Hi Kamila. Thanks for the careful patch, rationale, test, and all. Overall, it seems like a fine idea to me, and I'm happy to accept it. I agree with the general idea that providing this option to people in a convenient way, vs. overriding various variables, etc., is worth it.
But before I actually install the changes, some minor stuff: - for completeness, may as well support dist-shar and dist-tarZ too, I think? Not that I expect anyone will ever actually care about them in conjunction with AM_OPTIONAL_AUTOMAKE, but it just seems cleaner to support everything automake does. (I think those are the only two missing from your current patch.) - I think the doc could more strongly make the point that, in practice, the dist-* lists in AM_INIT_AUTOMAKE and AM_OPTIONAL_AUTOMAKE should not have any overlap, since it would be pointless. - I think the doc should mention some of the rationale, i.e., images or systems where people don't want to install compression tools just for the sake of make distcheck. Etc. Also an @comment with a url reference to this bug number (81040). - Finally, as for shells that don't support command -v, it seems fine to me to just silently fail on them. If a user of such ever cares, which seems highly unlikely to me, they can report the problem. However, the doc should mention this. Inserting that shell code to reimplement command -v seems annoyingly verbose. Also, it would need to use test instead of [ ... ], and use the usual x"$1" instead of bare "$1". Also, its test for absolute filenames will not work for all Windows paths. All in all, I'd much rather skip that whole thing than have to go through all the irritating special cases of finding executables for the millionth times. BTW, I was surprised to find that at least some Solaris 10 /bin/sh *does* support command -v. portal.cfarm.net has one Solaris 10 system still hanging on, cfarm210.cfarm.net, which is where I tried it. Thanks again, Karl
