Re: Detecting (g)m4

[Akim Demaille]

Le 16 juil. 09 à 19:27, Eric Blake a écrit :

Hi Eric!

> Akim Demaille <akim <at> lrde.epita.fr> writes:
>
> > > Maybe not quite a bug, and maybe not your fault, but configure
> > > apparently picks the first program called "gm4" in my path, >not<
> > > the first appropriate m4 or gm4.
> >
> > You are right, of course.  But it's more efforts to look for the
> > "best" m4 around, so improvements, unless contributed by some devoted
> > user (read my mind ;), are unlikely to be made soon.  Especially
>
> </me puts on ESP hat> Are you talking to me?

Well, I was more thinking about the original reporter (lost in the  
CC's btw).  But I'm fine with your reading of my mind :)

> > because I see no existing Autoconf macro to address this issue.   
> > Maybe
> > Autoconf itself has one (after all it is probably the most demanding
> > M4 application).
>
> Which version of bison?

I had forgotten the changes from Joel and you that moved to using  
Autoconf's macro.

> I can confirm that the latest released version, bison 2.4, used m4/ 
> m4.m4 dating
> from 2000, which does indeed have severe limitations at detecting a  
> good m4
> version.  You can work around them by calling './configure
> M4=/path/to/your/preferred/m4'.
>
> Meanwhile, the latest git version uses serial 5 of autoconf's m4/ 
> m4.m4, which
> does a MUCH better job at picking m4 1.4.5 or newer, regardless of  
> spelling.
> Autoconf itself has moved on to serial 6 (which now checks for the  
> m4 -g flag
> added in m4 1.4.12 or newer), so maybe it's time to do another  
> submodule update
> to pick that up.

Sounds like a good idea.  Thanks for the advice!