[gdal-dev] Crash on cancel overviews creation

[Even Rouault]

Javier,

Hopefully https://github.com/OSGeo/gdal/pull/4347 should solve that

Even

Le 26/08/2021 à 18:45, Javier Jimenez Shaw a écrit :
> Hi
>
> Some times (unfortunately it is not reproducible every time) we got a 
> crash when we cancel the overview generation.
> I tried to reproduce it in my Ubuntu 20.04, but couldn't. My 
> colleagues are able to reproduce some times on Mac and Windows.
>
> In C++, with GDAL 3.3.1, calling BuildOverviews("AVERAGE", ...) to 
> produce embedded overview and calling the cancel callback, it crashes 
> on overviews.cpp:
>
> ~PointerHolder() { CPLFree(ptr); }. (line 3929)
> The reason of the crash is : Thread 34: EXC_BAD_ACCESS (code=1, 
> address=0x1d)
> Yes, the debugger says the value of "ptr" is 0x1d... that does not 
> make so much sense.
>
> the debugger shows this steps on the back trace:
> poJob->oDstBufferHolder.reset(new PointerHolder(poJob->pDstBuffer));  
> (line 4930)
> GDALRegenerateOverviewsMultiBand (line 4638)
>
>
> Before calling to BuildOverviews I am calling
> CPLSetConfigOption("GDAL_NUM_THREADS", "ALL_CPUS");
> to make it multithreaded.
>
> The image is an RGBA 8bit with JPEG compression.
>
> Is anything wrong there?
>
> Thanks.
> .___ ._ ..._ .. . ._. .___ .. __ . _. . __..  ... .... ._ .__
> Entre dos pensamientos racionales
> hay infinitos pensamientos irracionales.
>
>
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/gdal-dev

-- 
http://www.spatialys.com
My software is free, but my time generally not.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20210826/646a5b3f/attachment.html>