Check wakeup_flags from wait_latch function to exit on `WL_POSTMAST…

…ER_DEATH` (#1938)

There are some cases when the `wait_latch` loop in background worker is
not being interrupted after the postmaster was exited.

It can be reproduced by running postgres directly using
`/opt/homebrew/opt/postgresql@17/bin/postgres -D
/opt/homebrew/var/postgresql@17` and after the background worker will be
started you will need to send `SIGKILL` to the postmaster process. (the
issue is not reproducible when managing postgres via `pg_ctl`, I have
encountered this issue during local development as homebrew service runs
postgres directly using `postgres` binary)

Then you can see that the postgres will be exited, but background worker
process will remain active.

> Make sure the WL_POSTMASTER_DEATH flag is set when calling that
function, and verify the return code for a prompt exit in the emergency
case that postgres itself has terminated.

(see #1938 for an example)

pgrx/src/bgworkers.rs

@@ -163,14 +163,16 @@ impl BackgroundWorker {
         unsafe {
             assert!(!pg_sys::MyBgworkerEntry.is_null(), "BackgroundWorker associated functions can only be called from a registered background worker");
         }
-        match timeout {
+        let wakeup_flags = match timeout {
             Some(t) => wait_latch(
                 t.as_millis().try_into().unwrap(),
                 WLflags::WL_LATCH_SET | WLflags::WL_TIMEOUT | WLflags::WL_POSTMASTER_DEATH,
             ),
             None => wait_latch(0, WLflags::WL_LATCH_SET | WLflags::WL_POSTMASTER_DEATH),
         };
-        !BackgroundWorker::sigterm_received()
+
+        let postmaster_died = (wakeup_flags & pg_sys::WL_POSTMASTER_DEATH as i32) != 0;
+        !BackgroundWorker::sigterm_received() && !postmaster_died
     }
 
     /// Is this `BackgroundWorker` allowed to continue?