Hello, everyone! While stabilizing tests for [0] I have noticed unclear (and confusing in my opinion) behavior of markers in the isolation tester.
I have attached a test with reproducer.
There are two permutations in the test:
permutation
after(before)
before
detach1
wakeup1
detach2
wakeup2
In that case, I see expected results:
step before: <... completed>
step after: <... completed>
But changing the order of steps slightly:
permutation
after(before)
wakeup1 <------- wakeup moved here
before
detach1
detach2
wakeup2
makes "after" to be completed before "before". Does that make sense?
In my cases, there are 2-steps of injection points for each of 3 sessions,
so, it is difficult to make it bullet-proof in that case without a lot of
additional injection points.
Best regards,
Mikhail.
[0]:
https://www.postgresql.org/message-id/flat/ZnoZ6GNwkJmq-gTh%40paquier.xyz#4d13f826fb1e62860cc3ae30067bd23a
v1-0001-test-Add-isolation-test-for-injection-points-mark.patch
Description: Binary data
