[EMAIL PROTECTED] writes: > Package: autoconf > Version: 2.59a-3 > > Watch: > > ---begin--- > zircon$ mkdir test > zircon$ cd test > zircon$ cat >configure.in > AC_PREREQ(2.50) > AC_INIT(test,1) > AC_CHECK_FUNCS(wombat) > sinclude(conf2.in) > AC_OUTPUT() > zircon$ autoconf > zircon$ grep wombat configure > for ac_func in wombat > zircon$ grep baboon configure > zircon$ cat >conf2.in > AC_CHECK_FUNCS(baboon) > zircon$ autoconf > zircon$ grep baboon configure > zircon$ rm -rf autom4te.cache/ > zircon$ autoconf > zircon$ grep baboon configure > for ac_func in baboon > ---endit---
There are at least two different issues here. The first is simply that Autoconf renames sinclude to m4_sinclude via M4sugar. I don't know why bare sinclude works for you; it does not work for me. If I replace sinclude by m4_sinclude, I can reproduce your problem. When the first issue is cleared up, the second arises. This is what you are pointing out: m4_sinclude does not do dependency tracking. This is because, unless Autom4te actually records the dependencies of a traced file that existed at the time, it has to assume that a dependency that does not exist may be missing because the file that included it was changed to no longer include it. Thus, a missing dependency always triggers an update. This interacts badly with m4_sinclude, which doesn't yield an error if the dependency is missing. The way to fix this would be to make autom4te record which optional (m4_sinclude) dependencies were present or absent on the last run, so that it could update the target only if that state changed. Autoconf 2.59 doesn't have any such logic, and a look at Autoconf CVS shows that it hasn't been added there either. My advice would be to use m4_include instead of m4_sinclude if possible. Otherwise, if you'd like to implement the above behavior and submit a patch to me, I'd be happy to review it, include it in the Debian Autoconf, and submit to upstream. (I have little interest in implementing this behavior myself.) -- Ben Pfaff email: [EMAIL PROTECTED] web: http://benpfaff.org -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
