checkpoints-test: fix comparing bad numbers

The test was comparing the number of files that the fd send to the
sd (this is the number displayed in the backup summary) to the number
of files that get restored (which is equal to the number of files that
the sd sends to the director).  This of course does not make sense and
can cause problems if the cancel does not happen fast enough.  Since
in that case the fd might continue sending files to the sd with the sd
never having the chance to send those to the director.

The fix is simple: we need to check the backup log and add up all
checkpoint batch inserts instead.  This should be equal to the number
of files that the sd send to the director.

For easier debugging I also added tracing to the test.

systemtests/tests/checkpoints/testrunner-checkpoints-on-cancel

@@ -29,6 +29,7 @@ slowjob="slow-backup-bareos-fd"
 cat <<END_OF_DATA >"$tmp/bconcmds"
 @$out /dev/null
 messages
+setdebug level=100 trace=1 all
 @$out $backup_log
 run job=$slowjob fileset=bigfileset level=Full yes
 quit
@@ -71,6 +72,7 @@ messages
 restore jobid=${slowjobid} where=$restore_directory all done yes
 wait
 messages
+setdebug trace=0 all
 quit
 END_OF_DATA
 
@@ -81,7 +83,12 @@ expect_grep "Termination:            Backup Canceled" \
             "$backup_log" \
             "Job was not canceled!"
 
-NumberOfBackedUpFiles=$(grep 'FD Files Written:       ' "$backup_log" | sed -n -e 's/^.*Written:       //p')
+NumberOfBackedUpFiles=0
+while read -a nums; do
+    for num in "${nums[@]}"; do
+	(( NumberOfBackedUpFiles += num ))
+    done
+done <<< "$(grep "entries start" "$backup_log" | cut -d" " -f12)"
 
 # Check that part of the files were written despite the cancel
 if [ "$NumberOfBackedUpFiles" -le 0 ]; then