(No comment was entered for this change.)

Do to technical difficulties, I unclaim the sound-juicer task and claim this one.

(No comment was entered for this change.)

Sebp:
I'm running into a slight difficulty. I can use
ModuleList.get_module_instance_from_name() for enabled modules, but
DisabledModuleList has no correspounding function. I've attached the diff file so
that you can see what I'm trying to do.

browser.diff 4.8 KB   Download

I see. I have to think about it. I let you now when I came up with a solution.

Sebp,
The disabled modules don't have a get_filename method because they aren't truely
loaded like the other modules are.

Ignore my last comment because its wrong.

The problem is that the DisableModulesList class keeps track of the classes of
disabled modules but not the module's themselves. In order to reload the modules we
need the module's file name.

You can usually get the file name from the module's __file__ attribute, but we can't
do that because we only have access to the classes in the files. The classes have a
built it __class__ attribute, but its a string and *not* a reference to the actual
class, so that doesn't help us.

There are two possible solutions and both are somewhat hackish. Both of them work by
saving the modules's filename in DisabledModuleList.

The first solution would be to add on an extra parameter to the
module-not-initialized signal containing the module's file name. We would then save
the file name in a new column in DisabledModuleList.

The second way is a bit more hackish, but its *much* simpler. We simply add on a new
attribute to the class before we emit the signal. It would look something like this.

def import_module(self, filename):
...
   module.filename = filename
   self.emit("module-not-initialized", module)

I know that the explanation isn't so clear... I'll post code soon showing what I mean.

The code I've attached should do the job. Its a bit hackish, but look it over and
tell me what you think. 

One problem with it is that the preferences window still shows the old modules. I
think I know what the problem is, but I'm too tired to fix it now.

Btw, the Epiphany Web History module is missing a requirments method.

browsers.diff 5.7 KB   Download

Patch attached.

browsers.diff 7.3 KB   Download

Patch attached. It fixes several small bugs.

browsers.diff 8.1 KB   Download

I reviewed you patch. I encountered two issues:
* When the new browser is neither firefox nor epiphany you should disable the web modules
* Instead of module.stop() and module.set_enabled(False) you should use
CoreImpl.stop_module

Some of the functions in ModuleList seem to be broken. I'm getting the following
traceback when using CoreImpl.Stop:
Traceback (most recent call last):
  File "/home/natan/deskbar-trunk/deskbar/core/ModuleList.py", line 139, in
module_toggled_cb
    self[self.get_position_from_context(module)[0]][self.ENABLED_COL] =
module.is_enabled()
TypeError: could not parse subscript as a tree path

Patch attached. I've fixed everything up. I'm not using CoreImpl.stop_module()
because it seems to be broken.

browsers.diff 8.5 KB   Download

You still have the problem that the new list of enabled modules isn't written to
GConf. I think best is to move DeskbarPreferences.update_gconf to CoreImpl class and
use it their to update the list. You might not need to disable the modules manually
then, too. In addition, I think you should add to
CoreImpl._on_enabled_modules_changed that newly added modules will be loaded.
Obviously, there's unfinished work in this method.

I tried to find the error when using CoreImpl.stop_module. Nevertheless, you should
use both CoreImpl.initialize_module and CoreImpl.stop_module. Duplicate code is a bad
thing.

To sum up it should work when you change CoreImpl._on_enabled_modules_changed that it
initializes new modules and if you only call CoreImpl.stop_module and
CoreImpl.initialize_module in _on_default_browser_changed method. Have a look at
DeskbarPreferences, too, how it handles enabling and disabling modules.

I'm sorry these issues didn't came to mind before. If you think it's too much work
now, you can resign from this task.

I'm definitly up to the task, and I don't mind the extra work. Besides, there aren't
very many open tasks at the moment, so its definitly better than working on nothing.

I'll post later if I have any questions.

I found the problem with CoreImpl.stop_module. The modules were being stopped
asyncrously and the code in _on_default_browser_changed assumed that they would be
stopped immediately. I'm now using async=False, and I don't think that the time delay
is significant enough to be of concern. The function should be called very rarely,
and when it is called a maximum of three modules are stopped.

I moved DeskbarPreferences.update_gconf to CoreImpl, but we still need disable the
modules manually. update_gconf only toggles a module's enabled status, but we also
need to move the disabled modules into DisabledModuleList which is for modules with
errors.

I made the other changes that you suggested and the patch is attached. It should be
in finished form, but I'll make any other changes that you think are necessary.

browsers.diff 11.5 KB   Download

It works like a charm now. Thanks a lot for your awesome work.

I have another small issue, it's not important, therefore you don't have to fix it,
you already completed the task. When you change the default browser when Deskbar is
not running it will not change the web modules. I don't know if it's possible that it
does, when starting Deskbar again. As I mentioned before, I think most users won't
come across this issue.

I'm going to commit your patch tomorrow. Thanks a lot for all your work.

No problem. I wouldn't mind taking another task for deskbar, if you write another one up.

I think the simplest way to fix that issue would be to add on another method to the
api called "alternatives" or something like that. We would call the method whenever a
module's requirments method returns false.

I'm not sure whether that's very practical or not. I can't think of that many other
cases where it would be useful.

Anyway, can someone please mark this task as completed?

(No comment was entered for this change.)

Hello Sebp,
I'm leaving this comment here because I'm not sure how else I should contact you. ;)

I'm doing some work on creating a universal applet format, and I think it would
pretty neat if we could get some kind of integration with the deskbar. There's an
overview of the reasons for this here: http://theesylum.com/2008/02/01/desktop-20/

Tell me what you think.