Glynn wrote:
Note that the monitor sockets no longer live in <mapset>/.tmp/...,
but
in /tmp/grass-<user>.
That change was made in 5.0 since pre2?
That seems like a good change, but not necessarily better then saving
under a user's ".grass" directory, which was the original option.
Hopefully this was done in such a way that it won't cause me major
update problems. Clearly while this change might be benign it should
serve as an example of how changes in GRASS could make it very difficult
to upgrade a production GRASS installation.
Further, this is a pretty fundamental change. This along with some
relatively recent changes in the monitor procedures just reinforce my
impression that GRASS5.0 is *not* yet stable.
Roger Miller
On Mon, Mar 25, 2002 at 09:16:34AM -0700, rgrmill@rt66.com wrote:
Glynn wrote:
> Note that the monitor sockets no longer live in <mapset>/.tmp/...,
but
> in /tmp/grass-<user>.
That change was made in 5.0 since pre2?
Yes. It appeared in pre3. You find this documented in
http://grass.itc.it/grass5/source/NEWS.html
or also
http://freegis.org/cgi-bin/viewcvs.cgi/~checkout~/grass/NEWS.html?rev=1.129.2
That seems like a good change, but not necessarily better then saving
under a user's ".grass" directory, which was the original option.
Hopefully this was done in such a way that it won't cause me major
update problems. Clearly while this change might be benign it should
serve as an example of how changes in GRASS could make it very difficult
to upgrade a production GRASS installation.
Further, this is a pretty fundamental change. This along with some
relatively recent changes in the monitor procedures just reinforce my
impression that GRASS5.0 is *not* yet stable.
Mhhh, in the opposite an unmaintained system also doesn't get stable.
Comparing the messy driver to that what especially Glynn achieved the
last month, it just great.
The change of socket file(s) location is not severe and well tested.
Markus
On Mon, Mar 25, 2002 at 09:16:34AM -0700, rgrmill@rt66.com wrote:
Glynn wrote:
> Note that the monitor sockets no longer live in <mapset>/.tmp/...,
but
> in /tmp/grass-<user>.
That change was made in 5.0 since pre2?
That seems like a good change, but not necessarily better then saving
under a user's ".grass" directory, which was the original option.
Hopefully this was done in such a way that it won't cause me major
update problems. Clearly while this change might be benign it should
serve as an example of how changes in GRASS could make it very difficult
to upgrade a production GRASS installation.
Further, this is a pretty fundamental change. This along with some
relatively recent changes in the monitor procedures just reinforce my
impression that GRASS5.0 is *not* yet stable.
It was really needed to keep the file path below the 104 character limit
for socket file paths. Some people were hitting it. I'd hardly
characterize it as a "fundamental" change, since most everything
couldn't care less where the file really lives. Now, the change from
LZW to ZLIB for floating point raster *was* a fundamental change. The
work Radim has been doing for GRASS 5.1 vector are fundamental. This,
however, is a small implementation detail. There was some question if
the socket file could be reliably placed under $HOME, due to remote
mounts and such... Using /tmp is a pretty safe bet to be a local
mount.
--
Eric G. Miller <egm2@jps.net>