[tex-k] Autoconf-2.13 vs. 2.59

Olaf Weber olaf at infovore.xs4all.nl
Sun Jan 30 11:57:37 CET 2005

Karl Berry writes:

>> Step 3: get a hook into autoconf-2.59 that allows to insert the ac_include
>> ...
>> For step 3 above I'd like to request an enhancement (on the autoconf
>> discussion list) but before doing so I'd like to learn your reactions:

> My reaction is (as usual): very nice!  

> I seem to remember discussing it with djm and others back when I first
> wrote the ac_include stuff, and they were not at all interested.
> However, that was trying to get ac_include itself supported.  It was far
> too fragile to be an official feature.

> So we can hope that Akim, Alexandre, and crew will accept your much more
> generic patch -- although I'll be surprised if there is not a lot of
> comment and (perhaps) resistance.

One thing to note is that I've played with building an implementation
using the $extrasub that is part of the sed pipeline in the current
version.  (The comment preceding it says "Shell code in configure.ac
might set extrasub.  FIXME: do we really want to maintain this
feature?")  The problem I ran into when doing this is that the
following variables are not substituted until after that part of the


The other workaround I've found is to use a two-step process: put
@common_mk@ in the Makefile.in, and have a file common.in that is to
be inserted, then put this in configure.ac:


You do have to be rather careful about the order in which you do
things, or so I've been told by someone who uses the same workaround.

Personally, I think we'd be best off if there was a version of
AC_SUBST_FILE that did the substitutions itself, instead of the
current situation or the more generic hook Peter proposes.  The reason
I'm not that fond of the "hook" approach is that we end up inserting
non-trivial code in what is already a complicated sed expression.  It
too long ago since I did this to remember all the details, but while
working on trying to make the 'extrasub' code sufficienly robust I
ended up using/having to use a helper script that emits the actual sed
script to be used.  (But this may have been due to limited knowlegde
of sed.)

Olaf Weber

               (This space left blank for technical reasons.)

More information about the tex-k mailing list