Closed cjvogl closed 8 years ago
The reason for discarding files that don't appear to be from the latest run is so that they don't accidentally contaminate _PlotIndex.html
or an animation, in cases when you do a run with fewer output frames than the previous run.
It's checking for later files in the sequence that were closed more than 5 seconds before an earlier file. Are you seeing cases where e.g. fort.q0005 is closed more than 5 seconds before fort.q0004? For a large run that takes a long time to output, I would think this would be less of a problem than for a smaller run.
I think my case is a unique one: I have fort.q0136 with a timestamp of 16:42... and fort.q0137 with a timestamp of 11:50. This happened in two runs on HYAK, so either HYAK travelled back in time or there was a problem with the system clock. It seems to be somewhat isolated (ie there are runs from HYAK that don't have this problem). I see now that in runclaw, old fort files aren't removed if overwrite
is set to false, and now the utility of the check makes more sense to me. Thanks!
In frametools.py, line 1428, there's code to account for fort files that are closed "out of order":
delaytime = 5
. With some large runs, I'm seeing more than 5 seconds inbetween consecutive fort files. Is there a reason for this check with the output directory being cleared of fort files before each run?