"Magnus Hagander" <mha(at)sollentuna(dot)net> writes:
> The problem is that write_nondefault_variables() writes out variables of
> type PGC_INTERNAL. I've attached a patch that simply prevents it from
> doing this. It does make the problem go away, and I *think* it should be
> safe, but I'm not totally sure...
The issue then is whether there are other mechanisms in place to ensure
that backends correctly inherit the values associated with these
variables...
regards, tom lane