Re: newlib: require autoconf-2.69

[Corinna Vinschen <vinschen@redhat.com>]

On Jan  7 14:10, Mike Frysinger wrote:
> On 07 Jan 2022 10:56, Corinna Vinschen wrote:
> > On Jan  5 20:46, Mike Frysinger wrote:
> > > On 05 Jan 2022 13:10, Corinna Vinschen wrote:
> > > > On Dec 30 13:39, Mike Frysinger wrote:
> > > > > The top-level newlib dir is already using autoconf-2.69, and all the
> > > > > subdirs are using autoconf-2.68.  Update them all to use 2.69 as a
> > > > > prereq and then regenerate them accordingly.
> > > > > 
> > > > > The README file already says to use autoconf-2.69.
> > > > 
> > > > Yeah, but...
> > > > 
> > > > The AC_PREREQ only checks for the minimum required autoconf version for
> > > > *this* specific autoconf file.  AC_PREREQ([2.52]) doesn't mean we have to
> > > > use 2.52, but any later version will suffice.  All current autoconf
> > > > files are compatible with 2.69.  Neither binutils-gdb, nor gcc changed
> > > > their files to require 2.69.  So it's note quite clear to me what this
> > > > patch accomplishes.
> > > 
> > > that's incorrect.  gcc & binutils-gdb have config/override.m4 which requires
> > > autoconf-2.69 exactly is used.  no older versions, no newer versions, only
> > > 2.69 is permitted.
> > 
> > You're right, but that's not what I was up to.  Requiring 2.69 doesn't
> > mean we have to (or even should) change AC_PREREQ in every single file.
> > But, anyway, it probably won't hurt either, so just go ahead.
> 
> that's true.  but then i'd question what's the point of AC_PREREQ in the first
> place.  if we move to override.m4, maybe we delete all the AC_PREREQ calls ?
> -mike

I think it's ok to leave the files alone, unless a change is really
required.  The fact that they claim to run even under, say, autoconf
2.50, is still true and doesn't hurt the build system.  If it hurts the
build system because they use an outdated autoconf expression, then a
fix should also bump AC_PREREQ.

Of course, if binutils-gdb decide differently, we can follow suit.


Corinna