gdalwarp corrupted unsorted chunks on Centos 5.5 x86_64

[randrewspmx]

We have a gdal installation on 64bit Centos 5.5. Here is a list of the relevant version information:

gdal-1.7.3 gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-48)

gdalwarp crashes on occasion when invoked inside a shell script that is called when new data arrives on the image processing system. This leaves hanged instances of gdalwarp building up on the host until we manually killall old instances of gdalwarp.

See attached backtrace and memory map pulled from syslog on the host.

[randrewspmx]

gdalwarp backtrace and memory map of crashed instance

[warmerdam]

based on the "ERROR 1: TIFFReadDirectory:Cannot handle different values per sample for "BitsPerSample"" message I believe we would also need your data file(s) and commandline arguments in order to investigate the problem properly. It seems to be data dependent.

[randrewspmx]

${GDALDIR}gdalwarp -of Gtiff -s rcnodata "0 0 0" -dstalpha -te -144.667007 32.960999 -45.602901 82.636597 ${TEMPDIR}${DESTNAME}.tif -ts 1280 720 -rcs ${TEMPDIR}${DES TNAME}_canada.tif

[warmerdam]

I have been unable to reproduce this problem, even running gdalwarp 1.7.x under valgrind. I suspect it relates to the version of libtiff used on CentOS. I would suggest building your own GDAL (1.8) with the internal libtiff instead of the system libtiff and see if that works properly.

[randrewspmx]

Attached log of build from source. It can be seen that the internal TIFF code was used during the build.