On Sat, Jan 02, 2021 at 12:31:45PM +0500, Andrey Borodin wrote: > Do I understand correctly that this is bugfix that needs to be back-patched?
The slru-truncate-modulo patch fixes a bug. The slru-truncate-t-insurance patch does not. Neither _needs_ to be back-patched, though I'm proposing to back-patch both. I welcome opinions about that. > Thus we should not refactor 4 identical *PagePrecedes(int page1, int page2) > into 1 generic function? I agree with not refactoring that way, in this case. > Since functions are not symmetric anymore, maybe we should have better names > for arguments than "page1" and "page2"? At least in dev branch. That works for me. What names would you suggest? > Is it common practice to embed tests into assert checking like in > SlruPagePrecedesUnitTests()? No; it's neither common practice nor a policy breach.