[postgis-devel] Can we put back GEOS 3.5 support in 3.0?

[Paul Ramsey]

Yeah, I wonder, should symbol re-naming-fu be something on the TODO list? In the end, the world will be ruled by fat binaries, so…

P

> On Jan 23, 2019, at 3:15 PM, Even Rouault <even.rouault at spatialys.com> wrote:
> 
>> One of these days I'm going to propose that QGIS just bundles the
>> whole of the latest GEOS stable release inside the QGIS repo (maybe as
>> a git submodule) so that we're guaranteed to have the very latest GEOS
>> release alongside QGIS. It's honestly the only way forward that I can
>> see working. Maybe PostGIS should consider the same...
> 
> Beware of runtime clashes if you do that.
> Imagine that QGIS builds against its GEOS-latest, and also links against GDAL 
> from the distribution which links against GEOS-older. Runtime crash is ensured 
> / likely, at least on Linux due to how the dynamic loader works by agressively 
> using previously loaded symbols (that is it might make function A of GEOS-
> latest call function B of GEOS-older if GEOS-older was loaded before)
> 
> In GDAL we had similar situation a few years ago when the bundled libtiff in 
> GDAL was libtiff 4.X and most distributions shipped with libtiff 3.X. To 
> isolate things, the bundled copy in GDAL renames symbols (which was done with 
> a "#define foo GDALfoo" trick for all non-hidden symbols).
> 
> Even
> 
> -- 
> Spatialys - Geospatial professional services
> http://www.spatialys.com
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/postgis-devel