Attachment '20081216.txt'
Download 1 * mclasen apologizes
2 * mclasen wonders if the meeting takes place somewhere else
3 * mclasen would like to discuss "2.16 for gnome 2.26"
4 <tristan> erm, are we missing people ?
5 <tristan> mclasen, if you release this early 2.16, will it include buildable menus ?
6 <mclasen> we are a little short on people who have something to say, it seems...
7 <mclasen> It depends on what people want and what is good to be shipped before the end of the year, I think
8 <mclasen> will buildable menus allow you to have gtkbuilder support in glade ?
9 <mclasen> that would be a pretty convincing argument for me
10 <tristan> well *I* want to release a fully featured glade stable for gnome 2.26
11 <tristan> yes
12 <tristan> it will allow conversions of any glade files, that work
13 <mclasen> sounds like a nobrainer to me
14 <tristan> and if I get the action property in too, that will just be awesome
15 <mclasen> not so sure about that one
16 <tristan> :)
17 <mclasen> well, the action property is certainly easier than a fullblown Activatable interface
18 <tristan> hmmm, last meeting we discussed the patch, I got the impression it was all around agreed that the iface was a better approach
19 <tristan> so I hacked all night and cleaned out gtkaction.c
20 <mitch> +1
21 <tristan> mitch, did you see the patch finally ?
22 <mclasen> mitch ?
23 <mitch> i'm all for the interface approach too, but i was too busy and sick today to finally look at the patch
24 <tristan> I could try to sum up what it does in a few sentences
25 <tristan> It makes a property "related-action" on the iface
26 <tristan> (and a "use-action-appearance" too)
27 <tristan> when gtk_activatable_set_related_action () is called
28 <mitch> why "related"?
29 <tristan> the property changes and the iface implementor must hook onto the action
30 <tristan> mitch, filechooseractions
31 <tristan> thats why
32 <mclasen> mitch: we already have an "action" property
33 <mitch> oh ok
34 <mclasen> and gobject doesn't handle name clashes very gracefully
35 <mclasen> or at all...
36 <tristan> filechooser *uses* the GtkAction property :)
37 <tristan> anyway, the activatable hooks onto notify
38 <tristan> and calls gtk_activatable_update() from there
39 <tristan> I added a vfunc so that child implementations could be graceful
40 <tristan> one notify, one override in the implementing class
41 <mitch> so the activatable approach makes the widgets true models instead of being configured from the outside?
42 <mclasen> I'm still worried about the scalability of this many notfy handers...have you done a before-after comparison with the gimp ?
43 <tristan> umm, one last thing... also the iface implementors call gtk_action_activate()
44 <mitch> perhaps that method should be "update_action"?
45 <tristan> where gtkaction would have registered a signal
46 <tristan> activatable->update seems pretty implied its an action if you ask me :-/
47 <tristan> but
48 <tristan> if you wish
49 <mitch> i don't see how "activatable" implies "action" at all actually ;)
50 <mitch> a button is activatable
51 <tristan> its the activatable that updates
52 <tristan> not the action
53 <mitch> oh
54 <mitch> err yes sure
55 <tristan> the activatable gets notified and calls itself
56 <tristan> for subclasses to implement it cleanly
57 <tristan> for instance, togglebutton is interested in the toggle actions active state
58 <mitch> i wonder if it wouldn't make more sense to have some "set_action" methos and have the proxies connect themselves?
59 <tristan> but not button
60 <mitch> or am i on crak?
61 <mitch> +c
62 <tristan> the proxies do connect themselves with my patch
63 <tristan> and connect_proxy() is deprecated in favor of gtk_activatable_set_related_action()
64 <tristan> but as mclasen said, I didnt test this with gimp
65 <mitch> hmm
66 <tristan> and see if I could find performance differences
67 <tristan> but it seems to me to be the right thing to do
68 <mitch> so again, a GtkButton would not be an activatable?
69 <tristan> remember that action property changes are not a regular occurance, they happen sometimes
70 <tristan> yes it would be activatable
71 <mitch> maybe i should just sleep as planned and ask that stuff on daytime ;)
72 <mclasen> mitch: you can connect buttons to actions currently
73 <tristan> and by virtue of it being activatable, so would togglebutton
74 <mitch> yes you can, that's why i got confused
75 <mclasen> so if the new approach allows only to connect activatables, they better be
76 <tristan> I was interested to find out that expanders can be seen as a toggle state
77 <mitch> so how would e.g. a GtkToggleable interface fit in there?
78 <tristan> and they have an activate signal that toggles it
79 <tristan> mitch, my patch doesnt do that, but ... I have some ideas
80 <mitch> would it have GtkActivatable as prerequisite and somehow hook into set_related_action() ?
81 <tristan> one is type safety
82 <tristan> it could override the type of the action property
83 <mitch> yes for example
84 <tristan> and serve as an all purpose toggle
85 <tristan> and be resposable for the toggled signal
86 <tristan> etc etc
87 <tristan> but
88 <tristan> I would rather see that as out of scope for this patch
89 <tristan> and keep it simple
90 <mitch> but this patch should not block that other thing
91 <tristan> I think the toggles should be activatable by prerequisite though
92 <mitch> also not a GtkRadio interface
93 <tristan> exactly it should not
94 <mitch> yes i agree
95 <tristan> it was intended to be simple and scalable
96 <tristan> one thing is a little weird
97 <tristan> with recent chooser actions
98 <tristan> they call functions on their proxies
99 <mitch> so what i suggest it *trying* how a Gtktoggleable interface works on top of that, but leave it out of the current patch
100 <tristan> and set sort func etc
101 <mclasen> so, how do we proceed on this ? mitch reviews the patch and tristan does scalability tests with the Gimp ?
102 <tristan> thats the *only* reason gtkaction still maintains a proxy list
103 <mitch> i think that's best
104 <tristan> and needs a connect_proxy
105 <mitch> mclasen: i can't promise anything before christmas though, super busy mode
106 * mclasen too
107 <mclasen> tristan: oh, that proxy list was one of my scalability concerns, too
108 <tristan> mclasen, its still maintained, and was kept for compatability reasons
109 <tristan> people want to list the proxies etc
110 <mclasen> so, can I get some general feedback on the "2.16 in 2.26" idea ?
111 <tristan> I dont see how it can be removed, and its useful for weird ones, like recentchoosers
112 <mclasen> yeah, probably
113 <mclasen> is any of the stuff thats currently in trunk not baked enough to go out in a 2.15 release in January ?
114 <tristan> ummm, you cant release those margined labels ;-)
115 <tristan> heh
116 <mclasen> what labels ?
117 <tristan> sorry I mean menus
118 <tristan> left margin in menus, notably from combos
119 <mclasen> are you just talking about comboboxes, or do you disagree with margins in general ?
120 <tristan> no you need margins for you menu items that have icons dont you ?
121 <tristan> and check items and stuff
122 <tristan> but not when none of the menu items have anything
123 <tristan> right ?
124 <mclasen> the idea of the margins change was to always have a consistent margin, regardless if icons are presnt or not
125 <tristan> sorry, was that an intended feature ?
126 <tristan> hrm
127 <mclasen> the margin you showed in your combobox screenshot looked excessive though
128 <tristan> well, in combos surely excessive
129 <tristan> yeah
130 <tristan> I dont know if in menus it should be default behaviour
131 <tristan> well, if you expose a property there, I can easily let glade configure it... as usual...
132 <tristan> but yeah actually, the combos are chopping out some text with that margin
133 <tristan> my combo resizes to fit the menu width, and then the marin pushes my text off the menuitem
134 <mclasen> yeah
135 <mclasen> thats just a bug
136 <mclasen> mitch: did you do the progress entry work ?
137 <mitch> yes there is also stuff left to do :(
138 <mclasen> right
139 <tristan> oh yeah, that one thing about actions... is that people like actions with menus... and well... uimanager... *cough*
140 <tristan> anyway
141 <mclasen> the selection vs progress color clash may be hard to fix without introducing an extra color for the progress
142 <mitch> tristan: you will find a complex example in gimp for uimanager and menus ;)
143 <tristan> mitch, man, maybe we should talk about uimanagers in another meeting
144 <mclasen> yeah, any major action rework needs to survive the gimp test
145 <mclasen> mitch: what about the unfinished flipping work
146 <mclasen> can that go out in a 2.15 release as is ?
147 <tristan> hahaha, I want to see the gimp work with actions and menus and without uimanagers !
148 <tristan> grrrr
149 <mitch> mclasen: i can commit the outstanding patch
150 <tristan> ok I'll run gimp
151 <mclasen> mitch: but keeping everything abstract ?
152 <mclasen> I guess thats fine
153 <mitch> mclasen: eek, indeed
154 <mclasen> it just makes it less useful than it could be
155 <mitch> needs more work indeed
156 <mitch> will tackel that after xmas
157 <mitch> tackle
158 <mclasen> ok
159 <mclasen> so, I assume from this discussion that there is no major disagreement over doing an early 2.16
160 <mitch> need sleep now
161 * mitch <- sick
162 <mitch> nite all
163 <mclasen> gute besserung
164 * mitch is now known as miZzz
165 <miZzz> danke :)
166 <mclasen> lets close the meeting then, I guess
167 <mclasen> last meeting of 2008, probably
168 <mclasen> see you all in 2009
169 <miZzz> happy new year ;)
170 <miZzz> Zzzz...
171 <mclasen> I'll send some mail about 2.16 plan changes to gtk-devel
172 <Lethalman> thanks to everyone for considering the bug :)
Attached Files
To refer to attachments on a page, use attachment:filename, as shown below in the list of files. Do NOT use the URL of the [get] link, since this is subject to change and can break easily.You are not allowed to attach a file to this page.