Alvaro Herrera wrote:
> Background worker processes
I had tested this on EXEC_BACKEND some time ago, and it worked fine, but
I had neglected since then, and now I find it fails with a pretty
strange message on startup. Also, Andres and I have been talking about
other possible problems in that scenario (mostly that the order in which
shared libraries are loaded might not be deterministic, and this would
cause the whole approach to fall down).
I'm considering reverting this; but since it won't cause a visible
problem unless and until a worker is loaded, leaving it in place might
enable someone else to peek at it while I come up with some idea to
handle the broken EXEC_BACKEND case ...
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services