I prefer that LuceneTestCase will not assume anything about how it's being run. E.g. if someone writes a script, not through the runner, which forks JVMs, it would be good if LTC was still safe.
I don't mind giving mkdirs() a try. I think that it's better than createNewFile() -> delete -> mkdirs(). Shai On Wed, Dec 19, 2012 at 10:54 PM, Dawid Weiss <[email protected]>wrote: > > Well, mkdirs() does state that it returns true iff all path components > were > > created, false otherwise. It doesn't say anything about atomicity, which > > makes me nervous ... > > I'd assume it's the filesystem that guarantees this. Or so I hope. > > > But if others think that's fine, then this will be more resilient to > > multiple JVMs sharing the same temp.dir. > > But this should never be the case -- we're physically splitting cwd's > of all forked JVMs and any temporary folders should be created inside > those (the security manager and Uwe are making sure you can't write > out of your assigned directory)? > > D. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
