[GRASS5] mkfifo start for cygwin

[Malcolm Blue]

John,

I have been following the cygwin mailing lists for a while and it seems
that some of the core developers have made some starts at fifo
implementation but haven't decided which approach to use.  Since you
have been in contact with some of them, you may have more recent
information.  Are they close to having something working, do you think?
I seem to recall that fairly recently there was an exchange of emails
posted that pointed out reasons for not using various approaches, but
there was no indication that they were still actively working on
anything.  If there is a clear indication that they will be implementing
something soon,  I can probably find more productive pursuits than
learning IPC messaging.  If fifo's are to be implemented as a part of
the standard cygwin distribution in the future then that is probably a
better path to follow.

I could (and probably will) keep on working on the IPC code a bit more
to see if I can get it working - even a little.  It seems to have a few
kludges built into it, but it may serve as a temporary fix for now.
Right now, I have been setting up some tests to see how well it behaves
as a msg queue server and have found some unexpected results (not the
good kind), even for the standalone tests that mimic the simple d.mon
programs and XDRIVER startup/status/close.  I haven't even gotten into
the message send and receive code in any detail yet (although my simple
tests show that these work ok if the queues are managed cleanly).  To
make it robust may take quite a bit of effort.  To just make it work
should not be too big a job, but will still take several days to make it
even close to reliable.  It seems to work very well in a basic mode of
operation, but does not behave well the way that the grass routines use
IPC message queues.   I will be able to work on it again in January if
others think that the cygwin IPC is worthwhile pursuing. (I probably
will anyway since I hate quitting anything in midstream.)

I will also take a look at the mkfifo implementation you pointed to
since there may be a cleaner solution there.  I will look to see if
these can be used with the existing code base, with a little hacking.
>From archived mailing list discussions it seems that fifo's are more
portable.  If a temporary fix with fifo's is available, that may be a
better alternative in the long run.

Either approach will probably still be experimental in nature.  I will
post my results to the mailing list so that you and other interested
parties can provide suggestions and so that I don't step on anyone's
toes if they are doing related tasks.

Speaking of related tasks:  Andreas Lange has said that he has started
to merge the IPC and fifo codes into one setup.  If he is doing this
mainly for the cygwin port, is this still something worth doing? ( I
expect that Andreas will let us know if there are other good reasons for
continuing.)

If you, or anyone else!!!, has any other relevant information,  please
pass it on.  I would rather contribute my efforts to a joint cause than
just get something working for myself.

Cheers,

Malcolm


John Huddleston wrote:

> Markus,  (Cc to grass)
>
> I went through some of the older Cygwin messages and
> found an email note from  someone was thinking of using
> Cygwin on his NT box for development.  He states that
> there was no reply from anyone and he has subsequently
> moved to Linux.  Included below is his reply to me with
> a URL to his source code.   I sent this earlier to Mike and
> Andreas but realized (with Malcom's email) that we all
> would be interested in this information.
>
> I downloaded the code from http://herzberg.ca.sandia.gov/namedpipes
> and it compiled cleanly.  There is a mkfifo.cc file used by two
> standalone executables.
>
> John Huddleston
>


---------------------------------------- 
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo at geog.uni-hannover.de with
subject 'unsubscribe grass5'