Re: Compiling current CVS Emacs on Windows NT

[Eli Zaretskii]

> From: Jason Rumney <address@hidden>
> Date: 05 Jan 2002 15:12:10 +0000
> 
> "Eli Zaretskii" <address@hidden> writes:
> 
> > What are the problems to support producing these two manuals on
> > Windows?  If it works for me on MS-DOS, Windows shouldn't be too
> > hard ;-)
> 
> makeinfo is not always available.

Is this an interesting situation?  Building a CVS version _requires_
makeinfo, on any system.  People who don't have makeinfo will be
unable to build the manuals on _any_ system.

> If someone (TM) wants to produce
> some makefiles and configure.bat rules that will build the manuals
> when makeinfo is available, and not produce an error (or better, build
> the manuals with Emacs if that works) when makeinfo is not installed,
> it would be a most welcome improvement.

If you need a test for whether makeinfo is installed, I can suggest
the following trick:

  - invoke makeinfo on some Texinfo file like this:

     makeinfo foo.texi --no-headers > junk

  - use "command.com /c COPY" to copy the file `junk' to some other
    file;

  - test if that other file exists; if it doesn't, there's no
    makeinfo on this system.

This uses a ``feature'' of command.com's built-in COPY command
whereby it refuses to copy empty files(!).

Such a test could be run once, before the build starts, and then its
result used in the man, lispref, and lispintro directories.

> The easiest approach is probably to use the existing Makefile.in, and
> assume that makeinfo is not available when building with nmake/msvc.

I'd say the opposite: assume that if the Texinfo files are newer than
the Info files, makeinfo is _always_ present.  All you risk is an
error message.  The opposite case, where people who do have makeinfo
cannot produce the manuals is IMHO much worse.