Re: Can't `make install' Autoconf 2.53b if makeinfo is missing + an OSF make issue

[Alexandre Duret-Lutz]

>>> "Paul" == Paul Eggert <address@hidden> writes:

 >> From: Alexandre Duret-Lutz <address@hidden>
 >> 
 >> 
 >> The touch error is easy to explain.  The following line in
 >> Autoconf's doc/Makefile.am
 >> MAKEINFO = @MAKEINFO@ --no-split
 >> causes missing to think makeinfo will create the `--no-split'
 >> file.  So it runs `touch --no-split' and touch complains.

 Paul> Are you sure about that diagnosis?

No, thanks for correcting me.  I wrote this mail in a hurry :)

[...]

 Paul> This runs plain "touch", with no arguments.

Agreed.  Anyway, this seems to be a general issue with missing.
This script doesn't understand the entire syntax of all the
command it emulates.  It can be fooled when passed additional
options (like --no-split here).  The trouble is that such error
are usually found by users, not developers (who have all the
tools).

What could be done?

It would be nice if there was a way to test a package in
"missing" mode.  The idea is to make sure that missing properly
emulate all command used by the package.  I've no idea how this
can be done.

Maybe Automake should always use `-o' when a command support it.

 >> Why makeinfo is run is more fun.  It seems that because of the
 >> `.texi:' rule,


 Paul> To fix this problem, how about if we remove the .texi:
 Paul> rule from Automake?  The .texi: rule is obsolete.
 Paul> Nowadays people should be creating FOO.info from
 Paul> FOO.texi, and they should not be creating plain FOO
 Paul> without an extension.

I have many files in my /usr/share/info/ directory without
`.info' part.  Actually I think Automake should simply output
this rule if it is needed (i.e. if the user don't use `.info').

Automake could _warn_ about `.info'-less info files if you say
this is obsolete.

BTW, this problem would be solved as a side effect of two
pending patches I don't want to apply now (I'd like to make some
architecture changes in Automake first, and this will probably
wait until 1.7 is out).
http://mail.gnu.org/pipermail/automake-patches/2002-June/000858.html
http://mail.gnu.org/pipermail/automake-patches/2002-June/000867.html
-- 
Alexandre Duret-Lutz