Mapserver compilation flags should be added to the SWIG command line

[Steve Lime]

Ok, I did remove the USE_GEOS dependency in shape.i. For the couple of
functions
in question the msGEOS... functions will still produce results even if
GEOS support
is not included. If there is no non-GEOS workaround then the GEOS
functions will 
return an error about "No GEOS support..." and throw an error.

Feels kinda weird producing results from GEOS-named functions if GEOS
isn't 
enabled. We could move those wrappers elsewhere with a non-specific name
e.g. msIntersects calls msGEOSIntersects and alternatively provides a
brute-force
method, but that change could happen later if desired.

What bug are we tracking this under? 1881 doesn't seem quite right.

Steve

>>> "Umberto Nicoletti" <umberto.nicoletti at gmail.com> 08/30/06 9:30 AM
>>>
Hi Steve,
While doing the usual house keeping before 4.10 gets out I have looked
at the source of the swig interface files and found that they contain
preprocessor defines (#ifdef USE_GEOS for example).
Most of the mapscript laguages will ignore these defines because swig
is not passed the list of options built by configure. Java mapscript
used to pass the list to swig, but this feature was removed a long
time ago because Sean opted for removing the preprocessor defines from
the swig interface files to keep them tidier.
C# mapscript at this point is the only laguage still passing defines to
swig.
Perl, Ruby and Python invoke swig without passing the defines; also,
in the case of Perl and Python swig must be run manually by the user
(by typing the right command in the shell).

Now since most mapscript will ignore those preprocessor defines we are
facing a potentially critical issue and, in the case of GEOS,
mapscript won't use some GEOS functions.
Other possible errors could be memory errors due to the use of non
initialized variables (this is the case of POINT_Z_M).

I have reviewed the swig interface files and found these possibly
harmful defines:

swiginc/image.i:#if GD2_VERS > 1

Not sure about this one, maybe it is right. Dunno where GD2_VERS is
defined.

swiginc/image.i:#ifdef USE_GD_GIF
swiginc/image.i:#ifdef USE_GD_PNG
swiginc/image.i:#ifdef USE_GD_JPEG
swiginc/image.i:#ifdef USE_GD_WBMP

These were added by Sean in July 04 and should be removed, I guess.
They only affect TCL.

swiginc/point.i:#ifdef USE_POINT_Z_M
swiginc/point.i:#ifdef USE_POINT_Z_M
swiginc/point.i:#ifdef USE_POINT_Z_M
swiginc/point.i:#ifdef USE_POINT_Z_M
swiginc/point.i:#ifdef USE_POINT_Z_M
swiginc/point.i:#ifdef USE_POINT_Z_M

These are *rarely* used and they are not a real problem, at least for
4.10. If used they crash Java mapscript for sure.

swiginc/shape.i:#ifdef USE_GEOS
swiginc/shape.i:#ifdef USE_GEOS

These were added by sdlime on his lates rework of geos code this june.
They should be moved out of the swig interface file into the c code of
mapserver

I hope I made myself clear,
Umberto

On 8/28/06, Steve Lime <Steve.Lime at dnr.state.mn.us> wrote:
> Confused, what's the desired fix?
>
> >>> Umberto Nicoletti <umberto.nicoletti at GMAIL.COM> 8/28/2006 10:47:08
> AM >>>
> I have posted a similar mail today (search for Defines in swig
> interface files in the subject) . Tamas, sorry for the double posting,
> I missed this one.
>
> I appear to recall that it was once agreed that swig interface files
> should not have defines in them, but I at the moment the only
> reference I found is here:
>
> http://mapserver.gis.umn.edu/bugs/show_bug.cgi?id=895#c2
> (end of first para)
>
> Sean Gillies probably knows better.
>
> At this point C# is the only language that has 100% correct
> compilation today. Java mapscript had the flags passed to swig, but
> they were removed a long time ago.
> Other laguages like ruby and python require the user to generate the
> interface by running swig by hand. This makes it much difficult to
> pass defines to swig as the user should have to copy them from the
> Makefile or alternatively we should provide and mantain Makefiles for
> all mapscripts.
>
> This is the list of defines (still) in the swig interface files:
>
> swiginc/image.i:#if GD2_VERS > 1
>
> Not sure about this one, maybe it is right.....
>
> swiginc/image.i:#ifdef USE_GD_GIF
> swiginc/image.i:#ifdef USE_GD_PNG
> swiginc/image.i:#ifdef USE_GD_JPEG
> swiginc/image.i:#ifdef USE_GD_WBMP
>
> These were added by Sean in July 04.
>
> swiginc/point.i:#ifdef USE_POINT_Z_M
> swiginc/point.i:#ifdef USE_POINT_Z_M
> swiginc/point.i:#ifdef USE_POINT_Z_M
> swiginc/point.i:#ifdef USE_POINT_Z_M
> swiginc/point.i:#ifdef USE_POINT_Z_M
> swiginc/point.i:#ifdef USE_POINT_Z_M
>
> These are rarely used and they are not a real problem, at least for
> 4.10
>
> swiginc/shape.i:#ifdef USE_GEOS
> swiginc/shape.i:#ifdef USE_GEOS
>
> These were added by sdlime on his lates rework of geos code this june.
>
> I can try to take care of the defines in image.i and maybe steve of
> those in shape.i, but I dunno if I'm going or if we should do it for
> for 4.10?
>
> Umberto
>
>
> On 8/28/06, Daniel Morissette <dmorissette at mapgears.com> wrote:
> > Tamas Szekeres wrote:
> > > Folks,
> > >
> > > Currently most of the SWIG bindings does not add mapserver
> compilation
> > > flags to the
> > > SWIG command line and therefore the interface might not be
> generated
> > > properly.
> > > For example missing -DUSE_POINT_Z_M will prevent from generating z
> and m
> > > members
> > > for pointObj etc.
> > >
> > > I've added a bug for this issue:
> > > http://mapserver.gis.umn.edu/bugs/show_bug.cgi?id=1881
> > >
> > > And fixed for C#
> > >
> >
> > Does something need to be done for this for the other wrappers?
> There
> > used to be a 'mapscriptvars' file with all the compile flags used by
> the
> > Perl Makefile. Is this not used by the other bindings?
> >
> > Daniel
> > --
> > Daniel Morissette
> > http://www.mapgears.com/
> >
>