From: | Andres Freund <andres(at)anarazel(dot)de> |
---|---|
To: | pgsql-committers(at)lists(dot)postgresql(dot)org |
Subject: | pgsql: pgstat: Hide instability in stats.spec with -DCATCACHE_FORCE_REL |
Date: | 2022-04-08 01:22:58 |
Message-ID: | E1ncdKz-000osR-Lr@gemulon.postgresql.org |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-committers |
pgstat: Hide instability in stats.spec with -DCATCACHE_FORCE_RELEASE.
With -DCATCACHE_FORCE_RELEASE a few tests failed. Those were trying to test
behavior in the absence of invalidation processing and
-DCATCACHE_FORCE_RELEASE obviously adds a lot of invalidation processing. The
test already tried to handle debug_discard_caches > 0, by disabling it for
individual tests.
Instead hide potentially problematic function calls in a wrapper function that
catches the does-not-exist error. The error isn't the actually interesting
bit, it's whether the stats entry still exist afterwards.
I confirmed that the tests still catches leaked function stats if I nuke the
protections against that in pgstat_function.c.
Per buildfarm animal prion.
Discussion: https://postgr.es/m/20220407165709.jgdkrzqlkcwue6ko@alap3.anarazel.de
Branch
------
master
Details
-------
https://git.postgresql.org/pg/commitdiff/d6c0db14836cd843d589372d909c73aab68c7a24
Modified Files
--------------
src/test/isolation/expected/stats.out | 152 ++++++++++++++++----------------
src/test/isolation/expected/stats_1.out | 152 ++++++++++++++++----------------
src/test/isolation/specs/stats.spec | 38 ++++----
3 files changed, 172 insertions(+), 170 deletions(-)
From | Date | Subject | |
---|---|---|---|
Next Message | Kyotaro Horiguchi | 2022-04-08 01:32:55 | Re: API stability |
Previous Message | Tom Lane | 2022-04-07 23:52:19 | Re: API stability [was: pgsql: Fix possible recovery trouble if TRUNCATE overlaps a checkpoint.] |