On Mon, 25 Jun 2007, Malcolm J. Currie wrote:
> While running a few KAPPA tests so unusual errors appeared.
>
> % makesurface $KAPPA_DIR/m31 like=$KAPPA_DIR/m31
> OUT - NDF to contain surface > m5
> !! Search of free chip stack for a frame match exceeded stack size
> ! DAT_ERASE: Error erasing an HDS structure component.
> ! SUBPAR: Failed to find locator for action MAKESURFACE
> ! SUBPAR: Failed to find locator for parameter OUT
> ! SUBPAR: Failed to update GLOBAL file for parameter OUT
> ! Application exit status DAT__WEIRD, Unknown error
> ! /home/mjc/classic/bin/kappa/m31 like=/home/mjc/classic/bin/kappa/m31
> % nomagic n43 out=n7 repval=0
in the good old days you would have got a segv there so at least we've
made some progress :-) It's interesting that the corruption always looks
the same. Don't know enough about HDS internals to comment on the
significance of this though.
> There is a problem with my GLOBAL.sdf. After moving GLOBAL.sdf, the
> above two commands work without error. I've tried ERASEing a few things
> that would seem common or recently added, but there are many GLOBAL
> variables. Any idea what's wrong and how to identify specifically the
> offending data? My concern is that a KAPPA task (and my suspicion is
> LUTABLE) could be corrupting the global parameter file. The offending
> GLOBAL.sdf is attached.
>
Try running valgrind with --track-fds so that it will report open files on
exit. Need to do this in conjunction with setting HDS_DISABLE_CLEANUP
environment variable (that I added last week) which disables the tidy up
exit handler. Are you running the patched HDS which stops GLOBAL.sdf
corrupting in the pipeline? (but I don't think it would affect command
line usage)?
--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj
|