[gdal-dev] Reading ERS format with "HeaderOffset" field

[Frank Warmerdam]

Frank Horowitz wrote:
> Hi all,
> 
> We've been trying to read some ERS ("foo.ers") files in ERMapper raw 
> raster format. While many of them work perfectly (thanks for the great 
> software, BTW!) we're finding a bug in reading some of those files which 
> incorporate the "HeaderOffset" field.
> 
> The obvious thing is happening when reading a file with such a header: 
> the values are shifted right "HeaderOffset" pixels, and the header is 
> being read as data in the image.
> 
> Looking through the code, I had a go a hacking (emphasis on hacking) a 
> simple patch:
...
> +/* -------------------------------------------------------------------- */
> /*      Establish the data type.                                        */
> /* -------------------------------------------------------------------- */
>     GDALDataType eType;
> @@ -776,9 +786,9 @@
>                 poDS->SetBand(
>                     iBand+1,
>                     new RawRasterBand( poDS, iBand+1, poDS->fpImage,
> -                                       iWordSize * iBand * 
> poDS->nRasterXSize,
> +                                       headerOffset + iWordSize * iBand 
> * poDS->nRasterXSize,
>                                        iWordSize,
> -                                       iWordSize * nBands * 
> poDS->nRasterXSize,
> +                                       headerOffset + iWordSize * 
> nBands * poDS->nRasterXSize,
>                                        eType, bNative, TRUE ));
>             }
>         }
> 
> The trouble is, I'm shooting blind, without understanding details of 
> GDAL's data model, API, codebase, etc. etc. etc. This patch does not 
> appear to work, with results ranging from simple failure to account for 
> the header when actually reading such a file to segfaults.

The RawRasterBand() arguments are:

                  RawRasterBand( GDALDataset *poDS, int nBand, FILE * fpRaw,
                                 vsi_l_offset nImgOffset, int nPixelOffset,
                                 int nLineOffset,
                                 GDALDataType eDataType, int bNativeOrder,
                                 int bIsVSIL = FALSE, int bOwnsFP = FALSE );

the only one you need to offset by the header amount is nImgOffset.
It looks like you also modified the line offset.  Try it with:

                 poDS->SetBand(
                     iBand+1,
                     new RawRasterBand( poDS, iBand+1, poDS->fpImage,
                                        headerOffset + iWordSize * iBand * 
poDS->nRasterXSize,
                                        iWordSize,
                                        iWordSize * nBands * poDS->nRasterXSize,
                                        eType, bNative, TRUE ));

If that works, please submit a ticket with the patch and I'll apply
it in the master.   A small dataset demonstrating use of the header offset
would also be helpful.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org