Re: Reducing the chunk header sizes on all memory context types - Mailing list pgsql-hackers

On Tue, 30 Aug 2022 at 03:39, Tomas Vondra
<tomas.vondra@enterprisedb.com> wrote:
> The attached patch seems to fix the issue for me - at least it seems
> like that. This probably will need to get backpatched, I guess. Maybe we
> should add an assert to MemoryChunkGetPointer to check alignment?

Hi Tomas,

I just wanted to check with you if you ran the full make check-world
with this patch?

I don't yet have a working ARM 32-bit environment to test, but on
trying it with x86 32-bit and adjusting MAXIMUM_ALIGNOF to 8, I'm
getting failures in test_decoding. Namely:

test twophase                     ... FAILED       51 ms
test twophase_stream              ... FAILED       25 ms

 INSERT INTO test_prepared2 VALUES (5);
 SELECT data FROM pg_logical_slot_get_changes('regression_slot', NULL,
NULL, 'include-xids', '0', 'skip-empty-xacts', '1');
+WARNING:  problem in slab Change: detected write past chunk end in
block 0x58b8ecf0, chunk 0x58b8ed08
+WARNING:  problem in slab Change: detected write past chunk end in
block 0x58b8ecf0, chunk 0x58b8ed58
+WARNING:  problem in slab Change: detected write past chunk end in
block 0x58b8ecf0, chunk 0x58b8eda8
+WARNING:  problem in slab Change: detected write past chunk end in
block 0x58b8ecf0, chunk 0x58b8edf8
+WARNING:  problem in slab Change: detected write past chunk end in
block 0x58b8ecf0, chunk 0x58b8ee48

I think the existing sentinel check looks wrong:

if (!sentinel_ok(chunk, slab->chunkSize))

shouldn't that be passing the pointer rather than the chunk?

David