[GRASS5] [bug #2391] (grass) r.patch: creating support files fails when 1st map isn't local

[Glynn Clements]

Request Tracker wrote:

> Subject: r.patch: creating support files fails when 1st map isn't local
> 
> When running r.patch with "input=aspect at first_mapset,aspect at second_mapset",
> building support files fails with the following errors:
> 
> r.patch: percent complete:  100%
> CREATING SUPPORT FILES FOR aspect
> WARNING: unable to find [aspect] in [(null)]
> WARNING: category support for [aspect] in mapset [(null)] missing

main() calls G_raster_map_type(), which calls G_find_cell(), which
overwrites the qualified name with the unqualified version. The
unqualified names are then passed to the support() function.

Personally, I don't think that G_raster_map_type() ought to behave
like this. If someone was willing to check that none of the code which
calls it relies upon this behaviour, we could change it to use
G_find_cell2() instead.

More generally, we should eliminate the use of G_find_file in library
functions (other than G_find_cell etc). I suspect that this problem
isn't limited to r.patch.

But, so long as G_raster_map_type() has its current behaviour, the
only solution is to duplicate the elements of opt1->answers so that
the original values are available to support().

-- 
Glynn Clements <glynn.clements at virgin.net>