Mailing-List: contact cygwin-developers-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-developers-subscribe AT sources DOT redhat DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin-developers/>
List-Post: <mailto:cygwin-developers AT sources DOT redhat DOT com>
List-Help: <mailto:cygwin-developers-help AT sources DOT redhat DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-developers-owner AT sources DOT redhat DOT com
Delivered-To: mailing list cygwin-developers AT sources DOT redhat DOT com
Date: Sun, 16 Sep 2001 13:23:22 -0400
From: Christopher Faylor <cgf AT redhat DOT com>
To: cygwin-developers AT cygwin DOT com
Subject: Re: possible explanation for make hang
Message-ID: <20010916132322.B5329@redhat.com>
Reply-To: cygwin-developers AT cygwin DOT com
Mail-Followup-To: cygwin-developers AT cygwin DOT com
References: <20010916013932 DOT A30789 AT redhat DOT com> <Pine DOT NEB DOT 4 DOT 30 DOT 0109152245330 DOT 3759-100000 AT cesium DOT clock DOT org> <20010916020118 DOT A30902 AT redhat DOT com> <20010916111330 DOT A15312 AT cygbert DOT vinschen DOT de>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20010916111330.A15312@cygbert.vinschen.de>
User-Agent: Mutt/1.3.21i

On Sun, Sep 16, 2001 at 11:13:30AM +0200, Corinna Vinschen wrote:
>On Sun, Sep 16, 2001 at 02:01:18AM -0400, Christopher Faylor wrote:
>> On Sat, Sep 15, 2001 at 10:58:30PM -0700, Matt wrote:
>> >On Sun, 16 Sep 2001, Christopher Faylor wrote:
>> >
>> >> On Sun, Sep 16, 2001 at 03:09:29PM +1000, Robert Collins wrote:
>> >
>> >> You can't normally use WFMO for file handles.  I actually tried this an
>> >> hour or so ago to see if maybe it actually worked on Windows 95.  It
>> >> doesn't.
>> >
>> >Out of curiosity, are you testing on Win95 950, 950a, or 950b? I remember
>> >when I did QA, some API calls that did not work as documented on 950/950a
>> >would work fine on 950b. If you can't find 950b specifically, testing on
>> >win98 is almost equivelant (950b has the win98 "kernel", for the
>> >mostpart).
>> >
>> >If you don't have 950b or win98 handy, I can try an example test
>> >case if you have one compiled or in source.
>> 
>> It doesn't really matter.  If it doesn't work on one system, it isn't
>> useful.
>
>IMO, it does matter.  We have the same problem of non-interuptible IO
>on sockets.  It's solved in net.cc by using the `CancelIO' call which
>doesn't exist in 95.

I think you are all confused by my mention of Windows 95.

AFAIK, any pipe issues that exist with Windows 95 exist with Windows 98,
too.

I mentioned Windows 95 because that is what I have available for
convenient testing.  It's on my laptop.  Otherwise, I have to boot my
W2K system into Windows 98 or kick my kids off their game system.  Neither
is as convenient as just switching to my laptop for tests.

There is one specific issue with PeekNamedPipe not detecting EOF that
exists on Windows 95 -> Windows ME.  I was hoping that maybe if I could
do a WFSO on a pipe on these platforms that I could possibly work around
this problem.  That was not the case.

I guess it is possible that this would work on something > Windows 95.
I didn't test that.  I tend to doubt that Microsoft would make things
less compliant with their documentation as they "improved" the
OS so I didn't think it was worth tracking down.

>My humbly opinion:  95 is really, really old now.  It has been
>substituted by two following OSes in the meantime and the third
>is coming soon.  Even Microsoft has canceled support for 95 and
>I can understand them.  If 95 doesn't work in a specific part of
>Cygwin which works fine in 98/ME/NT/W2K/XP, we shouldn't care
>anymore.  We just stop to create workarounds which are really only
>for 95.

We have a couple of minor workarounds for Windows 95.  The CancelIo call
is one and one that I added yesterday is another.  I don't even know if
it is true that my problem exists only on 95.  Since my fix was trivial,
and since *people were complaining about it*, I fixed it.

Regardless of what Microsoft says about support, if we can trivially
support Windows 95, we will.

Maybe I'm mistaken but I thought that the vas majority of workarounds in
Cygwin were not specifically for Windows 95.  They were for Windows 95,
98, and ME.

So, to reiterate, the issue of dropping Windows 95 does not solve the
pipe problem in any way.

cgf