On Mon, 8 Oct 2001, Chris Torek wrote:

> I never touch the stuff, but, it appears to me that you have
> not found a problem with BSD-make; rather, you have found a(nother
> of the millions of them :-) ) problem with GNU autoconf.

I've runs into several of those already.  Fortunately I've been able to
work around them, including this one once you guys pointed me in the right
direction.  I haven't found any viable alternatives, though.  Whatever
scheme Joerg Schilling (author of cdrecord) uses seems to work, but it's
more than a bit impenetrable.

> Try "make -n clean" or "make -n distclean" to see what
> command(s) would have been run.  Commands "hidden" with @foo
> that exit with error code 1 are the only way the make would
> fail, and "make -n" will show you what they would be, even
> when "make" would run them silently.
> 
> This sort of thing is why I diligently avoid GNU autoconf,

The bug was in automake.

I made a fresh install of autoconf and automake from pkgsrc to a
non-development box here.  The diffs between Makefile.in generated by
automake 1.4 and 1.4-p5 seems to tell the story.


diff Makefile.in-zaphod Makefile.in-arthur 
1c1
< # Makefile.in generated automatically by automake 1.4 from Makefile.am
---
> # Makefile.in generated automatically by automake 1.4-p5 from Makefile.am
131c131
<         test "$$subdir" = "." && dot_seen=yes; \
---
>         test "$$subdir" != "." || dot_seen=yes; \
230c230
<           cp -pr $$/$$file $(distdir)/$$file; \
---
>           cp -pr $$d/$$file $(distdir)/$$file; \



-- 
David Griffith
dgriffi@cs.csubak.edu