pgsql: Fix --disable-spinlocks in 9.2 and 9.3 branches. - Mailing list pgsql-committers

From Tom Lane
Subject pgsql: Fix --disable-spinlocks in 9.2 and 9.3 branches.
Date
Msg-id E1asCqB-00045f-Op@gemulon.postgresql.org
Whole thread Raw
List pgsql-committers
Fix --disable-spinlocks in 9.2 and 9.3 branches.

My back-patch of the 9.4-era commit 44cd47c1d49655c5 into 9.2 and 9.3 fixed
HPPA builds as expected, but it broke --disable-spinlocks builds, because
the dummy spinlock is initialized before the underlying semaphore
infrastructure is alive.  In 9.4 and up this works because of commit
daa7527afc227443, which decoupled initialization of an slock_t variable
from access to the actual system semaphore object.  The best solution
seems to be to back-port that patch, which should be a net win anyway
because it improves the usability of --disable-spinlocks builds in the
older branches; and it's been out long enough now to not be worrisome
from a stability perspective.

Branch
------
REL9_2_STABLE

Details
-------
http://git.postgresql.org/pg/commitdiff/37f30b2510525155f4f1f94288a015163689cc69

Modified Files
--------------
src/backend/postmaster/postmaster.c |  9 +++++++
src/backend/storage/ipc/ipci.c      |  1 +
src/backend/storage/ipc/shmem.c     | 21 ++++++++++++---
src/backend/storage/lmgr/spin.c     | 52 ++++++++++++++++++++++++-------------
src/include/pg_config_manual.h      |  8 ++++++
src/include/storage/s_lock.h        |  5 +---
src/include/storage/spin.h          |  9 +++++++
7 files changed, 80 insertions(+), 25 deletions(-)


pgsql-committers by date:

Previous
From: Tom Lane
Date:
Subject: pgsql: Fix --disable-spinlocks in 9.2 and 9.3 branches.
Next
From: Tom Lane
Date:
Subject: pgsql: Further reduce the number of semaphores used under --disable-spi