Hamish:
> found it,
> https://trac.osgeo.org/grass/ticket/654
>
> I traced this back to r37726 where wind_format.c's
> format_double() uses G_projection() which wants to check
> the projection type (and that hasn't been created yet).
> Not sure of the best solution yet.
Glynn:
I suggest simply reverting r37726. Changing a function so
that it requires GRASS to have been initialised ($GISRC,
WIND, etc) when it didn't previously is a fundamental API
change.
I really hesitate to do that, because then lat/lon support in
the GUIs will again be subject to loss of precision for
resolution and the rounding errors that go along with that.
(%.8f is not enough to preserve the resolution once converted
to DMS and back again)
instead my proposed solution is to remove G_projection() from
format_double() [satisfying your concerns] and allow programmers
to specify "-1" for the target projection option to get full
%.15g instead of lying to the function about the projection type
in order to get the desired format.
Hamish