On 12 January 2016 at 13:12, Felipe Gasper <felipe at felipegasper.com> wrote:
> Same code, just different processes. > > We?ve just noted over the years with race conditions that that ?if it can > fail, it will?, so we try to be as bulletproof as we can. > Good policy :) After you unlink the temp file, I presume the schema creating process opens a new connection against the permanent file? I can't see your algorithm causing corruption, unless your affected users are using a filesystem which doesn't propagate POSIX file region locks across hard links. -Rowan