Re: Portability issue with `putenv'

[James Youngman]

On Tue, Feb 22, 2005 at 04:45:42PM +0100, Sam Lauber wrote:
> [ James Youngman wrote: ]
> 
> > You're right, I hadn't thought of that.  According to the POSIX
> > standard (IEEE Std 1003.1, 2003 Edition),
> > 
> >    |  The 'string' argument should point to a string of the form " name=
> >    |  value ".  The putenv() function shall make the value of the
> >    |  environment variable 'name' equal to 'value' by altering an
> >    |  existing variable or creating a new one. In either case, the
> >    |  string pointed to by 'string' shall become part of the
> >    |  environment, so altering the string shall change the environment.  
> 
> According to the DJGPP C Library manual, it seems to say 
> things are the other way around.  It says that the string 
> is _supposed_ to be copied.

Yes, but the DJGPP library intends to be a C library for DOS systems,
and does not intend, as far as I know, to provide a POSIX compilation
environment.  Therefore there is no need for it to comply with POSIX
and so it can do whatever it wants (putenv is not part of the standard
C library).

> But if there's at least one C library or compiler that's 
> not standards compliant, someone has to cater to them or 
> risk that construct being nonportable without anyone 
> knowing it, everyone using it, only to break on a platform 
> like DJGPP.  (I can justify that it's the job of Autoconf 
> to track down those things.).  

It's an open question as to whether Autoconf should replace an
OS-specific putenv() function just because it doesn't fully comply
with POSIX.

> P.S. Nowadays, I've noticed a lot of free software is 
> impolite to DOS systems.  

Well, the semantics of environment segments in DOS are baroque.  See
http://excession.spiral-arm.org/jay/programs/envset.c for an example
of the full horror of these things.

Regards,
James.