Re: AC_CONFIG_HEADERS vs. AC_CONFIG_FILES

[Ralf Corsepius]

On Mon, 2005-12-19 at 20:11 +0000, Harlan Stenn wrote:
> > On Mon, 2005-12-19 at 00:15 +0000, Harlan Stenn wrote:
> > > The problem with doing these things in configure is that one must rerun
> > > configure to regenerate the file.
> > > 
> > > Sometimes it is better do produce these things in config.status.
> > 
> > There is no need to do so, in this case, because the file is being
> > generated when running configure.
> 
> For certain definitions of "need", I agree with you.
> 
> > "config.status --recheck", runs configure, so changes to configure.ac
> > automatically get propagated to the generated file.
> 
> For packages that are "big enough" (ie, that have a large/slow run of
> "configure"), this is an unnacceptable solution.

1. This approach generates some sort of autoheader (Actually an
exported, generated (auto-) header). The only situation such headers
_can_ change is "changes to configure.ac" or "config.status --recheck".

A plain "./config.status ...." would only be required, if a user
removes/trashes the generated file (cf. ./config.status <file>).

2. The package, I apply this approach to, isn't necessarily "big" code
size-wise, but it is "complex and big", autotool-wise ;-)
(http://www.rtems.org/cgi-bin/viewcvs.cgi/rtems/cpukit/configure.ac)


Ralf