On 26 Feb 2002, Craig Hughes moaned: > FYI, your code example is not even close to concurrency-safe. If one
Oh, it's an astonishing kludge written in haste and isn't meant to be right. What's more, it doesn't help :( Further testing indicates that it's the previous state of the db that's at fault; a tiny script that just ties to the DB_File as spamassassin does and then unties works if it is given an empty database from a previous tie/untie call-pair, but falls over (with the silly incorrect error message) if it's handed a database of the kind that, say, spamassassin under the control of t/db_based_whitelist.t creates. I'm going to gdb my way through libdb-2.7.7; I'm fairly sure that it's that and not spamassassin at fault, now. Sorry for the false alarm. > instance checks (-f $path) and finds no file, then context switches, and > instance #2 performs the same check, then both instances will try to > create a new file, and you'll probably end up with some weird mangled DB > file. I seem to be getting a pretty mangled db file as it is, even with only one instance and the testbeds :( -- `Frankly I wonder whether you are not writing your posts from underneath a bridge.' --- Jason Clifford, to a particularly dense troll _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
