[GRASS5] grass cvs snapshot from 26. Feb.

[Eric G. Miller]

On Fri, Mar 02, 2001 at 02:39:35PM +0100, Andreas Lange wrote:
> Will the sockets setup be default for the stable 
> version?

???

> The "make install" with the sockets setup dies with 
> some error for a missing create_sockets.sh script,
> this happens on Linux, IRIX, FreeBSD and Cygwin, but obviously only 
> if ./configure --with-socket is used!

I don't know what this create_sockets.sh is about.  The sockets are
created as needed in the current mapset's .tmp directory.  Just comment
out the line in Makefile.in that looks like:

test ${XDRIVER} != socket && ${SHELL} -c "${SHELL}
${SRCDIR}/src/scripts/shells/create_${XDRIVER}s.sh ${INST_DIR}
2>/dev/null ; true"

I have no idea what the purpose of that is supposed to be, but the
"test" should be protected in an "if test ..." so it doesn't kill
the make when ${XDRIVER} == socket.

> If i issue "bash", next log into grass5 and do a "d.mon start=x0" 
> i get the following error message:
> Usage: x0 [nlev]
> No socket to connect to for monitor <x0>
> Trying again
> No socket to connect to for monitor <x0>
> ..

This problem can occur if there's a mismatch with the communications
method.  At the moment, if you do --with-socket for configure, and later
do, --with-fifo or --with-ipc, you have to make sure at least the d.mon
object files are deleted before continuing with "make".  Make can't know
it should recompile some of the d.mon sources as there's a USE_G_SOCKS
define that triggers conditional compilation in d.mon sources when
--with-socket is specified.  This conditional compile was necessary, as
the sockets drivers take at most two arguments (argv[0] and argv[1]),
where argv[0] is the name the driver was invoked as (and will be the
name used in creating the socket file) and argv[1] is the optional nlev
argument -- No paths to "fifos" are needed.

> The other tests in summary:
> 
> Linux, Red Hat 7.0, Kernel 2.2.17-7, i686, gcc 2.96:
> fifo: compiles and works. confirmed.
> ipc: compiles error free and works, confirmed.
> sockets: compiles, d.mon start=x0 does not work.
> Message:
> "No socket to connect to for monitor <x0>
> Problem selecting x0. Will try once more" (twice)

See above about cleaning d.mon and rebuilding it.

> ld32: ERROR 33: Unresolved text symbol "__INTADDR__" -- 1st referenced by
> /disk2/opt/grass.src/src/libes/LIB.mips-sgi-irix6.5/libgis.a(unix_socks.o).
>         Use linker option -v to see when and which objects, archives and
> dsos are loaded.  

Bob Covill mentioned something about needing a -lsocket linker option on
a Solaris build.  Maybe the same issue?  Guess we need another configure
check...

> FreeBSD/alpha, 4.2-RELEASE, gcc 2.95.2:
> fifo: compiles and works, confirmed.
> ipc: compiles error free except for NVIZ, works OK, confirmed. Slow.
> Problem with backing store?
> sockets: compiles and installs ok, but does not work:
>  d.mon start=x0
>  Usage: x0 [nlev]
>  No socket to connect to for monitor <x0>.
>  Problem selecting x0. Will try once more
>  No socket to connect to for monitor <x0>.

See above about cleaning d.mon sources first.

> Windows 2000 Professional, 5.00.2195, Cygwin 1.1.8, i686, gcc 2.95.2-6:
> fifo: compiles ok, working not possible
> ipc: compiles ok so far, does not work.
> sockets: does not work.
> 
> (on each machine i did:
> tar xzf grass_src, ./configure; make; make install for default fifo setup
> then:
> ./README.xdriver fifo|ipc|socket ; make ; make install)

Don't do this for sockets. You must do ./configure --with-socket.
Otherwise a symbol needed for d.mon conditional compile will not be
defined.  And, as already mentioned, d.mon sources need to be cleaned of
object files.

Yes, it's all pretty messy right now...

-- 
Eric G. Miller <egm2 at jps.net>

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