INGR driver fails on bigendian systems

[warmerdam]

The INGR driver fails the ingr.py tests quite thoroughly on my MacOS X system. I believe this is a big endian issue. Gdalinfo on gdalautotest/gdrivers/8bit_rgb.cot reports "not recognised as a supported file format".

Ivan, please let me know if you have the means to try and fix this bug (ie. access to a bigendian dev system). If not perhaps Mateusz or I will need to handle this.

[Even Rouault]

The enclosed patch fixes various endianness issues. Now, gdal_info and gdal_translate work fine on 8bit_rgb.cot, 8bit_pal.cot, uint32.col on a big endian host.

I've tested it on a MIPS Debian Etch running on qemu. See ​http://www.aurel32.net/info/debian_mips_qemu.php which is a very good tutorial to install a MIPS Debian Etch under qemu. I could also provide my QCOW2 image but it's 1.2 GB large.

In the process of testing the endianness issues, I also fixed :

• various memory leaks or incorrect memory usage (SetColorPalette was not cloning the input palette)
• INGR_SetTransMatrix : it wasn't consistent with INGR_GetTransMatrix. You can see it with a gdal_translate -of INGR in.cot out.cot; gdal_info in.cot; gdal_info out.cot
• INGR_SetIGDSColors and INGR_SetEnvironColors

[Even Rouault]

PS : once gzipped, the qcow2 image is 360 MB large.

[Even Rouault]

Commited in trunk in r12101

[Even Rouault]

epimetheus now succeeds on these tests. See ​http://buildbot.osgeo.org:8500/epimetheus-full/builds/292/step-test/0