Bug#691273: Acknowledgement (netcdf-bin: nccopy produces bogus output)

2012-10-23 Thread Paul Nienaber
It appears that this bug is triggered by attempting to overwrite an existing NetCDF file. We had previously created an output file of the same name with `ncks`, and nccopy appeared to overwrite it correctly, but instead left us with a mangled file, which should not happen. -- To UNSUBSCRIBE,

Bug#691273: netcdf-bin: nccopy produces bogus output

2012-10-23 Thread Paul Nienaber
Package: netcdf-bin Version: 1:4.1.3-6 Severity: grave Justification: renders package unusable It appears that nccopy generates bogus (numerically, and structurally). If this is not patched locally to Debian, this should be pushed upstream. Resulting files appear to have all missing values for

Bug#591854: (no subject)

2010-08-05 Thread Paul Nienaber
After further thought, depending on what is or isn't being set in userspace at present, this may be something that should be upstreamed to kernel folks, as setting this for init is kind of an obvious thing to do even if userspace is dumb and lacks fancy start-up infrastructure/etc. I.e., it sh

Bug#591854: (no subject)

2010-08-05 Thread Paul Nienaber
The values visible in and settable by /proc/*/oom_adj determine how vulnerable processes are to being OOM-killed. Accordingly, init's should be set to -17, the "safest" possible value. init getting OOM-killed results in the system either hanging or rebooting, which is very bad. The package '

Bug#591854: sysvinit: init oom_adj value set to 0 not -17

2010-08-05 Thread Paul Nienaber
Package: sysvinit Version: 2.88dsf-11 Severity: critical Justification: causes serious data loss *** Please type your report below this line *** Under no circumstances should init have its oom_adj set to anything other than -17. watchdog and other facilities have this set properly, it seems.