This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] fix compare_tests


On Wed, May 30, 2007 at 04:28:07PM -0700, Mike Stump wrote:
> On May 30, 2007, at 3:37 PM, Janis Johnson wrote:
> >I'm trying out Andrew's idea from the PR, adding the dump file  
> >suffix to the pass/fail line.  It makes the lines unique but has the  
> >disadvantage that it will cause compare_tests to fail for test  
> >results from before and after that testsuite change.  Is that  
> >acceptable?
> 
> Yes.  If you structure the line in a particular way, Geoff's  
> regression checker will even gloss over the fact.  I think the rule  
> would be, don't put the part that changes first on the line.

compare_tests notices any changes I make to the pass/fail line; do
Geoff's tools use compare_tests or something else?
 
> >An alternative is to use a slightly different regexp in one of the  
> >directives in gcc.dg/tree-ssa/pr17141-1.c and in forwprop-1.c, and  
> >to remember that for all future tests that might need to check for  
> >the same regular expression in two different dump files.
> 
> I think this sounds more like a work-around... if we didn't have the  
> above.  If we have the above, we don't need the work around.  Having  
> the framework produce uniqueness by itself is better than having a  
> hard to remember rule that one cannot have the same regexp (or  
> whatever).  If we didn't have the fix, I don't see anything wrong with  
> putting in the work-around.

Yes, it's definitely an ugly workaround.

Janis


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]