Re: why not check $PATH for m4?

[Chris Pickett]

Eric Blake wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> According to Chris Pickett on 2/2/2007 10:58 PM:
> > Hi,
> >
> > I'm not subscribed, please reply-all.
> >
> > I installed m4-1.4.8 locally.  However, in order for auto(re)conf find
> > it, I have to do `export M4=$HOME/$MY_PREFIX/bin/m4'
> >
> > Otherwise, with an empty $M4, I get this error from /usr/bin/m4:
> >
> > /usr/bin/m4: `changeword' from frozen file not found in builtin table!
>
> Sounds like you build $HOME/$MY_PREFIX/bin/m4 with changeword support, but
> that your installed version of autoconf was built at a time when only an
> older m4 version at /usr/bin/m4 was available.  The location of M4 is
> determined when you run ./configure on the autoconf tarball; if you want
> to use your newer m4 installation, but don't have root privileges to
> install it in the same places as where the existing autoconf is already
> looking, then I would suggest also configuring autoconf and installing it
> alongside m4 in $HOME/$MY_PREFIX/bin.

Thanks a lot!

> > Is this the right behaviour for some reason I don't know?  Can the $PATH
> > be checked for m4 first?
>
> The PATH check occurs at ./configure time for autoconf, not at runtime
> during autom4te, as a runtime check would noticeably slow down autoconf
> and automake.

Out of sheer curiousity, did anyone perform experiments to verify that a 
runtime PATH check is noticeably slower?  Why is it slower?  Is it 
because it must look for binaries named m4 in each directory on the 
path?  That seems about as slow as when I just type m4 and hit enter at 
a prompt, which is, needless to say, relatively unslow.

> Don't work too hard, make some time for fun as well!

I like your signature :)

Chris