Ralf Wildenhues wrote: > Let's use a specific example: > > AC_INIT([Pkg], [1.0], [bug], [pkg], [website]) > > Do you think that 'autoconf-2.63 -Wall -Werror' should have errored out > for the website argument?
Of course it should not and it would not error out. 2.63 is history, so there's no way a hypothetical feature we're discussing now to be available there (unless, of course, an even more hypothetical "energetic" distro maintainer manages to backport it, which presumes it is implemented at least in an autoconf git branch (or at least a public patch is available), and in which case it is a distro issue actually). > m4_version_prereq([2.64], > [AC_INIT([Pkg], [1.0], [bug], [pkg], [website])], > [AC_INIT([Pkg], [1.0], [bug], [pkg])]) Well, my extremely humble opinion is that those who use approaches like this deserve what they get. I don't think that my feature request encourages this use at all. But it's quite possible that I miss something (or many things). Going back to Eric's previous example: > That is, we intentionally document that AC_RUN_IFELSE([arg1]) and > AC_RUN_IFELSE([arg1], [], [], []) are identical, therefore > AC_RUN_IFELSE must not print a warning if $# is less than 4. Yes, and assuming my feature request is implemented, 2.68 should warn if you pass [arg5], while < 2.68 must silently discard it, the way it behaved historically. (2.68 and above should silently discard the 5th argument too if -Werror is not in use, I hope that's clear.) If the 5th argument is actually required, the package should use AC_PREREQ providing a version of AC_RUN_IFELSE that accepts (IOW: does not ignore) its 5th argument. Probably it's hard to implement this, and I'm not insisting at all. Apparently there's enough food for thought.
