Ncks segfault when using -d option
Command-line operators for netCDF and HDF files
Brought to you by:
zender
Menu
▾
▴
#20 Ncks segfault when using -d option
The ncks program will segfault when using the -d
option. I've tried compiling nco versions 3.0.2-1,
3.1.2, and 3.1.6 on two different systems, but this
still occurs.
Here are the build environments:
Dual Intel Xeon 3.4 Ghz (x86_64)
RHEL4 running the 2.6.9-34.ELsmp kernel
gcc version 3.4.6 20060404 (Red Hat 3.4.6-3)
Tried linking to both netcdf-3.6.2-beta4 and netcdf-
3.6.0
IBM Bladecenter Quad AMD Opteron (x86_64)
RHEL4 running the 2.6.9-34.ELsmp kernel
gcc version 3.4.5 20051201 (Red Hat 3.4.5-2)
Tried linking to both netcdf-3.6.2-beta4 and netcdf-
3.6.0
I'm not familiar with the details of the netcdf
dataset used (I'm a systems guy), but while using
strace I found it segfaults at the exact same spot if
I read in a blank file instead of the netcdf data.
What other information can I provide that would help
you out?
--Joey Mendoza
Logged In: YES
user_id=6408
Send us the exact command and error message along with the
smallest possible file which triggers the error.
If the file is to big to send, then make it available via wget.
Thanks,
Charlie
Logged In: YES
user_id=1625766
The smallest file I can find is around 30MB, you can
download it at http://vis.ucar.edu/~jam/wrfrst_d01_2005-04-
05_00.00.00_0255
The commandline and error is:
ncks -d south_north, wrfrst_d01_2005-04-05_00.00.00_0255
out.nc
Segmentation fault
Logged In: YES
user_id=6408
Your mal-formed -d argument causes the segfault.
The syntax is -d dimension,start,end,stride
The -d is meaningless without corresponding hyperslab limits.
http://nco.sf.net#-d
http://nco.sf.net#ncks
The segfault resulting from the dangling single comma had
not been reported before.
We will change it to die with a useful warning in the
next release.
In the meantime, fix your command and it should work fine.
Thanks,
Charlie
Logged In: YES
user_id=1625766
Ah, thanks for the info, I'll pass this on to the scientist.
--Joey
Logged In: YES
user_id=6408
I have committed code so that the operators die with a
useful error rather than a segfault when encountering
mal-formed hyperslab specifications of this type.
Will appear in 3.1.7. Closing this bug.
Charlie