Simplify handling of the timezone GUC by making initdb choose the default.

We were doing some amazingly complicated things in order to avoid running
the very expensive identify_system_timezone() procedure during GUC
initialization.  But there is an obvious fix for that, which is to do it
once during initdb and have initdb install the system-specific default into
postgresql.conf, as it already does for most other GUC variables that need
system-environment-dependent defaults.  This means that the timezone (and
log_timezone) settings no longer have any magic behavior in the server.
Per discussion.

Author: tglsfdc

doc/src/sgml/config.sgml

@@ -3915,9 +3915,10 @@ FROM pg_stat_activity;
         Sets the time zone used for timestamps written in the server log.
         Unlike <xref linkend="guc-timezone">, this value is cluster-wide,
         so that all sessions will report timestamps consistently.
-        If not explicitly set, the server initializes this variable to the
-        time zone specified by its system environment.  See <xref
-        linkend="datatype-timezones"> for more information.
+        The built-in default is <literal>GMT</>, but that is typically
+        overridden in <filename>postgresql.conf</>; <application>initdb</>
+        will install a setting there corresponding to its system environment.
+        See <xref linkend="datatype-timezones"> for more information.
         This parameter can only be set in the <filename>postgresql.conf</>
         file or on the server command line.
        </para>
@@ -4963,9 +4964,10 @@ SET XML OPTION { DOCUMENT | CONTENT };
       <listitem>
        <para>
         Sets the time zone for displaying and interpreting time stamps.
-        If not explicitly set, the server initializes this variable to the
-        time zone specified by its system environment.  See <xref
-        linkend="datatype-timezones"> for more information.
+        The built-in default is <literal>GMT</>, but that is typically
+        overridden in <filename>postgresql.conf</>; <application>initdb</>
+        will install a setting there corresponding to its system environment.
+        See <xref linkend="datatype-timezones"> for more information.
        </para>
       </listitem>
      </varlistentry>

doc/src/sgml/datatype.sgml

@@ -2286,7 +2286,7 @@ January 8 04:05:06 1999 PST
     but continue to be prone to arbitrary changes, particularly with
     respect to daylight-savings rules.
     <productname>PostgreSQL</productname> uses the widely-used
-    <literal>zoneinfo</> time zone database for information about
+    <literal>zoneinfo</> (Olson) time zone database for information about
     historical time zone rules.  For times in the future, the assumption
     is that the latest known rules for a given time zone will
     continue to be observed indefinitely far into the future.
@@ -2432,26 +2432,9 @@ January 8 04:05:06 1999 PST
      The <xref linkend="guc-timezone"> configuration parameter can
      be set in the file <filename>postgresql.conf</>, or in any of the
      other standard ways described in <xref linkend="runtime-config">.
-     There are also several special ways to set it:
+     There are also some special ways to set it:
 
      <itemizedlist>
-      <listitem>
-       <para>
-        If <varname>timezone</> is not specified in
-        <filename>postgresql.conf</> or as a server command-line option,
-        the server attempts to use the value of the <envar>TZ</envar>
-        environment variable as the default time zone.  If <envar>TZ</envar>
-        is not defined or is not any of the time zone names known to
-        <productname>PostgreSQL</productname>, the server attempts to
-        determine the operating system's default time zone by checking the
-        behavior of the C library function <literal>localtime()</>.  The
-        default time zone is selected as the closest match among
-        <productname>PostgreSQL</productname>'s known time zones.
-        (These rules are also used to choose the default value of
-        <xref linkend="guc-log-timezone">, if not specified.)
-       </para>
-      </listitem>
-
       <listitem>
        <para>
         The <acronym>SQL</acronym> command <command>SET TIME ZONE</command>

doc/src/sgml/ref/set.sgml

@@ -239,9 +239,7 @@ SELECT setseed(<replaceable>value</replaceable>);
          <listitem>
           <para>
            Set the time zone to your local time zone (that is, the
-           server's default value of <varname>timezone</>; if this
-           has not been explicitly set anywhere, it will be the zone that
-           the server's operating system defaults to).
+           server's default value of <varname>timezone</>).
           </para>
          </listitem>
         </varlistentry>

src/backend/bootstrap/bootstrap.c

@@ -333,10 +333,6 @@ AuxiliaryProcessMain(int argc, char *argv[])
 	{
 		if (!SelectConfigFiles(userDoption, progname))
 			proc_exit(1);
-		/* If timezone is not set, determine what the OS uses */
-		pg_timezone_initialize();
-		/* If timezone_abbreviations is not set, select default */
-		pg_timezone_abbrev_initialize();
 	}
 
 	/* Validate we have been given a reasonable-looking DataDir */

src/backend/commands/variable.c

@@ -259,23 +259,6 @@ check_timezone(char **newval, void **extra, GucSource source)
 	char	   *endptr;
 	double		hours;
 
-	if (*newval == NULL)
-	{
-		/*
-		 * The boot_val given for TimeZone in guc.c is NULL.  When we see this
-		 * we just do nothing.	If this isn't overridden from the config file
-		 * then pg_timezone_initialize() will eventually select a default
-		 * value from the environment.	This hack has two purposes: to avoid
-		 * wasting cycles loading values that might soon be overridden from
-		 * the config file, and to avoid trying to read the timezone files
-		 * during InitializeGUCOptions().  The latter doesn't work in an
-		 * EXEC_BACKEND subprocess because my_exec_path hasn't been set yet
-		 * and so we can't locate PGSHAREDIR.
-		 */
-		Assert(source == PGC_S_DEFAULT);
-		return true;
-	}
-
 	/*
 	 * Initialize the "extra" struct that will be passed to assign_timezone.
 	 * We don't want to change any of the three global variables except as
@@ -374,7 +357,7 @@ check_timezone(char **newval, void **extra, GucSource source)
 				return false;
 			}
 
-			if (!tz_acceptable(new_tz))
+			if (!pg_tz_acceptable(new_tz))
 			{
 				GUC_check_errmsg("time zone \"%s\" appears to use leap seconds",
 								 *newval);
@@ -427,10 +410,6 @@ assign_timezone(const char *newval, void *extra)
 {
 	timezone_extra *myextra = (timezone_extra *) extra;
 
-	/* Do nothing for the boot_val default of NULL */
-	if (!myextra)
-		return;
-
 	session_timezone = myextra->session_timezone;
 	CTimeZone = myextra->CTimeZone;
 	HasCTZSet = myextra->HasCTZSet;
@@ -490,20 +469,8 @@ check_log_timezone(char **newval, void **extra, GucSource source)
 {
 	pg_tz	   *new_tz;
 
-	if (*newval == NULL)
-	{
-		/*
-		 * The boot_val given for log_timezone in guc.c is NULL.  When we see
-		 * this we just do nothing.  If this isn't overridden from the config
-		 * file then pg_timezone_initialize() will eventually select a default
-		 * value from the environment.
-		 */
-		Assert(source == PGC_S_DEFAULT);
-		return true;
-	}
-
 	/*
-	 * Otherwise assume it is a timezone name, and try to load it.
+	 * Assume it is a timezone name, and try to load it.
 	 */
 	new_tz = pg_tzset(*newval);
 
@@ -513,7 +480,7 @@ check_log_timezone(char **newval, void **extra, GucSource source)
 		return false;
 	}
 
-	if (!tz_acceptable(new_tz))
+	if (!pg_tz_acceptable(new_tz))
 	{
 		GUC_check_errmsg("time zone \"%s\" appears to use leap seconds",
 						 *newval);
@@ -538,10 +505,6 @@ check_log_timezone(char **newval, void **extra, GucSource source)
 void
 assign_log_timezone(const char *newval, void *extra)
 {
-	/* Do nothing for the boot_val default of NULL */
-	if (!extra)
-		return;
-
 	log_timezone = *((pg_tz **) extra);
 }
 

src/backend/postmaster/postmaster.c

@@ -795,21 +795,6 @@ PostmasterMain(int argc, char *argv[])
 	 */
 	CreateDataDirLockFile(true);
 
-	/*
-	 * If timezone is not set, determine what the OS uses.	(In theory this
-	 * should be done during GUC initialization, but because it can take as
-	 * much as several seconds, we delay it until after we've created the
-	 * postmaster.pid file.  This prevents problems with boot scripts that
-	 * expect the pidfile to appear quickly.  Also, we avoid problems with
-	 * trying to locate the timezone files too early in initialization.)
-	 */
-	pg_timezone_initialize();
-
-	/*
-	 * Likewise, init timezone_abbreviations if not already set.
-	 */
-	pg_timezone_abbrev_initialize();
-
 	/*
 	 * Initialize SSL library, if specified.
 	 */

src/backend/tcop/postgres.c

@@ -3537,10 +3537,6 @@ PostgresMain(int argc, char *argv[], const char *username)
 	{
 		if (!SelectConfigFiles(userDoption, progname))
 			proc_exit(1);
-		/* If timezone is not set, determine what the OS uses */
-		pg_timezone_initialize();
-		/* If timezone_abbreviations is not set, select default */
-		pg_timezone_abbrev_initialize();
 	}
 
 	/*

src/backend/utils/error/elog.c

@@ -1805,24 +1805,20 @@ setup_formatted_log_time(void)
 {
 	struct timeval tv;
 	pg_time_t	stamp_time;
-	pg_tz	   *tz;
 	char		msbuf[8];
 
 	gettimeofday(&tv, NULL);
 	stamp_time = (pg_time_t) tv.tv_sec;
 
 	/*
-	 * Normally we print log timestamps in log_timezone, but during startup we
-	 * could get here before that's set. If so, fall back to gmt_timezone
-	 * (which guc.c ensures is set up before Log_line_prefix can become
-	 * nonempty).
+	 * Note: we expect that guc.c will ensure that log_timezone is set up
+	 * (at least with a minimal GMT value) before Log_line_prefix can become
+	 * nonempty or CSV mode can be selected.
 	 */
-	tz = log_timezone ? log_timezone : gmt_timezone;
-
 	pg_strftime(formatted_log_time, FORMATTED_TS_LEN,
 	/* leave room for milliseconds... */
 				"%Y-%m-%d %H:%M:%S     %Z",
-				pg_localtime(&stamp_time, tz));
+				pg_localtime(&stamp_time, log_timezone));
 
 	/* 'paste' milliseconds into place... */
 	sprintf(msbuf, ".%03d", (int) (tv.tv_usec / 1000));
@@ -1836,19 +1832,15 @@ static void
 setup_formatted_start_time(void)
 {
 	pg_time_t	stamp_time = (pg_time_t) MyStartTime;
-	pg_tz	   *tz;
 
 	/*
-	 * Normally we print log timestamps in log_timezone, but during startup we
-	 * could get here before that's set. If so, fall back to gmt_timezone
-	 * (which guc.c ensures is set up before Log_line_prefix can become
-	 * nonempty).
+	 * Note: we expect that guc.c will ensure that log_timezone is set up
+	 * (at least with a minimal GMT value) before Log_line_prefix can become
+	 * nonempty or CSV mode can be selected.
 	 */
-	tz = log_timezone ? log_timezone : gmt_timezone;
-
 	pg_strftime(formatted_start_time, FORMATTED_TS_LEN,
 				"%Y-%m-%d %H:%M:%S %Z",
-				pg_localtime(&stamp_time, tz));
+				pg_localtime(&stamp_time, log_timezone));
 }
 
 /*
@@ -1947,14 +1939,11 @@ log_line_prefix(StringInfo buf, ErrorData *edata)
 			case 't':
 				{
 					pg_time_t	stamp_time = (pg_time_t) time(NULL);
-					pg_tz	   *tz;
 					char		strfbuf[128];
 
-					tz = log_timezone ? log_timezone : gmt_timezone;
-
 					pg_strftime(strfbuf, sizeof(strfbuf),
 								"%Y-%m-%d %H:%M:%S %Z",
-								pg_localtime(&stamp_time, tz));
+								pg_localtime(&stamp_time, log_timezone));
 					appendStringInfoString(buf, strfbuf);
 				}
 				break;

src/backend/utils/misc/guc-file.l

@@ -292,7 +292,6 @@ ProcessConfigFile(GucContext context)
 	if (context == PGC_SIGHUP)
 	{
 		InitializeGUCOptionsFromEnvironment();
-		pg_timezone_initialize();
 		pg_timezone_abbrev_initialize();
 		/* this selects SQL_ASCII in processes not connected to a database */
 		SetConfigOption("client_encoding", GetDatabaseEncodingName(),

src/backend/utils/misc/guc.c

@@ -187,6 +187,7 @@ static bool check_log_stats(bool *newval, void **extra, GucSource source);
 static bool check_canonical_path(char **newval, void **extra, GucSource source);
 static bool check_timezone_abbreviations(char **newval, void **extra, GucSource source);
 static void assign_timezone_abbreviations(const char *newval, void *extra);
+static void pg_timezone_abbrev_initialize(void);
 static const char *show_archive_command(void);
 static void assign_tcp_keepalives_idle(int newval, void *extra);
 static void assign_tcp_keepalives_interval(int newval, void *extra);
@@ -2547,7 +2548,7 @@ static struct config_string ConfigureNamesString[] =
 			NULL
 		},
 		&log_timezone_string,
-		NULL,
+		"GMT",
 		check_log_timezone, assign_log_timezone, show_log_timezone
 	},
 
@@ -2827,7 +2828,7 @@ static struct config_string ConfigureNamesString[] =
 			GUC_REPORT
 		},
 		&timezone_string,
-		NULL,
+		"GMT",
 		check_timezone, assign_timezone, show_timezone
 	},
 	{
@@ -3817,7 +3818,7 @@ InitializeGUCOptions(void)
 	 * Before log_line_prefix could possibly receive a nonempty setting, make
 	 * sure that timezone processing is minimally alive (see elog.c).
 	 */
-	pg_timezone_pre_initialize();
+	pg_timezone_initialize();
 
 	/*
 	 * Build sorted array of all GUC variables.
@@ -4114,6 +4115,15 @@ SelectConfigFiles(const char *userDoption, const char *progname)
 	 */
 	SetConfigOption("data_directory", DataDir, PGC_POSTMASTER, PGC_S_OVERRIDE);
 
+	/*
+	 * If timezone_abbreviations wasn't set in the configuration file, install
+	 * the default value.  We do it this way because we can't safely install
+	 * a "real" value until my_exec_path is set, which may not have happened
+	 * when InitializeGUCOptions runs, so the bootstrap default value cannot
+	 * be the real desired default.
+	 */
+	pg_timezone_abbrev_initialize();
+
 	/*
 	 * Figure out where pg_hba.conf is, and make sure the path is absolute.
 	 */
@@ -8444,8 +8454,11 @@ assign_timezone_abbreviations(const char *newval, void *extra)
  * This is called after initial loading of postgresql.conf.  If no
  * timezone_abbreviations setting was found therein, select default.
  * If a non-default value is already installed, nothing will happen.
+ *
+ * This can also be called from ProcessConfigFile to establish the default
+ * value after a postgresql.conf entry for it is removed.
  */
-void
+static void
 pg_timezone_abbrev_initialize(void)
 {
 	SetConfigOption("timezone_abbreviations", "Default",

src/backend/utils/misc/postgresql.conf.sample

@@ -406,7 +406,7 @@
 #log_temp_files = -1			# log temporary files equal or larger
 					# than the specified size in kilobytes;
 					# -1 disables, 0 logs all temp files
-#log_timezone = '(defaults to server environment setting)'
+#log_timezone = 'GMT'
 
 
 #------------------------------------------------------------------------------
@@ -486,7 +486,7 @@
 
 #datestyle = 'iso, mdy'
 #intervalstyle = 'postgres'
-#timezone = '(defaults to server environment setting)'
+#timezone = 'GMT'
 #timezone_abbreviations = 'Default'     # Select the set of available time zone
 					# abbreviations.  Currently, there are
 					#   Default

src/bin/initdb/.gitignore

@@ -1,4 +1,5 @@
 /encnames.c
 /pqsignal.c
+/localtime.c
 
 /initdb