X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
From: 048321887-0001 AT t-online DOT de (Udo Kuhnt)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: Using DR-DOS fork in DJGPP
Date: 27 May 2004 00:35:13 -0700
Organization: http://groups.google.com
Lines: 27
Message-ID: <4d201f78.0405262335.2faffca3@posting.google.com>
References: <4d201f78 DOT 0405181653 DOT 16a677b6 AT posting DOT google DOT com> <4d201f78 DOT 0405191158 DOT 63550b51 AT posting DOT google DOT com> <7704-Wed19May2004233123+0300-eliz AT gnu DOT org> <4d201f78 DOT 0405191847 DOT 6697f90d AT posting DOT google DOT com> <u3c5vbous DOT fsf AT gnu DOT org> <4d201f78 DOT 0405200540 DOT 307bb15a AT posting DOT google DOT com> <9743-Thu20May2004201223+0300-eliz AT gnu DOT org> <4d201f78 DOT 0405210913 DOT 26a5ffcb AT posting DOT google DOT com> <uwu33ad61 DOT fsf AT gnu DOT org> <4d201f78 DOT 0405230527 DOT 3f17fb4b AT posting DOT google DOT com> <2719-Sun23May2004185224+0300-eliz AT gnu DOT org>
NNTP-Posting-Host: 217.93.38.138
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit
X-Trace: posting.google.com 1085643314 18250 127.0.0.1 (27 May 2004 07:35:14 GMT)
X-Complaints-To: groups-abuse AT google DOT com
NNTP-Posting-Date: Thu, 27 May 2004 07:35:14 +0000 (UTC)
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

"Eli Zaretskii" <eliz AT gnu DOT org> wrote in message news:<2719-Sun23May2004185224+0300-eliz AT gnu DOT org>...
> > From: 048321887-0001 AT t-online DOT de (Udo Kuhnt)
> > Newsgroups: comp.os.msdos.djgpp
> > Date: 23 May 2004 06:27:29 -0700
> > 
> > If a file had been opened by the parent process, it should also be open for
> > the child process which can access it using the same handle. If the child
> > closes the file in its VM using the handle, it will still be open in the
> > parent's VM.

That is probably the reason why the task manager in DR-DOS requires SHARE to
be loaded, but requiring SHARE is hardly a problem in this case.

> And the latter means that just copying is not good: it is a frequent
> case that the child closes one or more of inherited handles that it
> doesn't need.  You don't want that to invalidate the handle in the
> parent, since existing programs don't expect such a calamity.

It is not invalidated because each VM has its own set of file structures.

> Se perhaps writing a (16-bit real-mode) program to test this aspect
> would be a useful first step.

??? I do not see the point - DJGPP cannot produce such executables, and using
a different compiler would make the result harder to interpret. Besides, the
MT API in DR-DOS calls for a 32-bit implementation, so using a 16-bit test
program would probably not work anyway.