X-Authentication-Warning: new-smtp2.ihug.com.au: Host p178-apx1.syd.ihug.com.au [203.173.140.178] claimed to be acceleron
Message-ID: <00c801c12fb4$0a9748b0$0a02a8c0@acceleron>
From: "Andrew Cottrell" <acottrel AT ihug DOT com DOT au>
To: "Charles Sandmann" <sandmann AT clio DOT rice DOT edu>,
        "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>
Cc: <djgpp-workers AT delorie DOT com>, <wojciech DOT galazka AT polkomtel DOT com DOT pl>
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1010828092959 DOT 1704E-100000 AT is>
Subject: Re: ODP: Win2K/XP fixes - implementation review
Date: Tue, 28 Aug 2001 21:24:44 +1000
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.50.4807.1700
X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4807.1700
Reply-To: djgpp-workers AT delorie DOT com

> > dosexec, crt0, fstat are good for NT 4.0 also, not W2K specific.
> > dpmiexcp.c - setting PSP on exit, minor, not file related
> > open.c doesn't hurt anything, failure case only.
> > _rename.c - additional check for return failure code
> > utime.c - better than original code
> > _open.c, _creat*.c - more interrupts/file reopens (hurts NT4/LFN)
> >
> > The only ones which are W2K specific are the last 3.
>
> I think the set-PSP work-around is also specific to W2K.  It doesn't
> hurt, but it slows down by issuing several software interrupts.
>
> > They are only
> > on opening files.  If "open" of NUL shows get_dev_info of zero, then
> > we should use these patches, else not.  Since we already open nul for
> > the FSEXT stuff, we could start assuming no bugs, check for it, if we
> > see the bug close nul, turn on the bug-fix-switch continue.
>
> The FSEXT stuff doesn't get called unless the application uses the
> extensions.  By default, FSEXT doesn't get linked in.
>
> I agree that it's better to test for the presence of bugs directly,
> but this could lead to many tests and many flag variables which tell
> what bugs are present.  If we can lump the bugs in a couple of
> categories, I think it's okay; otherwise, lets stay with version test.
>
> For now, it sounds like we could get away with only 2 categories:
> NT+W2K and W2K-specific (the latter includes XP).  Maybe we should
> postpone these decisions a bit more.  I'm not sure we know the whole
> extent of the problems right now, although we are close.

Now for my 2 cents worth:

I agree with all that has been said. It looks like we have the standard
issue that you get from time to time:
   Do we cleanup the code and ensure that it then becomes easier to maintain
and expandable?
    or
   Do we release the code and fix it up while the end users are testing it
for us and giving us vital info on bugs which have not been fixed?

In the past I have erred towards the cleanup of the code if all of the known
issues and bugs have been fixed otherwise fix the bugs and issues first,
unless the code is so bad that the cleanup is required in order to fix bugs.
Once the bugs and issues have been fixed then cleanup the code while it gets
tested (alpha or beta release to the wider DJGPP users). The number of bugs
and issues is becoming less and less, but they are becoming harder and
harder to find and fix correctly.

I am in agreement with Eli in that we still do not know and have not done
enough testing to ensure that all of the problems with Win 2K and XP have
been resolved.