This is the mail archive of the pthreads-win32@sources.redhat.com mailing list for the pthreas-win32 project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: snapshot-2002-03-02 available


Hi,

"Crescioli, Phil" wrote:
> 
> Ross,
> Your win32pthreads lib saved me so much time porting my threaded code
> from unix to windows!

http://sources.redhat.com/pthreads-win32/contributors.html

>  Unfortunately some of my code works by sending data via
> message queues
> between threads of different processes.  Any idea if your going to
> support
> _POSIX_THREAD_PROCESS_SHARED in a near future release?
> Thanks,
> Phil Crescioli

As you know, there has been an attempt in the past to
include message queues.

Aurelio Medina <AMedina at xcaliber dot com> wrote an
implementation but unfortunately it relied on process shared
mutexes etc, which aren't supported by pthreads-win32.
Although, it did work for single process applications.

See thread starting at

http://sources.redhat.com/ml/pthreads-win32/2001/msg00085.html

Process shared objects would have to be the next major
ToDo for pthreads-win32 because there isn't a lot more that
can be added to the library without them. A few people have
been looking for message queues and possibly many others
who have considered pthreads-win32 and then silently moved
on.

I'm not sure if this is possible to do while supporting the
full POSIX semantics, but perhaps a restricted version might
be possible making use of Win32 named objects, putting
just the names into shared memory. The starting point
would have to be process shared semaphores, then the
other devices would start to fall into place.

I'll have to check over past discussion of this issue, but
I seem to recall some of the reasons it hasn't been done
yet are:

- technical feasibility. Anything is possible, but will it
  maximise the return on the investment :^);
- low priority. People who are willing to contribute
  time to the library don't need the feature themselves;
- it may start to pull in parts of the POSIX standard
  that are beyond the intended scope of the library,
  i.e. shared memory routines, even message queues.

Have you looked at other POSIX compatibility projects
such as Cygwin or UWIN?

Regards.
Ross

> 
> >>>The following POSIX 1003.1 2001 options are not defined:
> >>>_POSIX_THREAD_ATTR_STACKADDR
> >>>_POSIX_THREAD_PRIO_INHERIT
> >>>_POSIX_THREAD_PRIO_PROTECT
> >>>_POSIX_THREAD_PROCESS_SHARED
> 
> > -----Original Message-----
> > From: Ross Johnson [SMTP:rpj@ise.canberra.edu.au]
> > Sent: Saturday, March 02, 2002 1:24 AM
> > To:   pthreads-win32@sources.redhat.com
> > Subject:      NEW: snapshot-2002-03-02 available
> >
> > There are bugs fixes, new routines, new tests, and a lot
> > of reorganisation. Please see the ANNOUNCE, NEWS, and BUGS
> > files in the package or at:
> >
> > http://sources.redhat.com/pthreads-win32/announcement.html
> >
> > There is also an updated Copyright notice on the source
> > code, but NO CHANGE to the licensing, which is the GNU LGPL.
> >
> > The Copyright notice replaces the original inadequate notice
> > and reflects the heritage of the project.
> >
> > The Copyright notice is slightly different to the one that I
> > proposed on this list a few months back. This change was at
> > my own instigation. I don't expect anyone to have any
> > objection to it since there was no objection to the previous
> > one, but please let me know if you do.
> >
> > I've also made acknowledgements more prominent in the
> > CONTRIBUTORS file and ChangeLog to go with the change in the
> > copyright notice. Let me know if I've missed you out, or
> > missed noting any significant contributions.
> >
> > Enjoy!
> > Ross


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]