* Peter Geoghegan (pg@heroku.com) wrote:
> I seem to also recall
> Coverity correctly handling that, although perhaps I'm unfairly
> crediting them with taking advantage of the abort() trick because of
> the state of Postgres when I tried each of those two tools - it might
> be that scan-build *would* have taken advantage of that at the time,
> if only the trick was there.
With Coverity, we completely 'punt' on it because we make elog(FATAL)
into a 'program-exit' case when it clearly isn't. I've discussed this
w/ the Coverity folks but, aiui, there's no solution to this any time
soon..
Thanks,
Stephen