From: Jason Green <news AT jgreen4 DOT fsnet DOT co DOT uk>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: DJGPP neither running on ms-dos 6.2 nor on caldera-dos 7.03
Date: Sun, 13 Aug 2000 10:10:34 +0100
Organization: Customer of Energis Squared
Lines: 39
Message-ID: <guocps0l3eg5sesh2b8128pbu4dvv1qgmf@4ax.com>
References: <8mq6n6$6gu8v$1 AT ID-16903 DOT news DOT cis DOT dfn DOT de> <39912C43 DOT DF0BB5FD AT maths DOT unine DOT ch> <200008091553 DOT LAA13441 AT delorie DOT com> <tae6pscelfpqbs7lvoe7vsc5qcuriq9dun AT 4ax DOT com> <200008111728 DOT NAA06831 AT delorie DOT com>
NNTP-Posting-Host: modem-43.lithium.dialup.pol.co.uk
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Trace: news7.svr.pol.co.uk 966157845 7103 62.136.2.43 (13 Aug 2000 09:10:45 GMT)
NNTP-Posting-Date: 13 Aug 2000 09:10:45 GMT
X-Complaints-To: abuse AT theplanet DOT net
X-Newsreader: Forte Agent 1.7/32.534
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 is DOT elta DOT co DOT il> wrote:

> > Most of these FAQs probably stem from #include <iostream>, which has
> > streambuf.h nested in it.
> 
> This is only one of the problems.  Take a look at lang/cxx and
> lang/cxx/std: there are a lot of files there whose names exceed the
> DOS 8+3 limits.

Yep, I see them.  I am assuming that most of the FAQs to this group
are due to including the most common header - iostream. 

> > How about making a copy of streambuf.h with
> > an 8.3 filename and editing the nested include within iostream.h?
> 
> You suggest a manual solution.  This is error prone (what if someone
> who ports GCC forgets to do that?), and it doesn't solve the case when
> a new header is introduced with a new release of GCC that wasn't known
> before.

It's a bodge, yes, and a way to reduce the number of FAQs (maybe ;-).

> Anyway, thanks for the suggestions.

You are welcome.  Though I think the -remap option suggested by
yourself and Andris is the most promising, in fact I don't understand
why this is not already done.  Except, perhaps, the effort involved in
implementing it.

AFAIK, there is nothing in the C++ standard that requires the header
files to exist with the same names, or in fact to exist as files at
all.  Only that after #include <foo>, all the required names are
declared.  This, I believe, is partly the reasoning behind dropping
the ".h" and also why the C++ commitee allowed LFNs, which are not
permitted in ANSI C.

It seems very reasonable that gcc should translate #include statements
to a form compatible with the host filesystem.  This appears, to me at
least, to be exactly what is intended by the C++ standard.