[Alvaro Herrera <alvherre@alvh.no-ip.org>]

Hi

I got a report from Gabriele Bartolini and team that the pg_settings
view does not get the pending_restart flag set when a setting's line is
removed from a file (as opposed to its value changed).

The explanation seems to be that GUC_PENDING_RESTART is set by
set_config_option, but when ProcessConfigFileInternal is called only to
provide data (applySettings=true), then set_config_option is never
called and thus the flag doesn't get set.

I tried the attached patch, which sets GUC_PENDING_RESTART if we're
doing pg_file_settings().  Then any subsequent read of pg_settings will
have the pending_restart flag set.  This seems to work correctly, and
consistently with the case where you change a line (without removing it)
in unpatched master.

You could argue that this is *weird* (why does reading pg_file_settings
set a flag in global state?) ... but that weirdness is not something
this patch is introducing.

-- 
Álvaro Herrera         PostgreSQL Developer  —  https://www.EnterpriseDB.com/
"Hay dos momentos en la vida de un hombre en los que no debería
especular: cuando puede permitírselo y cuando no puede" (Mark Twain)

Attachments:

0001-fix-guc.patch (text/x-diff, 720B) download
From 71fa384a6bf7aef58bdbe6d382cbc1219dbaa420 Mon Sep 17 00:00:00 2001
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
Date: Mon, 26 Jul 2021 18:35:09 -0400
Subject: [PATCH] fix guc

---
 src/backend/utils/misc/guc-file.l | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/backend/utils/misc/guc-file.l b/src/backend/utils/misc/guc-file.l
index 986ce542e3..1fe3af6284 100644
--- a/src/backend/utils/misc/guc-file.l
+++ b/src/backend/utils/misc/guc-file.l
@@ -354,6 +354,7 @@ ProcessConfigFileInternal(GucContext context, bool applySettings, int elevel)
 											  gconf->name),
 									 NULL, 0,
 									 &head, &tail);
+			gconf->status |= GUC_PENDING_RESTART;
 			error = true;
 			continue;
 		}
-- 
2.20.1