Re: [Plplot-devel] Aspirin for autoheadache

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

On Thu, 30 Jan 2003, Rafael Laboissiere wrote:

> The promised patch is attached below.  It is not yet fully test, but at
> least bootstrap.sh works (both autoconf and autoheader).  I am also
> attaching the new include/plConfig.h.in file, but its contents may change.
> acconfig.h must be deleted from and include/plConfig.h.in must be added to
> the CVS repository.
>
> Here is a summary of the changes:
>
>     * In configure.ac, AM_CONFIG_HEADER is replaced by AC_CONFIG_HEADERS.
>       config.h is now the first element in the arguments list.
>
>     * All AC_DEFINE and AC_DEFINE_UNQUOTED macros in configure.ac and
>       sysloc.in have now three arguments, as mandated by the autoheader docs.
>
>     * Added at the top of plConfig.h:
>
>          #if HAVE_CONFIG_H
>          #  include <config.h>
>          #endif
>
>     * The strange call touch plDevs.h in bootstrap.sh has been removed.
>
>
> Could you please check if this patch is okay?

The patch applies cleanly on top of my recent changes to sysloc.in and
configure.ac.  I do get one error message, however, from bootstrap.sh.

configure.ac:770: required file ./config.h.in' not found

Also, no config.h.in (or config.h) is produced.

The workaround was to put touch config.h.in in bootstrap.sh.  Once that was
done, there were absolutely no messages from bootstrap.sh which is a first
for me, and quite encouraging.

I then went on with configure; make; make install, and all was well with
plplot-test.sh (produced identical postscript to 5.2.0).

So I am committing it because it is no worse than before, and I like the fact
that all those stupid warnings are gone.  BTW, have a look at
Rafael's extensive comments in the include/plConfig.in.h log for the AM-LT
branch.  (That file existed on that branch for a while before it was deleted
for that branch.  Apparently when I used cvs add on the current file, the
state information is okay since I can retrieve it from CVS without problems.)
Rafael's cvs log comments are very similar to recent threads explaining why
the bizarre touch command has to be used in ./bootstrap.sh, etc.

Now for the bad news.  The current commit does not solve the fundamental
problem that started all this thread:

Right in the middle of my initial make:

cd .. && /bin/sh /home/software/plplot_cvs/HEAD/plplot/missing --run autoheader touch ./plDevs.h.in

UGH!

Part of what is going on is the CVS date of ./plDevs.h.in is fixed in
December.  So this touch occurs when you first make, and you are okay until
the next time you do a CVS update (which I believe restores the old date).
plConfig.h.in is currently okay because the CVS date is today.  But if
configure.ac is updated again, then make will assume plConfig.h.in is
out of date and will run autoheader on it as well.

Note, I am no longer using the Debian autotools which have been patched
extensively by the Debian packager to allow a combination of old and new
autoconf stuff.  Instead, in the interests of making sure we all have the
same autotools versions, I have been using the latest stable releases of
everything, (autoconf-2.57, automake-1.7.2, and libtool-1.4.3) which I
downloaded from FSF and built from scratch.

I think we have clearly run up against a bug in autoconf since autoheader is
not supposed to be run on those files at all.  (It is only supposed to be
used to generate files from scratch such as config.h.in, not modify existing
files, and besides these are the second and third files in the list for
AC_CONFIG_HEADERS, and the documentation says autoheader should only be
invoked for the first file.)

Rafael, in your first post you had a suggestion for supressing autoheader
altogether.  Perhaps it is time to try that as a workaround for the bug. But
I am amazed (a) this problem hasn't bitten most of the many projects that
are using autoconf and cvs together, and (b) the resulting outcry didn't
generate strong motivation for a quick fix.

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
__________________________