[llvm-commits] [PATCH, PowerPC] Fix MC instruction encodings

[Hal Finkel]

----- Original Message -----
> From: "Rafael Espíndola" <rafael.espindola at gmail.com>
> To: "Hal Finkel" <hfinkel at anl.gov>
> Cc: "Ulrich Weigand" <Ulrich.Weigand at de.ibm.com>, "Commit Messages and Patches for LLVM" <llvm-commits at cs.uiuc.edu>
> Sent: Friday, November 16, 2012 2:46:30 PM
> Subject: Re: [llvm-commits] [PATCH, PowerPC] Fix MC instruction encodings
> 
> >> However, right now, there is *no* test whatsoever for any
> >> encoding,
> >> so IMHO it doesn't really make sense to block patches that fix
> >> obviously broken encodings until the infrastructure is in place.
> 
> Sorry, it is not obvious at all for someone that has not been working
> with ppc. Given that this is a case where LLVM has the necessary
> infrastructure for adding tests (MC), can we put a *requirement* that
> tests changing the encoding *must* include an assembly test?

I understand your viewpoint, but I think that this is a bad idea. As I've said in other threads, getting the necessary infrastructure in place for PPC is being actively worked on. Pulling people off of other tasks to assist with the implementation may not be practical or even helpful at this point.

Also, adding tests for these fixes by hand will prove to be a waste of time. As soon as the infrastructure is in place, I will automatically generate encoding tests from the test-suite binaries in such a way that these changes will automatically be covered. After that is complete, manually generated tests will be useful. Obviously, to some extent, I'm asking you to trust our judgment in this matter. All of the primary PPC contributors understand how the testing process is supposed to work, and we're working to make that process a reality.

Thanks again,
Hal

> 
> Cheers,
> Rafael
> 

-- 
Hal Finkel
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory