On Thu, Jan 14, 2010 at 7:18 PM, Mike Stump <[email protected]> wrote:
> On Jan 14, 2010, at 6:43 PM, Chris Lattner wrote:
>> On Jan 14, 2010, at 6:06 PM, Mike Stump wrote:
>>
>>> Author: mrs
>>> Date: Thu Jan 14 20:06:42 2010
>>> New Revision: 93500
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=93500&view=rev
>>> Log:
>>> Refine unreachable warnings.  WIP.
>>
>> How does it refine it?
>
> We avoid issuing warnings for code that can be reached from other code that 
> itself is dead, and when we report on dead loops, we try and select a better 
> location to report.  Better means, as close to the top of the file as we can, 
> giving preference to the main source file over other files, and any file over 
> system headers.
>
>> Testcase?
>
> Checked one in r93503.
>
>> Also, the codesize for std::sort is quite high, is it needed?  Why are you 
>> sorting based on locations?
>
>
> When there are multiple dead locations inside a single function, the sort 
> helps ensure a more sane ordering to the produced diagnostics.  I didn't 
> think randomly jumping back and forth had quite the polish I was looking for. 
>  If the diagnostic subsystem buffered and selected an ordering, we wouldn't 
> have to buffer or sort them here.

Can it use array_pod_sort?

 - Daniel

> _______________________________________________
> cfe-commits mailing list
> [email protected]
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>

_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to