On 2024-Jan-29, Alvaro Herrera wrote:
> It's terrifying that SerialAdd() doesn't seem to be covered by any
> tests, though.
I realized that there's some coverage when compiling with
TEST_SUMMARIZE_SERIAL, so I tried that and it looks OK.
One other change I made was in the comment that explains the locking
order. I had put the new lock at the top, but when I tested adding some
asserts to verify that the other locks are not held, they turn out to
fire soon enough ... and the conflicting lock is the last one of that
list. So I added the new lock below it, and the SLRU lock further down,
because SerialAdd does it that way.
I pushed it now.
--
Álvaro Herrera PostgreSQL Developer — https://www.EnterpriseDB.com/
