On 04/04/2025 01:11, Tom Lane wrote:
So that's a long laundry list and we haven't even dug hard.
Is it worth it? If you feel like doing the legwork then
I'm willing to support the project, but I really wonder if
we shouldn't cut our losses and just remove the module.
(I hesitate now to look at the rest of contrib/spi/ :-()
You wrote a note that I decided to omit. As I mentioned, the patch does not even fix the cascade update problem—there are still broken cases—because it seems impossible to address it in a gentle way (the code was patched 20 years ago; it's truly legacy).
I considered removing it entirely, but that seemed too drastic a solution (and, at the very least, I don't have enough expertise to make that decision). If everything looks acceptable, I would prefer to cut the module. The check_primary_key and check_foreign functions are clearly unused, are buggy, and no one has reported any obvious problems—so refint.c can be safely removed. Autoinc.c also looks problematic.
There are some question. When should we remove the module? Should we mark it as deprecated for now and remove it later? Should we handle it in another thread? Should we apply this patch in that case?
Best regards,
Dmitrii
