be and for whom, but I'm not opposed to including it if Fuschia has a use for it.

Any of the previous reviewers have a comment before I commit?

• Make LIBUNWIND_WEAK_PTHREAD a formal Cmake option.

I've now put the weak-reference version behind a non-default build option. Is this acceptable?

• Add an option to make pthread symbol references weak.

Hi Nick--hope your vacation was good.

In D60285#1461209, @ldionne wrote:

@kledzik

Nick, what do you think about adding weak symbols just for achieving this?

ping?

This is better than a no-op RWMutex because then the same binary works both when the application is multi-threaded and when it isn't. Therefore the build doesn't have to distinguish between the two cases.

Thanks. I've hidden a couple more symbols Please take another look.

Make cpu_model a hidden symbol, to match libgcc.
Also cpu_features2 and __cpu_indicator_init.

• Also hide __cpu_inicator_init for similar reasons.

I have now switched it to use --unwindlib=[libgcc|libunwind]

• Add --unwindlib=[libgcc|compiler-rt] to parallel --rtlib= [take 2]
• Change option argument --unwindlib= from "compiler-rt" to "libunwind".
• Fix tests syntax and driver checks for --unwindlib=

• Undo cpu model change.

This looks correct to me, thanks for the fix.

rsmith: Do you have any comments on the naming here?

This was rolled back with r352524. I'll work out something better.

Sorry for the breakage everyone, I'm preparing a rollback now.

• Correct help text for --unwindlib options.

Committed with r350900.

Committed as r347614.

Add a test case.