On Tue, Aug 17, 2004 at 10:50:29AM +0900, Stephen J. Turnbull wrote:
> >>>>> "Jan" == Jan Harkes <jaharkes_at_cs.cmu.edu> writes:
>     Jan> Why is the automake cache a problem?
> 
> Auto* is essentially a regression test for configuration.  If
> something changes, you need to run the tests again, all of them,
> because auto* provides no dependency information for the cache.  In
> practice, bootstrap.sh failed occasionally for me on Linux, often on
> the Mac.  Always silently, too.  Since getting rid of autom4te.cache,
> no problems.
> 
> There's also an issue of standards conformance.  I don't know what the
> Coda standard is, but in the GNU Makefile standard, the distclean
> target should remove all files generated by process of building the
> distribution.

Right, but what we distribute is whatever is in the tree after
"cvs checkout ; bootstrap.sh". So this would include the autom4te.cache
files as well. The other tarballs (lwp/rpc2/rvm) use 'maintainerclean'
as the target that removes autogenerated files.

I guess I will add rm -rf autom4te.cache to the beginning of the
bootstrap.sh file, it just seems like a better place for it.

Jan