PATCH/RFA : texinfo bootstrap failure

[Jeffrey A Law]

  In message < 4.3.2.20000417114011.00dee460@mail.lauterbach.com >you write:
  > At 06:35 17.04.00, Philipp Thomas wrote:
  > >* Alexandre Oliva (aoliva@cygnus.com) [20000417 06:07]:
  > >
  > > > Actually, it should be installed, just not updated (assuming that, if
  > > > we stop shipping GCC with Texinfo, we're going to ship it with the
  > > > .info files)
  > >
  > >If we're going to do the latter, then I'd agree with you. I started from t
  > he
  > >assumption that we'd only ship the texinfo sources.
  > >
  > >I'd still argue for a prominent diagnostic telling the user what will/won'
  > t
  > >happen.
  > 
  > I would argue for the following (I don't mention "install-info" here for 
  > compactness):
  > 
  > - "make/make all/make bootstrap": don't generate *.info
  > - "make info" with correct makeinfo version: generate *.info
  > - "make info" with wrong makeinfo version: don't generate *.info, warn user
  >  
  > (non-fatal)
  > - "make install": call the "info" target (probably as the final make target
  > )
  > 
  > While we're at it, we should assure that the MAKEINFO (and maybe 
  > MAKEINFOFLAGS?) environment values are correctly handled by configure and 
  > the values are passed down to the sublevel Makefiles (eg. 
  > MAKEINFO="makeinfo --no-split").
We will continue to include pre-built info files in our releases (and
possibly snapshots).

The pre-built info files will not be in the CVS tree.

So, if no suitable texinfo is available and we're building from CVS we need
to either give an error, or a warning.  I'm not sure which is best, though
I'd lean towards a warning since the info files are not critical on a daily
basis for people working on the CVS tree.

For releases (and possibly snapshots) we'll be providing suitable info files
so it shouldn't be necessary to rebuild them.  Though doing so won't hurt if
the "make" logic to avoid building unnecessary info files is hairy.

jeff