D31018: Handle boundry condition missed by r297682

Mon Mar 20 14:28:23 PDT 2017

On 3/18/2017 3:42 AM, Artyom Skrobov wrote:
> Yes, I had to revert r297820 because this test case still caused infinite compilation time on the TK1 build bots (but none of the others): http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20170313/437811.html
> I suspected a problem specific to their host compiler; these build bots are using Clang 3.8.0 (tags/RELEASE_380/final) from Ubuntu packages.
> I've now rebuilt LLVM with my patch on an ARM host, using http://releases.llvm.org/3.8.0/clang+llvm-3.8.0-armv7a-linux-gnueabihf.tar.xz (same tags/RELEASE_380/final) as the host compiler; but I'm still unable to reproduce that hang :-|
>
> Do you have any suggestions as to how I could debug it?

Are you configuring it the same way?  See 
https://github.com/llvm-mirror/zorg/blob/5c695726b8e25a461a581158d207e4d781232fc4/buildbot/osuosl/master/config/builders.py#L257 
.

Strictly speaking, your patch has undefined behavior (negating INT_MIN); 
maybe try committing a version without the undefined behavior?

-Eli

-- 
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project