Build errors on Clang 3.7 because of new move diagnostics

[dean0x7d]

Trying to compile with Clang 3.7 results in the following errors because of -Werror and the new move warnings which are included with -Wall in the 3.7 release.

~/pyston_deps/llvm-trunk/include/llvm/ADT/APFloat.h:352:12: 
error: redundant move in return statement [-Werror,-Wredundant-move]
    return std::move(Value);
           ^

../../src/codegen/compvars.cpp:2548:16:
error: moving a local object in a return statement prevents copy elision [-Werror,-Wpessimizing-move]
        return std::move(elts);
               ^

Note that the first error already has a fix in LLVM.

[sYnfo]

👍 I'm bumping into the same issue.

[undingen]

Could you please test if #1019 fixes the issue if you have time or if there are additional warnings which need to get fixed.
Thanks

[sYnfo]

Unless I'm doing something wrong the patch doesn't seem to help:

https://gist.github.com/sYnfo/826bad71a20360700a3f

[undingen]

Ok I updated the patch to include fixes for the warnings in your log. Could you please try again?
If more problems show up I will install 3.7 :-D

[dean0x7d]

The patch does help: it clears the LLVM specific warnings.
But there are a few other -Wredundant-move warnings not related to LLVM and a large number of new -Wshift-negative-value messages show up now.

I've been looking into this and I can submit a pull request with the clang 3.7 fixes a bit later today.

[undingen]

I don't know if you have seen it but I replied a few seconds before your last replay :-D to let you know that I updated the patch to fix the remaining warnings (hope that those where the last ones..).

[dean0x7d]

Ah, I missed your previous post. But there is actually a lot more. I have the fixes on my fork.
There are 2 fixes for -Wredundant-move: 3205005
And 10 for -Wshift-negative-value: 60644ab

That takes care of all the warnings, but then something truly weird happens. The compiler crashes :-/
As far as I can tell the crash happens on these two lines with delete obj; here and here. If I comment out those two lines, the build completes successfully.

I'm not sure if it's just something with my system or a genuine bug in Clang 3.7.

[kmod]

Ok I set up a container with clang-3.7 and I'll take a look :)

[kmod]

I'm running into the same clang bug -- I think this is a legit compiler bug so I filed it on the clang tracker. I think I found a workaround, and the build is succeeding in my container, so I opened #1020. The tests aren't all passing though, but I'm guessing it's because I did this in a 15.10 container; I haven't looked into the test failures much (io_test is failing, set comparisons aren't being allowed in some places CPython is letting them, and stack_limits is failing).