On Thu, 16 Aug 2012, Even Rouault wrote:
The rgdal source is part C (for proj), part C++ for OGR/GDAL, but C++
written by C-minded people, so no new classes, but using C++ OGR/GDAL
classes, things like: pDataset->~GDALDataset();.
Ah ok, so rgdal uses the GDAL C++ API (when feasable, the use of the C API is
encouraged to avoid ABI issues).
This was Tim Keitt's design choice in 2002. On reflection, the C API might
be an improvement, when time permits. Maybe a GSoC project next year!
By the way pDataset->~GDALDataset() is a weird syntax. I've never used
that in any C++ code, so I am wondering if this completely destroys the
object. I suspect that the code of the destructor is called, but
potentially not its super destructor, so there must be a small memory
leak. Why not using delete pDataset, or better, to avoid cross-heap
issues on Windows, GDALClose((GDALDatasetH)pDataset) ?
Once again, Tim's choice. Thanks for the tip about alternatives!
As I understand you,
1.x.a and 1.x.b share the C++ ABI, but say 1.x.b and 1.y.a will probably
not.
Yes, 1.x.a and 1.x.b share the same C++ ABI, but other combinations
might not.
This suggests that I may use GDALCheckVersion(GDAL_VERSION_MAJOR,
GDAL_VERSION_MINOR, NULL); to get something that might help users debug
performance failures, but that GDAL_VERSION_REV is too fine-grained.
Yes.
Good, implemented and committed to SVN.
--
Roger Bivand
Department of Economics, NHH Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: [email protected]
_______________________________________________
gdal-dev mailing list
[email protected]
http://lists.osgeo.org/mailman/listinfo/gdal-dev
