"make dist" calls "automake" ??

[Guido Draheim]

I did continue my checks with an update to the autotools series,
now using autoconf-2.52/libtool-1.4b/automake-1.4-p5

Apart from that other combination, the call to automake happens
to be encoded directly into the "distdir" target, reading


distdir: $(DISTFILES)
        -rm -rf $(distdir)
        mkdir $(distdir)
        -chmod 777 $(distdir)
        here=`cd $(top_builddir) && pwd`; \
        top_distdir=`cd $(distdir) && pwd`; \
        distdir=`cd $(distdir) && pwd`; \
        cd $(top_srcdir) \
          && $(AUTOMAKE) --include-deps --build-dir=$$here 
--srcdir-name=$(top_srcdir) --output-dir=$$t
        [...]

why is that? the original Makefile.in is not good enough for distribution?
Is it supposed to have "make dist" be run only within a sane
development environment with the same autotools available?

I'd like to call it a severe feature flaw, as this scheme can easily 
be used to have developers work out all their stuff right, followed
by happily building a tarball that differs from the build sources
he was using. 

possible solutions?
Hmm, adding a check to "missing" that it looks at the version number of
the tool in question, "e.g. missing --run --version 1.4-p5 automake ..."
and missing would then call "automake --version" *before* trying
to run it - and it would only be run, if the versions are identical.
A bit hard to implement though, but the best to have atleast
those tools be run that are there, and otherwise ignored. This is
a good thing for implicit autoheader atleast.

The harsh way might be to have all calls to autotools be wrapped with
a check about maintainer-mode. Therefore, if the make-dependencies
work out a way to rule that wants to update maintainer-files then
it would not call the autotool but issue a warning that it should
only be done in maintainer-mode. I remember a time that most of the 
code that wants to run autotools had been prefixed with an ifdef-check
(@maintainer_mode_true@) - why not for these too?

-- guido                               http://guidod.4t.com
31:GCS/E/S/P C++$++++ ULHS L++w- N++@  d(+-) s+a- y++ 5++X-