{5} Assigned, Active Tickets by Owner (Full Description) (13 matches)

CherryPy unconditionally sets the log level for loggers used by cherrypy to DEBUG for error logs and INFO for access logs. This may override log levels configured by the user since this is done during startup.

From IRC:
[04:40am] askadar: I just want to log everything to sys.stdout, but not the requests since the access log is being written by lighttpd anyway
....
[04:40am] askadar: so no matter what I do the root logger, access records are still being written to stdout
....
[04:43am] Lawouach_: setting the log.screen : False doesn't work as expected?
[04:44am] askadar: I want to disable the default logging
[04:44am] askadar: and have only the root logger write to stdout
[04:44am] askadar: log.screen: False turns of all logging
[04:44am] askadar: by introducing the root logger with level WARNING, cherrypy's internal stuff suddenly starts to log again,too
[04:44am] askadar: and that is very unexpected
[04:45am] Lawouach_: I see
[04:45am] askadar: I had hoped to filter all access records by setting the root logger level to WARNING, since access stuff is level INFO
[04:45am] Lawouach_: If you think there is a design issue here you ought to open a ticket with an alternative (patch are even better but not mandatory

There is no reason why CherryPy should set these levels, especially not in production environments. The solution is to just leave the setting alone. Patch is attached.

When socket_port is specified as 0, an exception is thrown as follows:

File
"/localhome/raghu/localwork/cherrypy/svn.cherrypy.org/trunk/cherrypy/_cpserver.py",
line 69, in quickstart
    self.start()
  File
"/localhome/raghu/localwork/cherrypy/svn.cherrypy.org/trunk/cherrypy/_cpserver.py",
line 95, in start
    self._start_http(httpserver)
  File
"/localhome/raghu/localwork/cherrypy/svn.cherrypy.org/trunk/cherrypy/_cpserver.py",
line 118, in _start_http
    self.wait(httpserver)
  File
"/localhome/raghu/localwork/cherrypy/svn.cherrypy.org/trunk/cherrypy/_cpserver.py",
line 159, in wait
    wait_for_occupied_port(*bind_addr)
  File
"/localhome/raghu/localwork/cherrypy/svn.cherrypy.org/trunk/cherrypy/_cpserver.py",
line 247, in wait_for_occupied_port
    raise IOError(msg)
IOError: Port 0 not bound on 'localhost'

I am using the latest code from the trunk. It appears to me that even though wsgiserver is properly passing 0 to socket.bind (and hence binding to an unused port), other parts of the code still use "configured" port which is 0.

The session library in CP isn't very good: it's hard to extend, partly because IMO it tries to do too much. For example, it shouldn't try to allow both eager and lazy locking. It also does too much at the class level instead of the instance level; that is, instead of instantiating a Session object per request, it should just stick all that state into cherrypy.request--it would then be free to instantiate a FileSession? object per Application, rather than using class variables to manage config across requests.

Using CherryPy 3.0.3.

I have a method setup as such:

@cherrypy.expose
@cherrypy.tools.response_headers(headers = \
           [('Content-Type','application/data')])
def my_method:
        ...
        yield "blah"
        ...

my_method._cp_config = { 'response.stream': True }

Whenever the method is executed by cherrypy's default dispatcher, the Content-Type is text/html, not application/data as I expected.

However, if I changed the method setup to be like this:

@cherrypy.expose
def my_method:
        ...
        yield "blah"
        ...

my_method._cp_config = {
        'response.stream': True,
        'tools.response_headers.on': True,
        'tools.response_headers.headers': [('Content-Type', 'application/data')]
}

...then it works.

Ideas?

The Autoreloader class has the entire logic for reloading in its run() method. The patch moves the match check into a separate method. With the patch subclasses of Autoreloader can create a custom check method based on the name and module.

Index: cherrypy/process/plugins.py
===================================================================
--- cherrypy/process/plugins.py (revision 9407)
+++ cherrypy/process/plugins.py (working copy)
@@ -432,11 +432,15 @@
         Monitor.start(self)
     start.priority = 70

+    def match_hook(self, name, module):
+        """Check if module should be considered for reloading."""
+        return re.match(self.match, name)
+
     def run(self):
         """Reload the process if registered files have been modified."""
         sysfiles = set()
         for k, m in sys.modules.items():
-            if re.match(self.match, k):
+            if self.match_hook(k, m):
                 if hasattr(m, '__loader__'):
                     if hasattr(m.__loader__, 'archive'):
                         k = m.__loader__.archive

In wsgiserver.tick, some socket.error exceptions are caught and ignored ("Bad file descriptor", "Socket operation on non-socket", "Resource temporarily unavailable").

When using in the same application gtk and cherrypy, displaying a gtk dialog (such as a fileselector) while cause the server to crash with the "Interrupted system call" exception.

Catching this exception solves the issue :

--- /tmp/__init__.py    2008-09-24 18:13:55.000000000 +0200
+++ /var/lib/python-support/python2.5/cherrypy/wsgiserver/__init__.py   2008-09-24 17:40:30.000000000 +0200
@@ -913,6 +913,8 @@
             if msg == "Resource temporarily unavailable":
                 # Just try again. See http://www.cherrypy.org/ticket/479.
                 return
+            if msg == "Interrupted system call":
+                return
             raise
     
     def _get_interrupt(self):

But considering also #860 (use error code rather than error message), the except clause should rather catch the error code 4.