Em qui., 18 de set. de 2025 às 13:40, Tom Lane <tgl@sss.pgh.pa.us> escreveu:
I wrote: > Sigh ... so the answer is this used to work (since commit 39df0f150) > and then I carelessly broke it in commit a391ff3c3. If you try this > test case in versions 9.5..11 you get a spot-on rowcount estimate. > Serves me right for not having a test case I guess, but I'm astonished > that nobody complained sooner.
The attached fixes things so it works like it did pre-a391ff3c3.
I spent some time trying to devise a test case, and was reminded of why I didn't have one before: it's hard to make a case that will be robust enough to not show diffs in the buildfarm. I'll keep thinking about that though.
One question, make difference return -1.0 (float point notation)?