[PATCH] tracd should log to syslog when daemonized

[edahlman@…]

When running tracd -d it would be really nice if it logged to syslog so that one could look at what is going on as it runs.

[Christian Boos]

Well, you have other valid choices:

• file
• winlog on Windows.

But maybe we should warn if the mode is stderr.

[Matthew Good]

Replying to cboos:

But maybe we should warn if the mode is stderr.

You'd have to open each project being served by tracd in the initialization before it's been daemonized, which seems kind of ugly. Since "stderr" is not the default logging mode I wouldn't think it's common for people to accidentally set the mode to "stderr" when running tracd as a daemon.

[Christian Boos]

Replying to mgood:

Replying to cboos:

But maybe we should warn if the mode is stderr.

You'd have to open each project being served by tracd in the initialization before it's been daemonized …

An alternative solution would be to fallback to syslog if stderr can't be used:

--- a/trac/env.py       Mon Aug 28 10:37:17 2006 +0200
+++ b/trac/env.py       Wed Sep 06 09:15:18 2006 +0200
@@ -15,6 +15,7 @@
 # Author: Jonas Borgström <jonas@edgewall.com>

 import os
+import sys

 from trac import db_default, util
 from trac.config import *
@@ -263,6 +264,8 @@ class Environment(Component, ComponentMa
         logfile = self.log_file
         if logtype == 'file' and not os.path.isabs(logfile):
             logfile = os.path.join(self.get_log_dir(), logfile)
+        if logtype == 'stderr' and not sys.stderr.isatty():
+            logtype = 'syslog'
         self.log = logger_factory(logtype, logfile, self.log_level, self.path)

     def get_known_users(self, cnx=None):

[Christian Boos]

An alternative solution would be to fallback to syslog if stderr can't be used:

trac/env.py

@@ -263 +264 @@
         logfile = self.log_file
         if logtype == 'file' and not os.path.isabs(logfile):
             logfile = os.path.join(self.get_log_dir(), logfile)
+        if logtype == 'stderr' and not sys.stderr.isatty():
+            logtype = 'syslog'

Scratch that, stderr could redirected to a file or a pipe (tee).

Well, since the original request, nobody has shown much interest in that feature which can be easily "fixed" by using a proper configuration.

[Marcin Wisnicki <mwisnicki@…>]

Unless I'm missing something, you can only configure per instance/environment logging but global messages (like "Server starting in PID") go either to stdout or null if daemonized ?

[Peter Suter]

So this is not about "logging" and more about print statements and the call​ to daemonize​ that redirects them to /dev/null.

Should it redirect stdout (and stderr?) to /dev/log instead? (​Or ‘/var/run/syslog’ on OS/X?) Offer a command line option? Use the logging framework instead of print statements?