[GRASS-user] Error running g.mapset

[Glynn Clements]

Markus Neteler wrote:

> > The underlying bug (init.bat doesn't set GIS_LOCK) is still present in
> > 6.5-svn. It isn't applicable to 7.0, where init.sh/init.bat have been
> > replaced by a Python script (which does set GIS_LOCK).
> 
> I have found this GPL program to the the PID which is needed
> to set GIS_LOCK:
> http://www.scheibli.com/projects/getpids/
> 
> Would that be of interest? Or is %RANDOM% together with something
> else sufficient?

AFAICT, GIS_LOCK needs to be set to an integer, but the actual value
is irrelevant on Windows.

g.mapset reads GIS_LOCK, passes it to $GISBASE/etc/lock, and checks
that etc/lock returns a zero exit code.

On Windows, etc/lock does nothing:

	int main(int argc, char *argv[])
	{
	...
	    if (argc != 3 || sscanf(argv[2], "%d", &lockpid) != 1)
		G_fatal_error("usage: %s file pid", argv[0]);
	...
	#ifdef __MINGW32__
	    G_warning("Concurrent mapset locking is not supported on Windows");
	    exit(0);
	#else

If GIS_LOCK isn't set, g.mapset will complain. If it doesn't parse
with "%d", etc/lock will complain. Other than that, it shouldn't
matter.

-- 
Glynn Clements <glynn at gclements.plus.com>