Second thoughts on TOAST decompression. On detecting a corrupted match tag, pglz_decompress() should just summarily return -1. Breaking out of the loop, as I did in dfc797730, doesn't quite guarantee that will happen. Also, we can use unlikely() on that check, just in case it helps.
Backpatch to v13, like the previous patch. Branch ------ REL_13_STABLE Details ------- https://git.postgresql.org/pg/commitdiff/7957e75c588c0b17210d4379afb50ea2673b0d20 Modified Files -------------- src/common/pg_lzcompress.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-)
