On Fri, Jun 26, 2015 at 4:02 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Combining this with my idea about preserving the ConfigVariable list,
> I'm thinking that it would be a good idea for ProcessConfigFile() to
> run in a context created for the purpose of processing the config files,
> rather than blindly using the caller's context, which is likely to be
> a process-lifespan context and thus not a good place to leak in.
> We could keep this context around until the next SIGHUP event, so that
> the ConfigVariable list remains available, and then destroy it and
> replace it with the next ProcessConfigFile's instance of the context.
> In this way, any leakage in the processing code could not accumulate
> over multiple SIGHUPs, and so it would be certain to remain fairly
> negligible.
That seems like a nice idea.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company