* Joel E. Denny wrote on Fri, Jan 28, 2011 at 08:31:05PM CET: > On Fri, 28 Jan 2011, Ralf Wildenhues wrote: > > * Joel E. Denny wrote on Fri, Jan 28, 2011 at 08:06:38PM CET: > > > A way to get the best of both worlds might be to permit shell functions > > > containing AT_CHECK to be defined outside of AT_SETUP. I believe that is > > > not possible with autotest now, and I have no idea if it's feasible to > > > implement. > > > > That is possible right now, either in atlocal.in or in PREPARE_TESTS > > diversions: > That's interesting, and I wasn't aware of it. However, when I tried > expanding AT_CHECK inside that function, autotest gave the usual > complaint: > > tmp.at:4: error: AT_CHECK: missing AT_SETUP detected Ah, I think it doesn't work to expand AT_CHECK inside a function. Sorry for overlooking that requirement of yours. It might be possible to rework _AT_CHECK so that it will work in such a setup, should mostly be a matter of assigning some values that are currently held in m4 macros only to shell variables in AT_SETUP, and then referencing them in _AT_CHECK. This would probably enlarge testsuite code a bit, but maybe negligible. Cheers, Ralf _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
