Gajim-devel November 2009

gajim-devel@gajim.org

4 participants
6 discussions

#2 Refactoring Newsletter
by Stephan Erb 28 Nov '09

28 Nov '09

Hey everybody, this mail has been sitting in my Drafts folder for much to long (since 16.11).... == What I have done "last" week == I have focused mostly on PEP and Contacts. * The Groupchat-related behavior of the Contact class was moved to a separate GC_Contacts class. (Similar to what was done to the Metacontact handling) * The classes Contacts and GC_Contacts are now 'account insensitive'. This means, they no longer handle contacts for different accounts internally, but instead we create an instance for each account. This removed quite some duplicated code. * Contacts now don't just know the account name, but instead have access to the newly created Account class. This means, we can now move account related methods and data to the Account class and can access them if we have a contact instance at hand. * By introducing several PEP related classes it was possible to cut the overall PEP code size in half (mostly due to polymorphism): 523 insertions(+), 956 deletions(-) * Adding a new PEP-Type can be done by creating a subclass and implementing three methods. The only remaining cumbersome job is due to RosterWindow which still requires that a new column is added. All the drawing/updating should work out of the box == What I encountered == Not much - just a lot of copy-pasted code. Removing the cruft was not that funny. Often I had the impression that I was unifying code that should have been unified from the beginning. == What I learned == connection_handlers.py and connection.py can be simplified by using more of what xmpppy gives to us: specific registration on stanzas per stanza, type, namespace. More tests would have been helpful. Doing my work was a shotgun surgery because there where changes in many different places with almost no real API. Having a local server and some highlevel roster tests saved me several times. Looks like I have to write a few more highlevel tests next time. == Additional, bigger refactorings since 16.11.09 == * Asterix was able do drop a lot of Zeroconf code by introducing a common Connection class and by dropping almost the complete ConnectionBytestream implementation used by Zeroconf. * Asterix re-refactored a few of my changes: Goal is that no caller has to know whether an account is a local or normal one. * Asterix dropped egg.trayicon. We now use the default GTK statusicon * I started an initial attempt to make dependencies of our connection classes more explicit: In this particular case ConnectionPEP now knows the objects on which it calls methods. Before, it just assumed that "I will be subclassed and that the subclass defines a few methods". First goal is to get rid of false-positives in pylint and other statical checkers. * Alexander updated all our docstrings to be consistent and conforming to widely known coding standards. * Thibaut started some cleanup work in the jingle area. (including coding standards) I hope I didn't miss anything. Best Regards, steve

1 0

some issues with current gajim hg (default branch)
by Halim Sahin 25 Nov '09

25 Nov '09

Hi, I have two Problems with current gajim from hg: 1. Doing an audio call works fine but it consumes much cpu power. I have an amd x2 with 2x 1.9 GHz. The load is 2 during an audiocall and top show about 80/95%. 2. a11y: I wasn't able to open the menu from a chatwindow to change the gpg encryption setting. In the past alt+a could be used but this doesn't work any more. Regards Halim

1 0

traceeback in gajim hg default branch
by Halim Sahin 19 Nov '09

19 Nov '09

Hi, Traceback (most recent call last): File "/usr/share/gajim/src/gui_interface.py", line 2922, in autoconnect message = self.roster.get_status_message(show, on_message) File "/usr/share/gajim/src/roster_window.py", line 2084, in get_status_message on_response('', empty_pep) File "/usr/share/gajim/src/gui_interface.py", line 2920, in on_message self.roster.send_pep(a, pep_dict) File "/usr/share/gajim/src/roster_window.py", line 1908, in send_pep connection.send_activity(activity, subactivity, activity_text) AttributeError: ConnectionZeroconf instance has no attribute 'send_activity' Regards Halim -- Halim Sahin E-Mail: halim.sahin (at) t-online.de

2 1

#1 Weekly Refactoring Newsletter
by Stephan Erb 11 Nov '09

11 Nov '09

Hey everybody, == What I have done last week(s): == I have worked on several smaller issues in many places of our code base. The highlights are: * Introduced a common super class for Contact and GC_Contact so that both can share code. * Contacts now know their associated account. There might still be old code doing heavy lifting to work around the missing account attribute. This old code can now be removed. * Made it more explicit how we use the Contacts class. Now there are several methods to distinguish the creation of certain kinds of contacts, e.g. self contacts (other resources of us) * Extracted the Metacontact related methods of Contacts to a separate class * Wrote tests for the caps module * Internally refactored the caps module to unify the handling of the old and new XEP version * Removed the 'nonexistant' workaround for caps features scattered over our codebase by using a feature blacklist to handle those special cases * The global caps cache now resides under caps.capscache instead of gajim.capscache. This removed the dependency of the caps.py on gajim.py * Moved the Interface() class from gajim.py to gui_interface.py * Added a pylint and a build-test script to scripts/dev * Reorganized our tests. Tests requiring network or GTK have been moved to a separate module. == What I encountered == Our startup logic is twisted and complex. It involves gajim, common/gajim and the constructor of the Interface class. It is not really obvious where we do certain initializations and pre-start actions (e.g. check that we are the default jabber client). The problem is tightened by modules which perform actions without explicitly stating them. Without looking into each module it is not obvious which dependencies have to be reflected in the import order. A first step to simplify this could be to make module import side-effects explicit: import i18n i18n.register_gettext_translation_builtin(localdir) instead of just import i18n == What I learned == It gets worse before it gets better. Sometime you just have to try something to be able to think about a problem or an area of code more clearly. Sometimes it is even advisable to throw the refactored code away and to start again from scratch. Small incremental steps help to prevent major breakage. == What I plan for the next days == My current plans are to take another closer look at how we handle and create contacts. Furthermore, vArDo and I will have a closer look at the plugin branch. Comments and ideas welcome :-) Best Regards, steve

2 1

Next meeting
by Yann Leboulanger 10 Nov '09

10 Nov '09

Hi all, Next meeting will be on monday 16/11 at 8:00 pm GMT (9:00 pm west Europe) We need to talk about current branches status (plugin, refactoring, archiving, whiteboard), and what to do in trunk. See you on monday! -- Yann

1 0

Gajim 0.13-rc1
by Yann Leboulanger 03 Nov '09

03 Nov '09

Hi all, Gajim 0.13 has its first release candidate! Don't hesitate to test and report all bugs you find. You can find changelog here: http://hg.gajim.org/gajim/file/cb35a23ac836/ChangeLog You can find tarball here: http://www.gajim.org/downloads/0.13/gajim-0.13-rc1.tar.gz And windows installer here: http://www.gajim.org/downloads/0.13/gajim-0.13-rc1.exe Thanks in advanced for your reports -- Gajim team

1 0

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

Gajim-devel November 2009