I checked in a fix for this; please try it out and if it still doesn't work, send me your patch. -- David
Michael Cullen wrote: > Hi all... > I have been suffering from a mean file_deleter problem on my project > ever since upgrading to (a slightly old) version of trunk. Before you > start, file_deleter hasn't changed since. I switched from server_stable > because I needed the newer CUDA stuff. > > Anyway, the problem is that if a client fails to return a workunit > because of an error (for instance, divide by zero) the file doesn't > exist, but the problem is that the directory doesn't always exist > either, so the program doesn't get as far as the "if client error, don't > expect a file to necessarilly exist" code, and as a result, it keeps > trying (no sign of the hour or even 5 second wait) to delete the file, > until it has filled the disk with it's log! The obvious choice here > would be to check if the workunit failed at the directory error stage as > well. I have a patch to do this if you're interested. > > --Michael > _______________________________________________ > boinc_dev mailing list > [email protected] > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev > To unsubscribe, visit the above URL and > (near bottom of page) enter your email address. _______________________________________________ boinc_dev mailing list [email protected] http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
