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]

Re: Testcase for -fssa bug


On Fri, May 19, 2000 at 12:30:37PM -0400, Kaveh R. Ghazi wrote:
>  > From: Jeffrey A Law <law@cygnus.com>
>  > 
>  >   In message <u8snvey37i.fsf@gromit.rhein-neckar.de>you write:
>  >   > 
>  >   > As discussed with Richard and Ulrich on gcc-bugs, here's a testcase to
>  >   > show a bug with -fssa.
>  >   > 
>  >   > Can I commit this?
>  >   > 
>  >   > Andreas
>  >   > 
>  >   > 2000-05-19  Andreas Jaeger  <aj@suse.de>
>  >   > 
>  >   > 	* gcc.dg/ssa-1.c: Testcase for SSA bug from Ulrich Drepper.
>  > 
>  > Actually, I think I'd prefer the test in c-torture/execute and that
>  > we change the c-torture framework to test -fssa.
>  > 
>  > Other opinions?
>  > jeff
> 
> Agreed.
> 
> What I don't like about the dg framework is that it only runs tests
> through one -O# level.  I've seen many cases where individual test
> cases originally were meant for one -O# level but years later
> triggered at some other one.  We lose this sanity check in the dg
> framework.

Not cycling through -O levels can be desirable, too.  I put
preprocessor tests in gcc.dg partially because it doesn't waste time
compiling the same chunk of code over and over again.  These are tests
where optimization can't possibly make a difference, the only thing we
care about is whether the preprocessor mangled the source.  It also
means you get one failure report instead of six.

zw

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