On 12/17/2013 03:57 AM, Kevin Grittner wrote:
> I agree with others that this patch is not suitable for
> back-patching. I wonder whether some other solution might be worth
> back-patching, since it is clearly a bug. The "never delete
> internal pages" might be safe enough to consider. It would lead to
> some degree of bloat, but perhaps bloat is better than errors. Of
> course, the argument can certainly be made that people hit the bug
> so rarely that we're better off just leaving it alone in stable
> branches. Anyway, that would be a different patch.
I don't think "never delete internal pages" would be acceptable. In
particular, it would bloat indefinitely for a smallish FIFO queue table
with a timestamp column.
- Heikki