[PATCH] Fix crash in handling back-end warnings

[Duncan P. N. Exon Smith]

LGTM.  Thanks both!

> On 2015 May 8, at 11:48, Diego Novillo <dnovillo at google.com> wrote:
> 
> Sure. If Duncan is also OK with the patch, I'll take care of it.
> 
> 
> Diego.
> 
> On Fri, May 8, 2015 at 8:47 AM, Geoff Berry <gberry at codeaurora.org> wrote:
>> 
>> No problem.  Could either of you commit this for me?
>> 
>> Thanks,
>> 
>> --
>> Geoff Berry
>> Employee of Qualcomm Innovation Center, Inc.
>> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
>> 
>> -----Original Message-----
>> From: Diego Novillo [mailto:dnovillo at google.com]
>> Sent: Friday, May 08, 2015 11:42 AM
>> To: Duncan P. N. Exon Smith; gberry at codeaurora.org
>> Cc: reviews+D9449+public+904931d5e5bc6215 at reviews.llvm.org; cfe commits
>> Subject: Re: [PATCH] Fix crash in handling back-end warnings
>> 
>> On Fri, May 8, 2015 at 8:28 AM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
>> 
>>> I probably broke this at some point when I was transitioning to the
>>> new debug info hierarchy.  The old hierarchy's accessors checked for
>>> NULL and give default answers, whereas the new hierarchy just uses
>>> pointers directly.
>>> 
>>> At some point I switched the accessors to start asserting and fixed
>>> any bugs that uncovered, but I guess there wasn't any test coverage
>>> here so I missed it.
>> 
>> Yeah. The backend tests that we have for diagnostics all include debug info and if you go via the FE, it always makes sure that debuglocs are always generated.
>> 
>> Geoff, your patch LGTM.  Thanks for the fix!
>> 
>> 
>> Diego.
>>