[llvm-commits] [llvm] r71013 - in /llvm/trunk: lib/Transforms/Scalar/LoopStrengthReduce.cpp test/Transforms/LoopStrengthReduce/2009-05-04-Overflow.ll

[Chris Lattner]

On May 6, 2009, at 10:37 AM, David Greene wrote:

>> while running: grep {cmpq       $-478,} change-compare- 
>> stride-1.ll.tmp
>> child process exited abnormally
>
> Hmm...Can't we at least give folks a chance to fix things first?
> Doing this 5 minutes after sending a notice to the commits list seems
> a bit excessive.

Having a patch be reverted should not be seen as a negative thing.   
There are many people pulling from mainline svn: if there is a problem  
with a patch, I think it makes sense to revert it immediately.

However, I disagree with Bill that "testing would have fixed this".   
There is no reasonable way to test every configuration before  
committing the code, and most patches don't have issues.  The flip  
side of this is that we should aggressively revert patches that *are*  
found to have minor issues.

-Chris