On Sat, Aug 22, 2020 at 8:33 PM Tom Lane wrote:
>
> Amit Kapila writes:
> > On Thu, Aug 20, 2020 at 10:28 PM Tom Lane wrote:
> >> 2. On the other hand, the code is *releasing* the
> >> ReplicationSlotControlLock before it calls
> >> ProcArraySetReplicationSlotXmin, and that seems like a flat out
Amit Kapila writes:
> On Thu, Aug 20, 2020 at 10:28 PM Tom Lane wrote:
>> 2. On the other hand, the code is *releasing* the
>> ReplicationSlotControlLock before it calls
>> ProcArraySetReplicationSlotXmin, and that seems like a flat out
>> concurrency bug.
> It is not clear to me how those value
On Thu, Aug 20, 2020 at 10:28 PM Tom Lane wrote:
>
> While trying to make sense of Adam Sjøgren's problem [1], I found
> myself staring at ReplicationSlotsComputeRequiredXmin() in slot.c.
> It seems to me that that is very shaky code, on two different
> grounds:
>
> 1. Sometimes it's called with P
