On Thu, Sep 29, 2016 at 8:46 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> We could probably refactor things enough so that we do pq_init()
> before PGSharedMemoryReAttach(). It would be a little bit ugly,
> and it would fractionally increase the chance of a reattach failure
> because pq_init() palloc's a few KB worth of buffers. I'm not quite
> sure if it's worth it; thoughts? In any case the mentioned comments
> are obsolete and need to be moved/rewritten.
Just speaking off the cuff without reviewing the code in detail...
Alternately we could call pq_init in the error path if it hasn't been
called yet. I'm sure there are problems with doing that in general but
for the specific errors that can happen before pq_init it might be
feasible since they obviously can't have very much shared state yet to
have corrupted.
--
greg