Geometry Boost List: [ggl] suspicious line 320 in extensions/gis/projections/proj/stere.hpp

Subject: [ggl] suspicious line 320 in extensions/gis/projections/proj/stere.hpp
From: Barend Gehrels (barend.gehrels)
Date: 2010-11-11 09:09:57

hi Kris,

Good catch!

On 11-11-2010 14:28, Krzysztof Czainski wrote:
> Hello,
>
> I have been playing with converting Geographic <-> Cartesian, tried 
> projections. When I tried stere_ellipsoid, I found a warning on line 
> 320. That was yesterday at work, and I don't remember the warning, but 
> it said something about UB.
>
> Anyway, when I look at this line in the current (r66490) 
> sandbox/geometry, it looks suspicious, so I changed it locally:
> $ svn diff stere.hpp
>
>  Index: stere.hpp
> ===================================================================
> --- stere.hpp    (wersja 66490)
> +++ stere.hpp    (kopia robocza)
> @@ -316,7 +316,7 @@
>              void setup_stere(Parameters& par, par_stere& proj_parm)
>              {
>                  proj_parm.phits = pj_param(par.params, "tlat_ts").i ?
> -                    proj_parm.phits = pj_param(par.params, 
> "rlat_ts").f : HALFPI;
> +                    pj_param(par.params, "rlat_ts").f : HALFPI;
>                  setup(par, proj_parm);
>              }
>
> What do You guys think?

See
http://svn.osgeo.org/metacrs/proj/trunk/proj/src/PJ_stere.c

Our projection sources are converted automatically from proj4. The 
source of proj4 already contains this possible suspicious construct.

However, this is, though suspicious and redundant, correct C code 
because "the assignment x = y returns the value of the assignment". But 
indeed it looks very ugly.

I will report this to proj4, because if this is committed at our side, 
the same construct will be there again after re-generation. But in the 
meantime we can commit it. Will do that.

Thanks, Barend