Re: BUG #18146: Rows reappearing in Tables after Auto-Vacuum Failure in PostgreSQL on Windows

On Sat, Dec 21, 2024 at 12:15:56PM +1300, Thomas Munro wrote:
> Here. Better/tidier ideas welcome.

Indeed, the state of back-branches is incorrect this way. I don't
know of any out-of-core callers of smgrtruncate(), but well, the world
is a wide place.

extern void smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks,
- BlockNumber *old_nblocks,
BlockNumber *nblocks);
+extern void smgrtruncate2(SMgrRelation reln, ForkNumber *forknum, int nforks,
+ BlockNumber *old_nblocks,
+ BlockNumber *nblocks);

Please don't rely on my naming sense, still I'm OK with what you are
using here.

Some inconsistent notes in the code with smgrtruncate2() in place:
src/backend/access/heap/visibilitymap.c:
* otherwise the caller is responsible for calling smgrtruncate()
src/backend/catalog/storage.c:
* Second, the call to smgrtruncate() below will in turn call
src/backend/access/heap/vacuumlazy.c:
* processed the smgr invalidation that smgrtruncate sent out ... but

That's not really critical, so leaving them as they are is equally OK
for me. Looks good to me, otherwise.
--
Michael