Re: Fix distcheck for Autoconf

[Benoit Sigoure]

On Nov 6, 2007, at 8:43 AM, Ralf Wildenhues wrote:

* Benoit Sigoure wrote on Tue, Nov 06, 2007 at 06:17:19AM CET:

On Nov 6, 2007, at 5:20 AM, Ralf Wildenhues wrote:

AC_CONFIG_FILES should not live in the source tree and should not be
packaged.

Our manual even states:

| hierarchy: the test suite ought to be shipped!

| Be sure to distribute @file{package.m4} and to put it into the source

What's wrong with having package.m4.in in your source tree and in
AC_CONFIG_FILES and letting configure (well, actually config.status)
generate package.m4?

recreate the testsuite (which depends on package.m4) using autom4te.

have.

For the Autoconf package itself, that is not so much of a problem: it

of Autoconf.

testsuite should be shipped (so autom4te is not needed), which in turn

depend on files that are not distributed, because it will otherwise

for your own package.

Then config.status generates package.m4 (unconditionally!), so make must But autom4te is a developer tool, which we assume the end-user does not ships autom4te. But for every other package, you just now need Autoconf installed on the end-user's system. That is one of the non-design- goals means all its prerequisites should be shipped. And none of those should break builds from read-only source trees. distcheck should tell you so

--
Benoit Sigoure aka Tsuna
EPITA Research and Development Laboratory

Yes I know all this, I just did not realize that since package.m4 is generated, it'd trigger a rebuild of testsuite. That's indeed a problem. Is there really no way of, say, distributing the package.m4 produced by config.status? Overall this is not very important, it's just that I like this solution better because the Makefile.am does not need to contain this rule and its hacks, and this looks like a perfect job for config.status.

PGP.sig
Description: This is a digitally signed message part