pg_stat_statements: Fix handling of duplicate constant locations
authorÁlvaro Herrera <alvherre@kurilemu.de>
Wed, 29 Oct 2025 11:35:02 +0000 (12:35 +0100)
committerÁlvaro Herrera <alvherre@kurilemu.de>
Wed, 29 Oct 2025 11:35:02 +0000 (12:35 +0100)
commit16edc1b94fc2db6e6a376471e280b50a418907c2
tree19a8654301046f90ae6fb3ff6aa773421efa7e27
parent94f95d91b025cb6752b4118bb0b30851e3d64df9
pg_stat_statements: Fix handling of duplicate constant locations

Two or more constants can have the same location.  We handled this
correctly for non squashed constants, but failed to do it if squashed
(resulting in out-of-bounds memory access), because the code structure
became broken by commit 0f65f3eec478: we failed to update 'last_loc'
correctly when skipping these squashed constants.

The simplest fix seems to be to get rid of 'last_loc' altogether -- in
hindsight, it's quite pointless.  Also, when ignoring a constant because
of this, make sure to fulfill fill_in_constant_lengths's duty of setting
its length to -1.

Lastly, we can use == instead of <= because the locations have been
sorted beforehand, so the < case cannot arise.

Co-authored-by: Sami Imseih <samimseih@gmail.com>
Co-authored-by: Dmitry Dolgov <9erthalion6@gmail.com>
Reported-by: Konstantin Knizhnik <knizhnik@garret.ru>
Backpatch-through: 18
Discussion: https://www.postgresql.org/message-id/2b91e358-0d99-43f7-be44-d2d4dbce37b3%40garret.ru
contrib/pg_stat_statements/expected/squashing.out
contrib/pg_stat_statements/pg_stat_statements.c
contrib/pg_stat_statements/sql/squashing.sql