Eli Zaretskii <e...@gnu.org> writes: > Hmm... no, this doesn't solve the problem. The expansion of AC_LIBOBJ > for waitpid is gone from the configure script, but the value of > LIBOBJS in libiberty/Makefile still includes waitpid.o. What else is > related to this?
After re-reading the sources a bit, I come to the following conclusions... * $funcs is a list of functions libiberty should provide if the host doesn't have them. * We can override what the host *has* but not what it *shouldn't* have. Since (or "if") nobody will (should) use waitpid() on mingw anyway, and since libiberty really wants to include waitpid.o, how difficult would it be to use some #ifdefs to have waitpid() just return an error on mingw? That at least gets past the mingw build problem. > One caveat: I needed to hack config/override.m4 to allow me to run > autoconf 2.69 I have installed, because otherwise it insists on > autoconf 2.64 which I don't have. I hope this isn't the reason for > the incomplete solution. I have many versions of autoconf installed, each in their own directories, and add the right one to my $PATH on a per-project basis. Autoconf works just fine that way, and there have been plenty of cases of autoconf output changing in, er, "unexpected" ways across autoconf releases. If you regen configure and an "svn diff" (or git diff) shows unexpected changes, check your autoconf :-) Same for automake and autogen.