Re: SSI and Hot Standby

"When I test your example, though, I'm getting the serialization
failure on T3 rather than T2, so I'd call that a bug. Will
investigate. Thanks again for your tests! You seem to be able to
shake out issues better than anyone else! Once found, fixing them
is not usually very hard, it's coming up with that creative usage
pattern to *find* the problem which is the hard part."

Thank you very much, but I do not deserve this honor. I was just constructing an example for myself so that I could understand why read only transaction might pose a problem. I posted it to help other people to see a concrete example of the problem. I had no idea this would show an actual bug in the code.

"OK if I add this one to our dcheck test suite, too?"

It is of course OK. And if you want to add this as an example in the documentation, it would be great. This is a simple, but concrete example of why read only serializable transaction might cause an anomaly. If I am not mistaken, there isn't any complete example in the documentation. It was hard for me to grasp why there might be a problem and I don't think I am alone.

- Anssi