Date: Wed, 13 Mar 1996 17:59:13 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Charles Marslett <charles DOT marslett AT tempe DOT vlsi DOT com>
Cc: djgpp-workers AT delorie DOT com
Subject: Re: Library docs-questions
In-Reply-To: <3145AC3E.4D53@tempe.vlsi.com>
Message-Id: <Pine.SUN.3.91.960313174852.10443E-100000@is>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


On Tue, 12 Mar 1996, Charles Marslett wrote:

> Permitting the operation of a nested shell to affect the outer shell 
> would be confusing, I think, since Unix shells, DOS batch files and 
> every version of make I am aware of work just like this.  Why would 
> you want to break such a pervasive standard?

Unix shells do that, but DOS batch files don't: if you chdir in a batch 
file you'll wind up in that directory when the batch file terminates.  I 
don't know what versions of Make did you see that enforce this; the 
current version of DJGPP Make doesn't (it doesn't use the code in 
`system' that was the cause of my complaint).

DJGPP is not only about porting Unix Makefiles and making DOS look more
like Unix; it's also about developing native DOS apps.  People who work 
under DOS are used to the global effects of `chdir' and expect it when 
they write Makefiles.  Take a look at the Makefiles in the DJGPP archives 
of ported GNU programs and tell me how many of them can do without `cd'.
I did check, and the answer is: almost none.  I'd speculate that 
Makefiles written originally under DOS would use this even more.

On Unix, you'd write a `cd' and a command on one line with a semi-colon 
to make it affect the command, but DOS shell cannot grok multiple 
commands on a single command line.