[gdal-dev] MSG Driver: two memory leaks

[Simon Richard Proud]

Hello,
We actually stopped using the GDAL MSG driver some time ago (as it is very slow) and wrote our own one instead. My mail was mainly to find out if others have experienced similar problems.

The problem is not from the EUMETSAT libraries, by the way. We have been using them with very small memory use (they only decompress the data). The memory use is coming from somewhere in the reflectance calculations. They seem well hidden but, as Christian mentioned, a valgrind could well find them.

All the best,
Simon
________________________________________
Fra: Even Rouault [even.rouault at mines-paris.org]
Sendt: 26. juni 2014 16:48
Til: gdal-dev at lists.osgeo.org
Cc: Simon Richard Proud
Emne: Re: [gdal-dev] MSG Driver: two memory leaks

Le jeudi 26 juin 2014 16:14:03, Simon Richard Proud a écrit :
> Hi,
> Have you noticed any other memory leaks in the MSG driver? We have seen
> some huge memory uses when working with the visible MSG channels
> (converting from DN to Reflectance) but have not been able to track down
> where the memory is being used. Would be good to have a second opinion!

Simon,

I just wanted to mention I can be available for contractual support on GDAL if
you need.

It might not be to be excluded that the leak might be in the Eumetsat library
itself.

Best regards,

Even


>
> All the best,
> Simon
>
> ________________________________________
> Fra: gdal-dev-bounces at lists.osgeo.org [gdal-dev-bounces at lists.osgeo.org]
> på vegne af Christian Authmann [authmanc at Mathematik.Uni-Marburg.de]
> Sendt: 26. juni 2014 05:23
> Til: gdal-dev at lists.osgeo.org
> Emne: [gdal-dev] MSG Driver: two memory leaks
>
> Hello,
>
> valgrind'ing my application revealed two memory leaks in msgdataset.cpp.
>
> a) MSGDataset::pszProjection shall always point to a valid string during
> the lifetime of the object; the old string needs to be released before
> getting a new one from exportToWkt().
>
> b) A spatial reference is exported to a temporary string as WKT, but the
> string is never released.
>
>
> Attached you'll find a patch fixing both; valgrind confirms that both
> memory leaks are gone.
>
>
> --
> Christian Authmann
> Philipps-Universität Marburg
> Fachbereich Mathematik und Informatik
> AG Datenbanksysteme
> Hans-Meerwein-Straße
> D-35032 Marburg
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/gdal-dev

--
Geospatial professional services
http://even.rouault.free.fr/services.html