This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [_Complex test 1/4] support_complex_tests in gdb.exp and pass _Complex args to func
> Date: Thu, 19 May 2011 15:22:32 +0200 (CEST)
> From: Mark Kettenis <mark.kettenis@xs4all.nl>
>
> > Date: Thu, 19 May 2011 21:01:26 +0800
> > From: Yao Qi <yao@codesourcery.com>
> >
> > On 05/19/2011 04:14 PM, Mark Kettenis wrote:
> > > The other thing is that the testsuite should not be "conservative".
> > > We should run as many tests as possible on as many platforms as
> > > possible. Tests that are known to fail should be KFAILed (if there is
> > > a known bug/issue with GDB itself) or XFAILed (if there is an issue
> > > with the platform). Only if running the tests causes serious problems
> > > (crashing the OS, causing excessive timeouts) we should consider
> > > skipping them.
> >
> > No. I don't want to break running non-complex-type tests in
> > funcargs.exp or varargs.exp on platforms that don't support _Complex
> > after this patch.
>
> Ok, I see what you're trying to avoid now. In that case you should
> first build the test binary without defining TEST_COMPLEX, run the
> existing tests, rebuild the binary with TEST_COMPLEX and then run the
> new _Complex tests. Or simply seperate the _Complex tests into files
> of their own.
Of course the feature-based test that Joseph suggested is fine as well.