Re: [saga-gis-developer] Attempt to read past EOF

[Volker W. <wic...@la...>]

Hi Lucas,

thanks a lot for your debugging and fixing and for providing the 
patches! I've just applied them in trunk in order to facilitate further 
fixes if necessary.

Thanks,
Volker


On 06/06/2011 09:32 PM, Lucas C. Villa Real wrote:
> Hi Volker,
>
> As I stated in the email below, the NoData was indeed the problem with
> the cropped subset of the DEM that I had to use. However, another
> problem appeared when I switched to the intermediary  DEM, which is
> single SAGA grid of 19Gb: the resulting images, after processed by the
> "Fill Sinks (Wang & Liu)" module, are all black.
>
> I spent the weekend debugging the issue and I came last night to a fix.
> There are a couple of places in SAGA where integer overflow may happen.
> In my test bed, the DEM resolution is 71452 x 36739, and the result of
> that multiplication alone already overflows a 32-bit signed integer. The
> CSG_Grid class in special had the most pertinent issues with regard to that.
>
> I've splitted my fixes in a series of 6 patches. I'm submitting them to
> the SourceForge's Patches tracker page.
>
> Thanks,
> Lucas
>
> On Sun, Jun 5, 2011 at 12:13 AM, Lucas C. Villa Real
> <lu...@go... <mailto:lu...@go...>> wrote:
>
>     Hi Volker,
>
>     The NoData value was indeed the problem here. It was set to
>     -340282346638528859811704183484516925440.000, and that value was
>     interpreted as being a valid DEM elevation when generating the color
>     palette for the output image. In the end, the color range assigned
>     to the actual data was  very small and caused the output image to
>     look bogus.
>
>     Thanks much for your help, once again :-)
>
>     Lucas
>
>
>
>     On Thu, Jun 2, 2011 at 9:52 AM, Volker Wichmann
>     <wic...@la... <mailto:wic...@la...>> wrote:
>
>         Hi Lucas,
>
>         good to hear that you could track down the GDAL problem.
>
>         Regarding the "Fill Sinks (Wang & Liu)" module, I could imagine
>         this to
>         be a problem with the NoData value used in your grid. In case it
>         isn't
>         -99999.0 you could try to reclassify it to that value.
>
>         Volker
>
>         On 06/02/2011 04:42 AM, Lucas C. Villa Real wrote:
>          > Hi Volker,
>          >
>          > I just found that GDAL has an issue when seeking to offsets >
>         4Gb. It
>          > looks like the offset calculation in
>         SAGARasterBand::IReadBlock() and
>          > SAGARasterBand::IWriteBlock() returns a temporary signed
>         32-bit value,
>          > and that value is sent to GDAL's file system abstraction
>         layer's seek()
>          > method. On the platform I'm running this code on, that method
>         expects to
>          > receive an unsigned 64-bit value as offset. Fortunately, the
>         fix is
>          > simply a matter of applying the proper cast to ensure the
>         result doesn't
>          > get truncated by the compiler.
>          >
>          > I'll send out a patch to the GDAL community. Thanks for your
>         suggestion,
>          > it helped much to narrow the problem.
>          >
>          > The only pending issue I have now is that the "Fill Sinks
>         (Wang & Liu)"
>          > module is not generating a reasonable set of basins -- that
>         must have
>          > something to do with the data format, which is Float32. I'll
>         investigate
>          > that now.
>          >
>          > Cheers,
>          > Lucas
>          >
>
>         ------------------------------------------------------------------------------
>         Simplify data backup and recovery for your virtual environment
>         with vRanger.
>         Installation's a snap, and flexible recovery options mean your
>         data is safe,
>         secure and there when you need it. Data protection magic?
>         Nope - It's vRanger. Get your free trial download today.
>         http://p.sf.net/sfu/quest-sfdev2dev
>         _______________________________________________
>         saga-gis-developer mailing list
>         sag...@li...
>         <mailto:sag...@li...>
>         https://lists.sourceforge.net/lists/listinfo/saga-gis-developer
>
>
>
>
>     --
>     Lucas
>     "If you're looking for a reason I've a reason to give: pleasure,
>     little treasure"
>
>
>
>
> --
> Lucas
> "If you're looking for a reason I've a reason to give: pleasure, little
> treasure"