Re: Useful patch for GNUmake on Windows

[Greg Chicares]

Eli Zaretskii wrote:
> 
> On Thu, 19 Sep 2002, Earnie Boyd wrote:
> 
> > > the DJGPP port does not
> > > automatically honor SHELL set in the Makefile, and it does import SHELL
> > > from the parent environment.
> >
> > I'll have to think about that.  What kinds of issues were caused by
> > this?
> 
> Mainly, two considerations:
> 
>   - Unlike on Unix, SHELL is normally not set on DOS/Windows systems.  So
>     if someone does set them, it's a good sign that she actually wants
>     that shell to be used everywhere, not only in interactive sessions.
> 
>   - Many Unix-born Makefile's say "SHELL = /bin/sh" (in particular, this
>     is required by GNU coding standards), but most DOS/Windows machines
>     don't have a /bin directory on every disk, and many don't even have a
>     Unix-compliant sh.exe installed anywhere.

With MSYS,
  $ set |grep -iw shell
  SHELL=/bin/sh
  $ ls /bin/sh 
  /bin/sh
so I'd think MSYS should follow the *nix behavior.

The make manual could be revised to say that, but
it's not trivial to cover all cases--my cygwin
installation (perhaps a bit dated) says
  $ set | grep -iw shell
  SHELL=/bin/bash
  _=shell
  $ ls /bin/sh
  c:\cygwin\bin\ls.exe: /bin/sh: No such file or directory
  $ ls /bin/bash
  c:\cygwin\bin\ls.exe: /bin/bash: No such file or directory

But if the manual's only changed to say in essence
"here's the default dos/windows behavior, that's
what djgpp uses, MSYS follows the *nix behavior,
and other 'environments' may do something else",
is that an improvement? Is this behavior determined
at build time, at run time, or a little of both?
(Yeah, I could read the source, but I'm way
behind on a GPL project I'm getting paid for.)

The manual is important. I've read it more times
than Lord of the Rings, and that's saying something.
I can write up a patch for it, but need to ask
for this guidance first.