Date: Sun, 27 Feb 2000 03:44:53 -0800 (PST)
From: Martin Buchholz <[EMAIL PROTECTED]>
Users who aren't totally paranoid are doing the wrong thing. Running
"make distclean" isn't paranoid enough - it involves trusting the
maintainer. The thing most likely to work, and the only thing I do
when building Other People's Packages
I have not yet found a problem with `make distclean' in a package
which uses automake.
rm -rf foo-1.9
tar xzf foo-1.9.tar.gz
cd foo-1.9
configure
make
make install
For a package which uses autoconf, I always configure in a separate
directory:
tar xzf foo-1.9.tar.gz
mkdir objdir
cd objdir
../foo-1.9/configure
make
make install
But well-written packages should aspire to help even non-paranoid
users. Good packages provide a `distclean' target. I also think good
packages should be written in such a way that reconfiguring forces a
rebuild of all architecture-dependent files. In fact, the Makefiles
are likely to be generated by configure, and to be really safe, object
files should be rebuilt, so maybe *.o should depend on the Makefile
that builds them.
I agree that autoconf should provide some way to detect whether the
user ran configure again, and I agree that that should normally force
a rebuild of all files.
I agree that this is a level of quality that most packages won't
achieve - there are more important things in the free software world
to fix. But it would be a good public service if someone decided to
fix up all the GNU packages purely from the point of view of good
configure/Makefile hygiene. AFAIK, no one has ever decided to do
global code maintenance across all GNU packages.
Isn't that what automake is for?
Ian
