* Eric Blake wrote on Mon, Jul 13, 2009 at 04:10:22PM CEST: > Paolo Bonzini <bonzini <at> gnu.org> writes: > > > > Part of me thinks it might be nice to change AT_XFAIL_IF to use similar > > > semantics, but then the other part worries about backwards compatibility > > > > I think it's not trivial either to change it. I doubt it makes a > > difference in > > practice, there are not many Autotest users and even fewer AT_XFAIL_IF > > users. > > Then let's leave AT_XFAIL_IF alone until (unless?) someone complains about it > being different in practice.
I think I can remember a couple of occasions where this particular semantics of AT_XFAIL_IF was either unexpected or undesired; at least within the Libtool testsuite it has caused some churn before. However, I also think that changing the semantics is a problematic thing to do, as it very likely breaks compatibility. I am not sure if it is better to document the current semantics, and maybe provide another macro that updates xfail status at the point it is invoked. Cheers, Ralf
