[Gdal-dev] gdalwarp segfaults when -ot "COMPRESSION=..." used

[Maciek Sieczka]

Hi All!

Long time no troubles with Gdal, happily - till today. I wanted to resample
a 15 m, Float32, ArcInfoGrid DEM to 1 m. When no compression used it goes
fine. But when -ot "COMPRESSION=..." is used a segfault occures (for either
LZW, DEFLATE and PACKBITS). And I need the compression badly because the
output is much too big. The input, called "kpngrid", is 18 MB alltogether.
Here's the command I used:

[maciek at localhost maciek]$ gdalwarp -t_srs '+init=epsg:2180' -s_srs
'+init=epsg:2180' -rcs -tr 1 1 -ot Int16 -co COMPRESS=DEFLATE"
/home/maciek/kpngrid kpngrid_r1.tif
Creating output file that is 44370P x 23925L.
:0..Segmentation fault

However, when a small subset of my input AIG DEM is chosen, using "-te
266034 324959 268067 327060", all compression algorithms work perfect.

Here's the gdalinfo of the input:

[maciek at localhost maciek]$ gdalinfo kpngrid
Driver: AIG/Arc/Info Binary Grid
Size is 2958, 1595
Coordinate System is `'
Origin = (244061.875000,340065.125000)
Pixel Size = (15.00000000,-15.00000000)
Corner Coordinates:
Upper Left  (  244061.875,  340065.125)
Lower Left  (  244061.875,  316140.125)
Upper Right (  288431.875,  340065.125)
Lower Right (  288431.875,  316140.125)
Center      (  266246.875,  328102.625)
Band 1 Block=370x4 Type=Float32, ColorInterp=Undefined
  Min=340.000 Max=1600.000
  NoData Value=-3.40282e+38

My gdal is cvs-20050115. Configured as follows:

./configure --with-hdf4=/usr/local/lib --with-grass=/usr/local/grass-6.0.cvs
 --with-libz=internal --with-png=internal --with-libtiff=internal --with-geotiff=internal
 --with-jpeg=internal --with-gif=internal

Built with following switches, gcc 3.3.4:

LDFLAGS="-s -pipe"
CPPFLAGS="-I/usr/include/python2.3/Numeric -I/usr/include/mpatrol -I/usr/include/freetype2"
CXXFLAGS="-march=athlon-xp -O2 -mfpmath=sse -s -pipe"
CFLAGS="-march=athlon-xp -O2 -mfpmath=sse -fomit-frame-pointer -s -pipe"

My system is Mandrake 10.1, kernel 2.6.8, 512 MB RAM, 2 GB swap, athlon-xp
1400.

Can anybody guess what could be the reason of that segfault?

Thanks for any answer. And thank you Frank for GDAL!

Maciek