Fixed, thanks for the report!

This is my code, if anyone is interested, but I believe that the problem is 
almost certainly in downward:


      (let* ((outfile (merge-pathnames "downward-output" scratch-dir))
             (status (excl:run-shell-command
                      (format nil "~a ~{--heuristic '~a' ~}--search '~a' < 
output"
                              search-prog
                              heuristic search)
                      :output outfile :error-output :output
                      :directory scratch-dir)))
	(when verbose
          (format t "Search ~:[failed~;succeeded~].  Output from program was:~%"
                  (zerop status))
          (dump-file outfile))
	(unless (zerop status)
          (error "Downward fails..."))

I have been running downward wrapped in a function that checks for its exit code 
(apologies for the CL source), where I check to see if the return status is 
zero.

On terminate and preprocess, my code correctly detects errors.  However, when I 
get an out-of-memory error running downward, my code incorrectly thinks that the 
planner has run successfully (but failed to find a plan):


terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
warning: could not determine peak memory
Peak memory: -1 KB
caught signal 6 -- exiting
/home/rpg/stratus/stratus-trunk/code/externals/fast-
downward/src/search/downward: line 184: 28602 Aborted                 "$PLANNER" 
"$@" \
< $TEMPFILE

I believe that this is because the downward bash script doesn't check exit 
status and pass it on.

I am no one's idea of a bash programmer, but I will see if I can fix this and 
provide a patch.