Re: Why doesn't GiST VACUUM require a super-exclusive lock, like nbtree VACUUM?

Hello, Matthias!

Updated patches attached.

Changes:
* cleanup test logic a little bit
* resolve issue with rescan in GIST (item->blkno == InvalidBlockNumber)
* move test to the main isolation suite
* add test for SpGist
* update comment I mentioned before
* allow GIST to set LP_DEAD in cases it is a safe even if LSN is updated

Also, seems like SP-GIST version is broken, it fails like this:

TRAP: failed Assert("BufferIsValid(so->pagePin)"), File:
"../src/backend/access/spgist/spgscan.c", Line: 1139, PID: 612214
FETCH(ExceptionalCondition+0xbe)[0x644a0b9dfdbc]
FETCH(spggettuple+0x289)[0x644a0b3743c6]
FETCH(index_getnext_tid+0x166)[0x644a0b3382f7]
FETCH(+0x3b392b)[0x644a0b56d92b]
FETCH(+0x3887df)[0x644a0b5427df]
FETCH(ExecScan+0x77)[0x644a0b542858]
FETCH(+0x3b3b9b)[0x644a0b56db9b]
FETCH(+0x376d8b)[0x644a0b530d8b]
FETCH(+0x379bd9)[0x644a0b533bd9]
FETCH(standard_ExecutorRun+0x19f)[0x644a0b531457]
FETCH(ExecutorRun+0x5a)[0x644a0b5312b5]
FETCH(+0x6276dc)[0x644a0b7e16dc]
FETCH(+0x628936)[0x644a0b7e2936]
FETCH(PortalRunFetch+0x1a0)[0x644a0b7e229c]
FETCH(PerformPortalFetch+0x13b)[0x644a0b49d7e5]
FETCH(standard_ProcessUtility+0x5f0)[0x644a0b7e3aab]
FETCH(ProcessUtility+0x140)[0x644a0b7e34b4]
FETCH(+0x627ceb)[0x644a0b7e1ceb]
FETCH(+0x627a28)[0x644a0b7e1a28]
FETCH(PortalRun+0x273)[0x644a0b7e12bb]
FETCH(+0x61fae1)[0x644a0b7d9ae1]
FETCH(PostgresMain+0x9eb)[0x644a0b7df170]
FETCH(+0x61b3e2)[0x644a0b7d53e2]
FETCH(postmaster_child_launch+0x137)[0x644a0b6e6e2d]
FETCH(+0x53384b)[0x644a0b6ed84b]
FETCH(+0x530f31)[0x644a0b6eaf31]
FETCH(PostmasterMain+0x161f)[0x644a0b6ea812]
FETCH(main+0x3a1)[0x644a0b5c29cf]

Best regards,
Mikhail.

>