On Fri, Nov 11, 2011 at 9:40 AM, Bruce Momjian <bruce@momjian.us> wrote:
> Robert Haas wrote:
>> On Thu, Nov 10, 2011 at 5:07 PM, Bruce Momjian <bruce@momjian.us> wrote:
>> > Tom Lane wrote:
>> >> Robert Haas <robertmhaas@gmail.com> writes:
>> >> > It's possible to compile the source tree with LOCK_DEBUG defined, but
>> >> > the resulting postgres promptly dumps core, due to the fact that
>> >> > user_lockmethod doesn't supply any value for trace_flag; thus, the
>> >> > first LockReleaseAll(USER_LOCKMETHOD) dereferences a NULL pointer.
>> >> > This is the result of the following commit:
>> >>
>> >> > commit 0180bd6180511875db046bf8ddcaa633a2952dfd
>> >>
>> >> +1 for just reverting that commit. ?I'm not sure how much use the
>> >> LOCK_DEBUG infrastructure has in exactly its current form, but I can
>> >> certainly imagine wanting to use it or some variant of it to debug
>> >> tough problems. ?If it's gone entirely, people would have to reinvent
>> >> most of it for that type of debugging. ?On the other side of the coin,
>> >> I don't have a clear enough use-case for it to want to spend time
>> >> right now on redesigning it, nor a clear idea of exactly what changes
>> >> might make it more useful. ?So I think we should just revert and
>> >> not spend additional effort now.
>> >
>> > I am confused. ? I thought it was lock_debug referencing user locks that
>> > was broken. ?Does lock_debug need user locks?
>>
>> It supports tracing them.
>>
>> The point is, they're not gone.
>
> Now that we know that the GUC trace_userlocks is used for advisory
> locks, should we rename it to trace_advisory_locks?
They referred to as USER_LOCKMETHOD in the code, and anyone who is
using this facility is probably also reading the code, so I think it's
clear enough as-is.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
