Hi again
(sorry for so many mails)
are there any objections to make "--with-socket" the
default driver in configure?
Markus
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'
Hi Markus
Markus Neteler wrote:
are there any objections to make "--with-socket" the
default driver in configure?
Well, seeing as most of the problems with process communication we are
currently experiencing (eg broken Makefile) is from trying to support 3
types of communication, I would suggest that we pick a single
implementation and strip out the others from the code. The question is
which implementation do we keep? The fifos don't work on all platforms
and the ipc code can leave message-queues allocated (if d.mon stop=x?
isn't used) that can only be freed with a system reboot. As far as I can
tell, Eric has fixed most of the problems (if not all) that were
reported for the sockets implementation, so unless there is still an
outstanding problem with them, it looks like sockets is an all win
implementation.
Does anyone have a reason for keeping more than one implementation?
Just my 2 cents worth.
--
Sincerely,
People who think they know everything are very irritating to those
of us who do. ---Anonymous
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'
On Mon, Mar 05, 2001 at 02:36:20PM +0700, Justin Hickey wrote:
Hi Markus
Markus Neteler wrote:
> are there any objections to make "--with-socket" the
> default driver in configure?Well, seeing as most of the problems with process communication we are
currently experiencing (eg broken Makefile) is from trying to support 3
types of communication, I would suggest that we pick a single
implementation and strip out the others from the code.
We should leave fifos and sockects in.
Fifos are known to work on some platforms we have not tested yet.
So they have to be provided as a fallback.
Sockets seems to work best on most our testing platforms, so I vote
for making them the default.
--
Professional Service around Free Software (intevation.net)
The FreeGIS Project (freegis.org)
Association for a Free Informational Infrastructure (ffii.org)
FSF Europe (fsfeurope.org)
Hi Bernhard
Bernhard Reiter wrote:
We should leave fifos and sockects in.
Fifos are known to work on some platforms we have not tested yet.
So they have to be provided as a fallback.
Can we identify these platforms and test them? Maintaining more than one
implementation is becoming complex and putting little hacks throughout
the initialization code and Makefiles. Besides, with CVS, we can always
restore them can't we?
Just a thought.
--
Sincerely,
People who think they know everything are very irritating to those
of us who do. ---Anonymous
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'
On Mon, Mar 05, 2001 at 06:01:46PM +0700, Justin Hickey wrote:
Hi Bernhard
Bernhard Reiter wrote:
> We should leave fifos and sockects in.
> Fifos are known to work on some platforms we have not tested yet.
> So they have to be provided as a fallback.Can we identify these platforms and test them? Maintaining more than one
implementation is becoming complex and putting little hacks throughout
the initialization code and Makefiles. Besides, with CVS, we can always
restore them can't we?Just a thought.
I agree with Justin,
currently we are not aware of any platform not supporting sockets.
Markus
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'
Hi developers,
to speed up things, the "sockets" are default now as no
reasons have been posted why to keep fifos default.
Still to decide is:
- keep IPC (I prefer not)
- keep fifo (maybe, but I cannot see much reasons)
The configure[.in] are changed accordingly to make sockets
default.
Thanks,
Markus
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'
On Mon, Mar 05, 2001 at 03:01:18PM +0000, Markus Neteler wrote:
Still to decide is:
- keep IPC (I prefer not)
I also vote for removing it.
- keep fifo (maybe, but I cannot see much reasons)
Because this is the old method which is known to work on obscure
platforms we cannot test right now.
Bernhard
--
Professional Service around Free Software (intevation.net)
The FreeGIS Project (freegis.org)
Association for a Free Informational Infrastructure (ffii.org)
FSF Europe (fsfeurope.org)
Hi Bernhard
Bernhard Reiter wrote:
> - keep fifo (maybe, but I cannot see much reasons)
Because this is the old method which is known to work on obscure
platforms we cannot test right now.
I don't know. It just seems to me that since we have positive results
from Cygwin, Mac, Linux, Solaris, IRIX, CRAY, FreeBsd, (others??) it
indicates that our socket code is fairly portable. I also find it hard
to believe that there is a UNIX system around that does not support the
sockets standard. They have been around for more than a decade. Do you
have any particular platform in mind? I'd just like to have clean code
rather than keeping something that may not be used or that is not
necessary.
Anyway, I guess it's up to Markus.
--
Sincerely,
People who think they know everything are very irritating to those
of us who do. ---Anonymous
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'
On Tue, Mar 06, 2001 at 05:23:05PM +0700, Justin Hickey wrote:
Hi Bernhard
Bernhard Reiter wrote:
> > - keep fifo (maybe, but I cannot see much reasons)
>
> Because this is the old method which is known to work on obscure
> platforms we cannot test right now.I don't know. It just seems to me that since we have positive results
from Cygwin, Mac, Linux, Solaris, IRIX, CRAY, FreeBsd, (others??) it
indicates that our socket code is fairly portable. I also find it hard
to believe that there is a UNIX system around that does not support the
sockets standard. They have been around for more than a decade. Do you
have any particular platform in mind? I'd just like to have clean code
rather than keeping something that may not be used or that is not
necessary.Anyway, I guess it's up to Markus.
I don't have a 20 years experience, but I am rather sure that people
try to use modern systems. So I agree with Justin: a cleanup is less
error prone than keeping historical code (even CVS keeps it for us).
If we should keep the fifos, we need the platform named which doesn't
support sockets.
Markus
----------------------------------------
If you want to unsubscribe from GRASS Development Team mailing list write to:
minordomo@geog.uni-hannover.de with
subject 'unsubscribe grass5'