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: Testsuite patch for Solaris 10


On Mon, Nov 29, 2004 at 12:31:26PM -0500, Daniel Jacobowitz wrote:
> On Mon, Nov 22, 2004 at 01:06:38PM -0800, Janis Johnson wrote:
> > On Fri, Nov 12, 2004 at 07:14:05PM -0500, Daniel Jacobowitz wrote:
> > > On Fri, Nov 12, 2004 at 04:07:05PM -0800, Janis Johnson wrote:
> > > > 
> > > >   { dg-do run [dg-target-list target
> > > >        [dg-and [dg-or [istarget i?86-*-*] [istarget powerpc-*-linux]]
> > > >                [is-effective-target ilp32]]] }
> > > > 
> > > > Would it be acceptable to do things like this within tests?
> > > 
> > > Just my two cents... If I'm reading what you're trying to accomplish in
> > > that example correctly, I'd rather write it like this:
> > > 
> > > { dg-do run { target i?86-*-* powerpc-*-linux } }
> > > { dg-require ilp32 }
> > > 
> > > Yes, it doesn't have half the expressive flexibility.  When we run into
> > > this as a real problem, then we can define additional predicates which
> > > do whatever; we have that option.  I don't think we need to have
> > > boolean logic embedded in tests.
> > 
> > Please see http://gcc.gnu.org/ml/gcc-patches/2004-11/msg01507.html
> > ([RFC] allow effective-target keywords in dg- directive target lists)
> > for a description of a different method of specifying an effective
> > target class in xfail and target lists.  As you may have noticed, I
> > REALLY want a way to do this.  I'd like to know if this new approach
> > is reasonable given the goal of being able to replace the test
> > framework; I'd rather not hear that it's inappropriate _after_ adding
> > the functionality and using it in tests, but so far that RFC has been
> > ignored.
> 
> I think that's a very nice idea.  How would you use it for the above,
> i.e. in target list and not 64-bit?

Only by adding an effective-target keyword that includes all of the
targets when used for 64-bit code generation.  Not optimal but it works,
and without requiring the ability to evaluate logical expressions.

Janis


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