From: "Vyacheslav O. Myskin" <V DOT O DOT Myskin AT inp DOT nsk DOT su>
Organization:  BINP RAS, Novosibirsk, Russia
To: djgpp AT delorie DOT com
Date:          Wed, 19 Feb 1997 14:30:58 +0600
Subject:       Re: Starnge rm behavour
Reply-To: V DOT O DOT Myskin AT inp DOT nsk DOT su
Message-Id: <19B60DC0919@csd.inp.nsk.su>

On 19 Feb 97 at 9:49, Eli Zaretskii wrote:

>
> On Tue, 18 Feb 1997, John M. Aldrich wrote:
>
> > The problem seems to be occurring because 'redir' doesn't perform
> > wildcard expansion on the command line it gets passed, yet it uses
> > spawnvp() to invoke the indicated program with the remaining arguments.
> > Obviously, this breaks any program which gets passed a wildcard,
> > including 'rm'.
>
> This analysis is correct.  `redir' doesn't glob the command line so you
> could use the same command with and without it (otherwise you would have
> to quote the wildcards when running programs from `redir').  When in v2.01
> arguments passed by `spawnXX' stopped from being expanded on the child
> side, this feature introduced a bug in some (admittedly rare) cases.
>

Sorry, but I don't understand... I thought that it was rm's job to expand
a wildcard. How could spawnXX expand anything without knowing what does
it mean for a particular child (rm in this case)? Or does rm refuse to
expand '*' only because it knows that it was passed via spawnXX (can't
see any sense)? So if rm sees an asterisk as an argument, why doesn't
it process it in a proper way (that means search for all files in a directory
and delete them)? What am I missing here?

It wouldn't be so bad if the problem was with redir only - what's the
use of launching rm via redir? But it makes impossible to use
rm -f somedir/* in makefiles, which is much worse.


______________________________________________
Vyacheslav O. Myskin  <myskin AT inp DOT nsk DOT su>
Budker Institute of Nuclear Physics (BINP RAS)
Novosibirsk, Russia