Re: Pike 8.0: Sql.pgsql objects don't get garbage collected?

Jeff Hungerford wrote:
>I've been testing with HEAD at 4887dfb, and overall things look
>pretty good; I have run into a couple of errors while doing a long
>running test of opening and destroying database connections over and
>over.

>The first is another case of access to an object after it has been destructed:

>-=-=-=-=-
>Cannot access variables in destructed object.
>/usr/local/pike/8.0.175/lib/modules/Sql.pmod/pgsql.pike:1198:
>    unknown function()
>-=-=-=-=-

That seems to be a race due to threading and destructs running
in parallel.  Fixed.

>The second appears to be something odd when throwing the magicterminate
>value as an error:

>-=-=-=-=-
><Invalid backtrace/error container: 42>
>/usr/local/pike/8.0.175/lib/master.pike:4416: describe_backtrace(42,80)
>/usr/local/pike/8.0.175/lib/master.pike:1355:
>    master()->call_describe_backtrace(42,UNDEFINED)
>    /usr/local/pike/8.0.175/lib/master.pike:1699: master()->handle_error(42)
>-=-=-=-=-

That is odd, I prevent the MAGICTERMINATE value from being rethrown
now; that should prevent this problem entirely.

Patches checked into 8.0 and 8.1.
-- 
Stephen.