[mapserver-dev] Dropping support for non-GDAL and non-PROJ builds [was Re: Modernizing code base: C99 and maybe C++ ?]

Seth G sethg at geographika.co.uk
Wed Nov 27 05:38:01 PST 2019

Hi Even,

I ran into this the other day when I had the failing FastCGI issue and wanted to debug without Proj to see if that was the cause. 
MapServer currently fails to compile unless both -DWITH_PROJ=1 and -DWITH_GDAL=1 are set. 

Turning off GDAL also means the following have to be turned off:


I then get the following compile error: 

msGetStringListFromHashTable - identifier not found (possibly related to the new layer->connectionoptions)

Turning off PROJ means the following options also have to be turned off:


I then get the following compile error: 

'msApproxTransformer': undeclared identifier

Fixing these may of course not resolve all issues. 

I agree we should either require PROJ and GDAL, or fix the current errors and add a new CI build without these options to ensure they don't become broken again. 

I like the idea of being able to compile a "lightweight" map rendering engine, but on the other hand I've never required one..
It might be worth having a poll on the users list?

Of course by default now both GDAL and PROJ are required so support has effictively been dropped on master at least. If users/companies require them then they'll need to supply a pull request or funding. At present it looks like only a couple of fixes would be needed to allow these builds.

Hope all is well - I'm not sure I've seen a rant by you before!


twitter: @geographika

On Wed, Nov 27, 2019, at 1:26 AM, Even Rouault wrote:
> And while I'm a rant-style RFC mode, what about dropping those #ifdef USE_GDAL 
> and #ifdef USE_PROJ that polutes the code base ? And we don't actually CI test 
> those configurations, hence my mantra "untested code is broken code"
> Who relies on non-PROJ and non-GDAL enabled code ? And if they are a few that 
> do, do we care about them ? We don't need to be nice with everyone. That 
> doesn't even buy us free beers.
> (I love those rant-style RFCs. They are so pleasant to write :-) And quite to 
> the point I think !)
> Even
> -- 
> Spatialys - Geospatial professional services
> http://www.spatialys.com
> _______________________________________________
> mapserver-dev mailing list
> mapserver-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/mapserver-dev

More information about the mapserver-dev mailing list