Re: [Plplot-devel] Make of plplot-5.2.0 fails on a freshly installed Red-Hat-8.0.

[Alan W. I. <ir...@be...>]

On Sun, 26 Jan 2003, Rafael Laboissiere wrote:

> This call to autoheader (through the `missing' script) should never, never,
> I repeat: _NEVER_ happen when regular users run make.  None of the Autotools
> should be called during a normal make, since this would defeat the very
> purpose of the Autotools.  The only tools that are required for
> configure/make are a Bourne shell and a regular make (not even GNU make).

I agree; that is the whole design philosophy behind autotools.

>
> Automake, for instance, should only be called from the Makefile when
> developpers change one of the autotools source files (like aclocal.m4,
> configure.ac, etc).  Something is really weird with the 5.2.0 distribution
> tarball.
>
> As I told before, I do not have time to plunge into this problem, but here
> is a first try on its diagnosis:

Thanks, Rafael, for starting the ball rolling here.  You inspired me to find
what I think is the correct solution (see end).

>
> When I run ./configure here on a freshly detarized 5.2.0 distribution, I get
> the following rules in include/Makefile:
>
> $(srcdir)/plConfig.h.in:  $(top_srcdir)/configure.ac $(ACLOCAL_M4) $(top_srcdir)/acconfig.h
>         cd $(top_srcdir) && $(AUTOHEADER)
>         touch $(srcdir)/plConfig.h.in
>
> $(srcdir)/plDevs.h.in:  $(top_srcdir)/configure.ac $(ACLOCAL_M4) $(top_srcdir)/acconfig.h
>         cd $(top_srcdir) && $(AUTOHEADER)
>         touch $(srcdir)/plDevs.h.in
>
> This is a complete non-sense, since there are no src/plConfig.h.in and
> src/plDevs.h.in to be built!  This is what is triggering the call to
> autoheader from make, which should never, I repeat, never occurs for regular
> users.

Actually, this makes sense.  Remember, $(srcdir) is the current directory,
i.e., include.  So these two rules say run $(AUTOHEADER) if plConfig.h.in or
plDevs.h.in in the include directory are out of date relative to
$(ACLOCAL_M4) (defined to be $(top_srcdir)/aclocal.m4 in the top-level
Makefile) and configure.ac or acconfig.h in the top level directory.  So
if bootstrap.sh (which invokes autoheader) updates plConfig.h.in and
plDevs.h.in, then the above
rules will only be run if the user fiddles with configure.ac,
acconfig.h, or aclocal.m4 in the top-level directory.

However, further investigation reveals bootstrap (and autoheader inside
it) did not do its job:

tar ztvf ../plplot-5.2.0.tar.gz | grep include |grep '\.in$'
-rw-r--r-- software/software    4635 2003-01-19 22:00:30 plplot-5.2.0/include/plConfig.h.in
-rw-r--r-- software/software    1056 2002-12-03 00:39:26 plplot-5.2.0/include/plDevs.h.in
-rw-r--r-- software/software   15875 2003-01-19 22:00:27 plplot-5.2.0/include/Makefile.in

plConfig.h.in and Makefile.in were properly updated, but plDevs.h.in has an
old date!  It turns out this December 3rd date is older than the relevant
top-level files so the $(AUTOHEADER) rule is run (in error), and I believe
this is the source of the trouble that Joao found.

How to fix it?

I think the problem here is that plDevs.h.in is under CVS control (as it
should be) so a file exists already and this confuses autoheader.  In fact
autoheader should *never* be invoked on this file.

Thus, I believe the complete solution is simply to do the following changes
in configure.ac:

AM_CONFIG_HEADER(include/plConfig.h include/plDevs.h) ==>
AM_CONFIG_HEADER(include/plConfig.h)

        src/Makefile
	include/Makefile

==>
	src/Makefile
	include/Makefile
	include/plDevs.h

i.e, Remove the generation of plDevs.h.in from autoheader control and put
the generation of plDevs.h under autoconf control.

Note on Joao's systems and on the RH 7.3 system available to me
the complaints were always about plDevs.h.in, and nothing was said about
plConfig.h.in which I think autoheader is handling without any problems
(since that file is not under CVS control there is nothing in the exported
tree from CVS there before ./bootstrap.sh creates it with autoheader).

Maurice and Rafael, do you agree with this analysis?

Alan

__________________________
Alan W. Irwin
email: ir...@be...
phone: 250-727-2902

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________