Michael Paquier wrote:
> On Thu, Mar 12, 2015 at 11:23 PM, Michael Paquier wrote:
> > On Thu, Mar 12, 2015 at 6:50 AM, Alvaro Herrera wrote:
> > In this part the only difference is a spawn of cmd /c, but I don't see
> > why it is useful to spawn a new command prompt here for this case, so
> > we could just drop this part. Looking at the log history, this has
> > been added since this code creation..
>
> Taking back my words here. We definitely want to spawn a new process
> in the case of pg_regress, so I think that the best thing to do would
> be to pass the to-be-launched command to get_restrict_token(), and be
> careful with WaitForSingleObject and GetExitCodeProcess() as there are
> cases where we cannot wait for a process, so we are going to need a
> control flag, or to let the callers of get_restricted_token() do the
> wait themselves. I would think that the latter is better, additional
> opinions being welcome.
Maybe we want a specialized routine for pg_regress; maybe
get_restricted_token would call it with some args set to null or false
as appropriate. My point is not to make the other callers of
get_restricted_token more complex.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services