On Saturday 21 August 2010, Ralf Wildenhues wrote: > * Stefano Lattarini wrote on Thu, Aug 19, 2010 at 03:05:20PM CEST: > > --- a/m4/init.m4 > > +++ b/m4/init.m4 > > > > @@ -70,10 +70,22 @@ _AM_IF_OPTION([no-define],, > > > > # Some tools Automake needs. > > AC_REQUIRE([AM_SANITY_CHECK])dnl > > AC_REQUIRE([AC_ARG_PROGRAM])dnl > > > > +dnl We pass the the $AUTOCONF and $AUTOM4TE commands in the > > environment +dnl of automake and aclocal calls in the generated > > Makefiles, so wrapping +dnl them with the `missing' script would > > defintely be a bad idea. > typo defintely Fixed.
> > +dnl For example, if aclocal proper is ever called, it will need > > a working +dnl autom4te to get traces from e.g. configure.ac, > > and in such a case the +dnl workarounds provided by `missing > > --run autom4te' are not enough; a +dnl flat-out aclocal failure > > is the best and most correct option. > > Hmm, but a slightly version-skewed autom4te for tracing should > usually be ok. The tracing algorithm hasn't been very > version-sensitive in the past. Maybe, but if `missing' detects a versioning error, it goes on with its workarounds (touching output files etc.), which are no good for what aclocal needs. So better leave aclocal fail, and `missing' detecting the failure of aclocal, no? And BTW, the new behaviour of rebuild rules is consistent with the old behaviour in the case where $AUTOCONF and $AUTOM4TE are not overridden at configure time... Regards, Stefano
