Fix some more bugs in signal handlers and process shutdown logic.
authorTom Lane <tgl@sss.pgh.pa.us>
Sat, 1 Feb 2014 21:20:56 +0000 (16:20 -0500)
committerTom Lane <tgl@sss.pgh.pa.us>
Sat, 1 Feb 2014 21:21:23 +0000 (16:21 -0500)
commit214c7a4f0b1784ce855512c2961b09c9f51dafd8
tree4eb56ad918fddf0a219e620156916875044a5c00
parent7e1531a4504248d9df5deae8c4aa9340adaa7ef5
Fix some more bugs in signal handlers and process shutdown logic.

WalSndKill was doing things exactly backwards: it should first clear
MyWalSnd (to stop signal handlers from touching MyWalSnd->latch),
then disown the latch, and only then mark the WalSnd struct unused by
clearing its pid field.

Also, WalRcvSigUsr1Handler and worker_spi_sighup failed to preserve
errno, which is surely a requirement for any signal handler.

Per discussion of recent buildfarm failures.  Back-patch as far
as the relevant code exists.
contrib/worker_spi/worker_spi.c
src/backend/replication/walreceiver.c
src/backend/replication/walsender.c