Mailman 3 January 2003 - Twisted

[Twisted-Python] Gtk version
by David LeBlanc Jan. 26, 2003

Jan. 26, 2003

Which version of gtk and pygtk should I be using to get twisted GUI's to work on Windows 2000? In this particular case, it's manhole. PyGtk-0.6.11 complains about GDK on every keystroke: This is after the GUI comes up: setting login report 'connecting...' -:1: error: invalid string constant "Console_stdout", expected valid string constant It does let me log in using the user and password entered when I made the tap. Every keystroke causes this: Traceback (most recent call last): File "J:\Python22\Lib\site-packages\pygtk\gtk.py", line 127, in __call__ ret = apply(self.func, a) File "J:\Python22\lib\site-packages\twisted\manhole\ui\pywidgets.py", line 63, in processKey isMeta = event.state & gtk.GDK.MOD4_MASK AttributeError: 'module' object has no attribute 'GDK' PyGtk-1.99.11-libglade dies with: J:\twila>manhole Traceback (most recent call last): File "J:\python22\scripts\manhole.py", line 33, in ? from twisted.scripts import manhole File "J:\Python22\Lib\site-packages\twisted\scripts\manhole.py", line 23, in ? from twisted.spread.ui import gtkutil File "J:\Python22\lib\site-packages\twisted\spread\ui\gtkutil.py", line 62, in ? class GetString(gtk.GtkWindow): AttributeError: 'module' object has no attribute 'GtkWindow' PyGtk-0.6.11 seems to be using gtk+-1.3.0 PyGtk-1.99.11-libglade uses gtk2_libraries-win32.zip Thanks, David LeBlanc Seattle, WA USA

3 3

[Twisted-Python] deferreds, and .rpy resources
by L. Daniel Burr Jan. 24, 2003

Jan. 24, 2003

On Fri, 24 Jan 2003 00:39:36 -0800, Dave Peticolas wrote: >On Thu, 2003-01-23 at 17:25, L. Daniel Burr wrote: >>Of course, you could also use twisted.enterprise.adbapi, >>but I've experienced a lot of memory leaks occurring when >>going that route, so I just use deferToThread instead. > >I'm about to start using t.e.adbapi pretty heavily, so I'm >interested in this report. Do you have more details? > >thanks, >dave To be clear, I definitely *haven't* been using adbapi heavily up until this point. The problem I experienced occurred around the time of Twisted 1.0.1 final, and deadlines prevented my doing any exhaustive testing. The scenario was pretty basic: Connect to an Oracle database, perform a query, return the result. Using adbapi from Twisted 1.0.1 and the cx_Oracle DBAPI module, the twistd process showed steady growth in size. Each time the code executed, the process would grow by 10-50k, never releasing any memory. The same scenario, using DCOracle2, showed similar leakage. Having insufficient time to debug the thing, I went with the process of elimination, and started using deferToThread instead of adbapi, and the leak was gone. That does not in any way prove that adbapi is the root cause, only that it may be involved in some way. This was all on Win32, by the way, so I have no idea if it is an issue on Linux or Solaris. In any case, I'll get a chance to revisit the issue shortly, and I'll report my findings. L. Daniel Burr

1 0

[Twisted-Python] deferreds, and .rpy resources
by L. Daniel Burr Jan. 24, 2003

Jan. 24, 2003

On Fri, 24 Jan 2003 18:36:37 +0100, Mario Ruggier wrote: >I suggest that one or both of these (corrected) examples be added >to the list of examples bundled with twisted, as i would expect >that many prospective users will not find this very obvious. If i >knew what i was doing i would offer to do it myself... > >All the best, mario You are right; Twisted has some good example code, but it needs more examples that "do stuff", and especially examples that show how to combine parts of Twisted (web and enterprise, for example) to accomplish common internet development tasks. I'll put my money where my mouth is on this one, and clean up the example I posted (resource + deferToThread). I'll also post an example using twisted.enterprise (resource + adbapi). By the phrase "cleaned up", I mean compliant with the documentation guidelines, development guidelines, epydoc, etc. If the results don't suck too badly, maybe somebody who understands CVS can throw them in the examples directory. L. Daniel Burr

1 0

[Twisted-Python] deferreds, and .rpy resources
by L. Daniel Burr Jan. 24, 2003

Jan. 24, 2003

On Thu, 23 Jan 2003, Mario Ruggier wrote: >Hi, sorry for the newbie questions but I do not understand how I >can combine the resource interface with deferred callbacks. Or, >specifically, can anyone point me to a simple example of a .rpy >that does a simple database call, and returns the result to the >http response? > >Many thanks, > >Mario I'm no expert, but you can do something like this (UNTESTED): ----------------------------------------------------------------- # replace this import with whatever DBAPI module you use... import cx_Oracle # twisted imports from twisted.internet import threads from twisted.web import resource from twisted.web.server import NOT_DONE_YET def databaseOperation(): con = cx_Oracle.connect("foo/bar(a)mydb.mysvr.org") cur = con.cursor() cur.execute("SELECT COUNT(*) FROM MyTable") databaseResult = cur.fetchall() cur.close() con.close() return databaseResult[0][0] def databaseResult(result, httpRequest): httpRequest.write(result) httpRequest.finish() class MyResource(resource.Resource): def render(self, request): d = threads.deferToThread(databaseOperation) d.addCallback(databaseResult, request) return NOT_DONE_YET resource = MyResource() ----------------------------------------------------------------- Of course, you could also use twisted.enterprise.adbapi, but I've experienced a lot of memory leaks occurring when going that route, so I just use deferToThread instead. There is probably a better way to code this, but I'm not far enough along in my Python/Twisted experience to provide it. L. Daniel Burr

4 4

RE: [Twisted-Python] First contribution...
by Sylane Jan. 23, 2003

Jan. 23, 2003

Hi, More than a week ago, I posted a little patch for pb saying I working on a pb example... Since this post, nobody ever reply it. Even to say the patch sucks or something and it's not applied to the CVS at this time. The example I worked on DONT work without this patch cause the cacheable implementation is broken... If nobody care of my contributions, OK say it and i will start a fork of spread for my own use. If I did bad, if it's not the way contributions must be done, OK say it too. If nobody received my post, you will not received this one too... and I don't understand anymore. P.S: Pleeeassse someone reply...... P.S.2: I'm not a paranoiac. And it's not because I'm a paranoiac that you're not all against me ! ! ! -- Best regards, Sylane mailto:twisted@sylane.com

3 2

[Twisted-Python] Error handling in Application.bindPorts
by Bernhard Herzog Jan. 23, 2003

Jan. 23, 2003

Hi, I'm trying to get one of my twisted programs to terminate immediately if it can't listen on a port on startup. It's main program is basically the same as in e.g. doc/howto/listings/application/app1.py, i.e (quoting from app1.py): app = twisted.internet.app.Application("daytimer") f = Factory() f.protocol = Daytime app.listenTCP(8813, f) app.run() My program uses a different protocol and factory, but the principle is the same: it calls app.listenTCP before it calls app.run(). Now, if the mainloop isn't running yet, listenTCP doesn't start listening immediately[1] but simply appends the arguments to a list. Then, when app.run() is called and that list is not empty it will call bindPorts before entering the mainloop. bindPorts simply catches the CannotListenError, logs a message and returns immediately even if not all ports have been bound yet. Unfortunately, even if bindPorts returns early, app.run still enters the mainloop. There doesn't seem to be a good way to avoid this. IMO bindPorts should reraise the exception after writing logging the error instead of simply returning. Bernhard [1] I think app.listenTCP doesn't try to listen immediately because this method is also used when creating taps and that should be possible even if the port is currently in use or the user doesn't have permission to listen on that port. -- Intevation GmbH http://intevation.de/ Sketch http://sketch.sourceforge.net/ MapIt! http://www.mapit.de/

1 0

[Twisted-Python] deferreds, and .rpy resources
by Mario Ruggier Jan. 23, 2003

Jan. 23, 2003

Hi, sorry for the newbie questions but I do not understand how I can combine the resource interface with deferred callbacks. Or, specifically, can anyone point me to a simple example of a .rpy that does a simple database call, and returns the result to the http response? Many thanks, Mario

1 0

[Twisted-Python] Test failures
by Jp Calderone Jan. 22, 2003

Jan. 22, 2003

Been playing around with different test orders (courtesy trials --random parameter) and I've found that with a seed of 4: =============================================================================== FAILURE: testWakeUp (twisted.test.test_internet.InterfaceTestCase) ------------------------------------------------------------------------------- Traceback (most recent call last): File "/home/exarkun/projects/python/Twisted/twisted/trial/unittest.py", line 122, in runOneTest method(testCase) File "/home/exarkun/projects/python/Twisted/twisted/test/test_internet.py", line 266, in testWakeUp self.assertApproximates(time.time(), start + 0.5, 0.5) File "/home/exarkun/projects/python/Twisted/twisted/trial/unittest.py", line 62, in assertApproximates raise AssertionError, (msg or "%s ~== %s" % (first, second)) AssertionError: 1043207178.43 ~== 1043207173.94 =============================================================================== ERROR: testFailedRETR (twisted.test.test_ftp.FTPClientTests) ------------------------------------------------------------------------------- Traceback! exceptions.AttributeError, TCPClient instance has no attribute 'connector' /home/exarkun/projects/python/Twisted/twisted/internet/default.py:566:doSelect /home/exarkun/projects/python/Twisted/twisted/internet/tcp.py:250:doConnect =============================================================================== ERROR: testFailedRETR (twisted.test.test_ftp.FTPClientTests) ------------------------------------------------------------------------------- Traceback! exceptions.AttributeError, 'NoneType' object has no attribute 'connectionLost' /home/exarkun/projects/python/Twisted/twisted/internet/default.py:579:doSelect /home/exarkun/projects/python/Twisted/twisted/internet/tcp.py:258:connectionLost /home/exarkun/projects/python/Twisted/twisted/internet/tcp.py:140:connectionLost ------------------------------------------------------------------------------- consistently comes up. Jp -- "There is no reason for any individual to have a computer in their home." -- Ken Olson, President of DEC, World Future Society Convention, 1977 -- 12:00am up 36 days, 9:48, 4 users, load average: 0.10, 0.14, 0.14

2 1

[Twisted-Python] lore bug
by Christopher Armstrong Jan. 21, 2003

Jan. 21, 2003

< and > in source listings that are inserted directly in the source .html don't work. See http://twistedmatrix.com/documents/howto/widgets.html and doc/howto/widgets.html for an example of this. If nobody gets around to this by tomorrow I'll try to dive in. -- Twisted | Christopher Armstrong: International Man of Twistery Radix | Release Manager, Twisted Project ---------+ http://twistedmatrix.com/users/radix.twistd/

2 2

[Twisted-Python] socket errors
by Jp Calderone Jan. 20, 2003

Jan. 20, 2003

Creating a socket can raise an exception in a few circumstances. Bad arguments for one (we don't have to worry about this, since we control the arguments in all circumstances); too many open files for another. Attached patch addresses the second circumstance, by catching the raised socket.error and wrapping it up in a CannotListenError. It seems unlikely that anyone has code to handle this error condition (and hence is relying on socket.error being raised), and CannotListenError seems to be a better thing to be raising in this case. Anyone disagree? Also, the XYZClient classes might be changed to catch the same socket.error and raise a ConnectError instead (attached doesn't address this). Jp -- Elitism: It's lonely at the top, but it's comforting to look down upon everyone at the bottom. -- 12:00am up 35 days, 9:48, 3 users, load average: 0.61, 0.33, 0.22

2 1