Re: [Denemo-devel] Trying to build from git

[Richard Shann]

On Fri, 2008-05-30 at 20:30 +0200, Jean-René Reinhard wrote:
> I hope to finish the core of the keymap migration tomorrow. I'll send Richard 
> a
> notice so that he can pull from my public repository.
> 
> Some news about it :
> I have rewritten the keymap type to :
> - reuse GtkActionEntry objects do describe commands
> - use hashtables to get an action from it's name
> - base its management of keybings on strings generater by gtk_accelerator_name
> (one level hashtable instead of two level)
> 
> I have to finish the accel management stuff, which should be easy now that I 
> have figured how to manage the quick edit stuff, and fix the keybinding 
> dialog. 
> 
> This modification enables to simplify the code base, since we can remove
> unmenued commands table and the global table denemo_commands, which duplicated
> information in the keymap for no reason at all (If I'm right) and to have a
> unified way of handling keybinding (only one configuration file)
This all sounds great.
> 
> some polishing still needs to be done (add strings in .po files, ...)
> 
Ah, do you understand this stuff? Should we be surrounding translatable
strings with _() or _N() ? I've seen both in the code but haven't had
time to look into it.
> During implementation, I discovered that there is still a big architectural 
> bug
> in the way keybindings are managed. The issue is that input of some commands 
> in
> denemo are modal, depends on the value of two modes (type (classic, edit, 
> insert) and mode(note, rest, blank)). For the time being these modes are 
> managed at two levels : 
> - specific actions are defined for actions, whatever the mode their are meant 
> to
>   be used in. This is an issue since an accel key can only belong to one 
> action.
>   So it requires to define accel keys for a insertion/edit/changing main
>   duration, which is awkward, all the more than the same callback is used for
>   these different actions, or changing accel keys for modal commands, which
>   is unnecessarily complicated
> - The callbacks does different things depending on the mode.
You're definitely right, there is a tangle here that needs simplifying -
many non-modal commands are implemented by switching mode, executing the
modal command and switching back.

> 
> Once I'm finished with the keymap, if it is ok for you, I plan to remove the 
> unnecessary actions and check that the callbacks are fine.
> 
Well, it should be easier to see what needs to be done once the keymap
complications have gone.
> Finally as a developper, it is most inconvenient to have versionned Makefiles.
> They appear as modified, since I tweaked my .configure. Does anybody now If I
> can make git ignore modifications on these files? gitignore stuff seems to 
> work
> only for untracked files...
What do other projects do? Its all to do with making it easy for people
to get the latest Denemo, right? The thing I see most often is you get
the stuff and do ./configure, make and make install, someone has already
done the ./autogen.sh for you. I suspect I am working in a very clunky
way - developing in one tree and merging into git and then commit and
push; the result is that I am making mistakes quite often. Another place
where things are clunky is when you add a file to denemo/examples, you
have to add to Makefile.am and Makefile.in
Is there a rule you could put in Makefile.am to take all of the files in
examples without having to list them?
It's great to have your help - Denemo may finally get itself into shape!

Richard