This is the mail archive of the 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: NetWare update

On Tue, Dec 21, 2004 at 10:50:17AM -0800, Janis Johnson wrote:
> On Tue, Dec 21, 2004 at 09:52:45AM -0800, Janis Johnson wrote:
> > On Tue, Dec 21, 2004 at 11:20:56AM +0100, Jan Beulich wrote:
> > > 
> > > To me it boils down to (as said before) compile_flags not being set at
> > > the time dg-skip-if (and from there check_conditional_xfail, where
> > > compile_flags are used) gets evaluated, as can be seen in the attached
> > > file, where "Compiler flags are: " is followed by nothing, which of
> > > course doesn't match against "* * *".
> > 
> > I've apparently done all of my testing passing -m32 or -m64 to all tests
> > so I hadn't seen the failure of dg-skip-if when there are no options.
> > I'll look into that to see if there's a fix that doesn't involve procs
> > within DejaGnu.  In the meantime, for that particular test try moving
> > the dg-option directive before dg-skip-if.
> Ouch, dg-skip-if is looking at the compiler flags for the previous test
> that was compiled (not skipped), or an empty string if this is the first
> test that is compiled.  It's magically doing the expected thing for
> tests that check for flags that apply to all tests, such as those set
> by RUNTESTFLAGS, for all but the first test.

Sorry for talking to myself here.

I'm testing a patch to dg-skip-if that will compare each of the flags
from that directive to the multilib flags plus either options that were
specified in a previous dg-options directive in the same test or else
the default options for this set of tests.  Are there other flags I
should look at?


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