This is the mail archive of the
cygwin-developers@cygwin.com
mailing list for the Cygwin project.
Re: time for 1.5.11?
- From: Corinna Vinschen <vinschen at redhat dot com>
- To: cygwin-developers at cygwin dot com
- Date: Mon, 5 Jul 2004 11:33:47 +0200
- Subject: Re: time for 1.5.11?
- References: <20040609170113.GA2866@coe.casa.cgf.cx> <20040609193110.GA28025@cygbert.vinschen.de> <Pine.GSO.4.58.0406091555190.20998@slinky.cs.nyu.edu> <Pine.GSO.4.58.0406091623070.20998@slinky.cs.nyu.edu> <20040609210144.GA10739@coe.casa.cgf.cx> <Pine.GSO.4.58.0406092042520.20998@slinky.cs.nyu.edu> <40C8B20B.906183E@ieee.org> <Pine.GSO.4.58.0406101518460.6113@slinky.cs.nyu.edu>
- Reply-to: cygwin-developers at cygwin dot com
On Jun 10 15:23, Igor Pechtchanski wrote:
> On Thu, 10 Jun 2004, Pierre A. Humblet wrote:
>
> > Igor Pechtchanski wrote:
> > >
> > > Ok, it turns out that to fix this properly, we'll need a much larger patch
> > > than the one line I was looking at... Here are some thoughts.
> > >
> > > First off, my guess above was wrong. The IO handle has nothing to do with
> > > this -- fhandler_dev_dsp doesn't use it. However, the intuition (I still
> > > think) was correct: some fields, namely the audio_in_ and audio_out_
> > > pointers, weren't replicated on dup(), so the new (duped) object ended up
> > > with NULL fields, and couldn't write to them.
> > >
> > > The simplest solution to the above is to simply copy the pointers. This
> > > will work for the test program I provided earlier, but won't for another
> > > legitimate test program (attached) that closes the original handle before
> > > using the duped one (I think this behavior is allowed).
> > >
> > > The "proper" way to fix this, IMO, would be to allow sharing the audio_in_
> > > and audio_out_ objects, and have some sort of reference count in them,
> > > instead of deleting them outright in close().
> >
> > FWIW, this reminds me of the archetype concept that cgf has built into cygwin.
> > It's currently used for ttys.
> >
> > Pierre
>
> Thanks, Pierre, I've looked at them. They seem much more complex than
> what's needed here, which is simply a way of sharing objects via
> pointers... Archetypes also seem to go out of their way to avoid an extra
> level of indirection, but since we have it here for free anyway, reference
> counting looks to be easier.
> Igor
Any patch in sight?
Corinna
--
Corinna Vinschen Please, send mails regarding Cygwin to
Cygwin Co-Project Leader mailto:cygwin@cygwin.com
Red Hat, Inc.