[geos-devel] Splitting _r and non _r functions in 2 different .h ?
even.rouault at mines-paris.org
Sat Jun 28 08:17:23 PDT 2014
Le vendredi 20 juin 2014 09:55:32, Sandro Santilli a écrit :
> Please also take a look at this pending pull request for
> the re-entrant API: https://github.com/libgeos/libgeos/pull/26
> It's blocked on naming, for the same reason you're doing this work
> (avoiding confusion)
Hum, I've had a look at the pull request but I'm highly confused by all the
commits and comments, so I didn't pursue further.
For now, I've just reordered the symbols as discussed. See
http://trac.osgeo.org/geos/ticket/695 for the proposed patch.
I've tested it successfully when compiling GDAL and MapServer.
> On Wed, Jun 18, 2014 at 06:52:58PM +0200, Sandro Santilli wrote:
> > On Wed, Jun 18, 2014 at 06:32:25PM +0200, Even Rouault wrote:
> > > I've hit an integration problem with QGIS using the non _r API and
> > > spatialite calling (accidently) finishGEOS() when closing a spatialite
> > > DB, causing later crashes in QGIS.
> > > So I've converted all the QGIS calls to GEOS to the _r API, and to do
> > > that the easiest solution was to alter geos_c.h to remove the non _r
> > > symbols and fix all compilation problems. But to make that robust in
> > > the length, it would be good to be upstreamed.
> > > GDAL and MapServer also now use the _r API for similar reasons. And it
> > > would also be good to be sure noone reintroduce calls to the non _r
> > > API.
> > I was actually thinking to deprecate the non _r API at some point,
> > but I've been too lazy to port PostGIS to the _r ones so far, so
> > I keep adding the non _r ones even for new signature :/
> > --strk;
> geos-devel mailing list
> geos-devel at lists.osgeo.org
Geospatial professional services
More information about the geos-devel