Gordon Rowell wrote: > It certainly is across all *nix variants - it's a fairly standard way to > ensure cleanup of temporary files.
If we are going to go this route, I'm certainly not interested in recreating File::Temp (any further than we already have). It's trivial to replace our existing temp_file and temp_dir with File::Temp's methods instead, and for good measure we could use unlink0() to do what you are suggesting. The tricky part is how to handle an error condition when we would want to keep a copy of the file. I want to get a much better handle on why we aren't cleaning up right now (though it may just be a byproduct of using 'svc -t' to restart the server). Only then can I make any sort of rational guess at what point in the transaction should failure retain a copy of the message: ~ post-data up to queue? ~ new hook available for plugin authors? ~ global parameter? ~ others I haven't thought of yet? John
