[llvm-commits] [llvm] r162034 - /llvm/trunk/include/llvm/Object/ELF.h

[Will Schmidt]

On Thu, 2012-08-16 at 15:06 -0400, Rafael Espíndola wrote:
> On 16 August 2012 14:33, Chandler Carruth <chandlerc at google.com> wrote:
> > Test cases?
> 
> We discussed this a bit on IRC. It is my impression that we already
> went too far with PPC ELF without testing. As Roman itself noticed two
> days ago, it is possible to break hello word on ppc elf and not notice
> it:

> 
> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20120813/148239.html
> 
> We have been too slow to add tests, even when we noticed that entire
> functions could be deleted and the tests would still pass:
> 
> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20111219/133931.html
> (which btw, is still the case on trunk).
> 
> So sorry for the current state for those starting to hack now on pcc
> elf, but to make sure we are in a better position tomorrow I think we
> have to start requiring tests as features are added and hopefully also
> cover the backlog we have so far.

Hi all, 
The changes so far are an attempt to get existing test cases to run.  So
I'd argue against the changes being considered new features.  :-)

2003-01-04-ArgumentBug.ll is the one chosen at random that Adhemerval
has been focused on.  That wasn't called out explicitly, and we do
expect some number of test cases to begin functioning better once this
initial batch of patches gets worked into the tree.

What got us in trouble here is that we didn't understand the
'test-suite' as being a separate batch of tests from 'make check', and
simply didn't know to run it.  (Until after Roman hit them and reported
the regressions to us).

Ok ? 


Thanks, 
-Will

> Cheers,
> Rafael
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>