Re: about documentation of AC_CHECK_HEADERS

[Vincent Torri]

On Sat, 14 Mar 2009, Ralf Wildenhues wrote:

> Hello Vincent,
>
> thanks for the report.
>
> * Vincent Torri wrote on Sat, Mar 14, 2009 at 05:51:21PM CET:
> > more precisely, if foo.h exists and bar.h does not:
> >
> > 1) with:
> >
> > AC_CHECK_HEADERS([foo.h bar.h], [have_headers="yes"], [have_headers="no"])
> >
> > have_headers value is "no"
> >
> > 2) with:
> >
> > have_headers="no"
> > AC_CHECK_HEADERS([foo.h bar.h], [have_headers="yes"])
> >
> > have_headers value is "yes"
> >
> > which is quite annoying, in my humble opinion
>
> Yes.  Please consider the situation where one is looking for the header
> of one particular API, but that API may be declared in one of several
> _alternate_ headers, foo.h or bar.h.  In that case, the current
> semantics are useful.
>
> It seems that with the current implementation, the semantics you would
> like are easily emulated, and AFAICS the current documentation matches
> the current semantics, too.  Consequently, I don't think this needs a
> change in Autoconf.

indeed, now I know what to do...

> Aside, a change in behavior of this prominent macro would be a sure way
> to destroy backward compatibility.  So if this were seen as broken, it
> would need a new macro with better semantics, plus deprecation of this
> one.

indeed, changing the macro behavior is not an option and I don't want 
that (that's the reason of the subject of my mail : the documentation). 
The problem is the documentation, I think. The 2 behaviors must be more 
detailed, I think. Something like:

"If action-if-not-found is given, it is executed when one of the header 
files is not found, even if action-if-found is given."

Then you can add the 2 examples above.

Otherwise, it's quite disturbing (well, it took me some time to understand 
why what i wrote was not working...)

Vincent Torri