[polly] r219275 - [Fix] Ignore forwarding alias sets in the alias set tracker.

Tobias Grosser tobias at grosser.es
Fri Oct 10 08:01:52 PDT 2014


On 08.10.2014 22:01, Johannes Doerfert wrote:
> On 10/08, Tobias Grosser wrote:
>> On 08/10/2014 19:23, Johannes Doerfert wrote:
>>> On 10/08, Tobias Grosser wrote:
>>>> On 08/10/2014 04:23, Johannes Doerfert wrote:
[...]

>>>> Also, was there a reason you did not commit a test case? I attached you
>>>> the one I used, in case you did not manage to reduce one.
>>> The test case for this is so unstable, I couldn't reproduce it with an
>>> llvm from yesterday.
>>
>> Interesting. It consistently failed on the LNT servers for several builds
>> and I could also reproduce it on my machine. I wonder why it is not
>> reproducible for you.
> Within the llvm changes of one day I could reproduce it but not before,
> I don't know why and I don't understand the AST good enough to guess.
>
>> If you have a reason why you do/can not want to commit a certain test case,
>> sharing this information in the commit message seems useful.
> I missed a test case and to properly explain the change, if you want you
> can rectify my mistake and add a test case that might introduce
> forwarding sets.

No need to correct a mistake. I would just propose for the next time to 
put a short comment in case you can for some reason not provide a test 
case. This shows people you did not forget about it, but had a reason 
why it was not possible/useful.

People can then still comment on this reason, but in most cases the 
reason is obvious to most people so we avoid the need for discussions. 
If the reason is not obvious, people may be able to hint on how to 
create a useful test case.

Cheers,
Tobias




More information about the llvm-commits mailing list