Bring some sanity to the trace_recovery_messages code and docs.

Per gripe from Fujii Masao, though this is not exactly his proposed patch.
Categorize as DEVELOPER_OPTIONS and set context PGC_SIGHUP, as per Fujii,
but set the default to LOG because higher values aren't really sensible
(see the code for trace_recovery()).  Fix the documentation to agree with
the code and to try to explain what the variable actually does.  Get rid
of no-op calls trace_recovery(LOG), which accomplish nothing except to
demonstrate that this option confuses even its author.

1 files changed, 11 insertions, 4 deletions

diff --git a/src/backend/utils/error/elog.c b/src/backend/utils/error/elog.c
index 2cafc31b2cd..feb0bd51fc9 100644
--- a/src/backend/utils/error/elog.c
+++ b/src/backend/utils/error/elog.c
@@ -42,7 +42,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/error/elog.c,v 1.224.2.1 2010/07/18 23:43:37 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/error/elog.c,v 1.224.2.2 2010/08/19 22:55:10 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -2798,12 +2798,19 @@ is_log_level_output(int elevel, int log_min_level)
 }
 
 /*
- * If trace_recovery_messages is set to make this visible, then show as LOG,
- * else display as whatever level is set. It may still be shown, but only
- * if log_min_messages is set lower than trace_recovery_messages.
+ * Adjust the level of a recovery-related message per trace_recovery_messages.
+ *
+ * The argument is the default log level of the message, eg, DEBUG2.  (This
+ * should only be applied to DEBUGn log messages, otherwise it's a no-op.)
+ * If the level is >= trace_recovery_messages, we return LOG, causing the
+ * message to be logged unconditionally (for most settings of
+ * log_min_messages).  Otherwise, we return the argument unchanged.
+ * The message will then be shown based on the setting of log_min_messages.
  *
  * Intention is to keep this for at least the whole of the 9.0 production
  * release, so we can more easily diagnose production problems in the field.
+ * It should go away eventually, though, because it's an ugly and
+ * hard-to-explain kluge.
  */
 int
 trace_recovery(int trace_level)