1 |
On Sat, 2006-10-28 at 17:12 -0400, Alec Warner wrote: |
2 |
> Caleb Cushing wrote: |
3 |
> >> |
4 |
> >> cairo and openexr, good idea, AFAIK. udev has come up before but from |
5 |
> >> that discussion, the flag means slightly different things in some cases. |
6 |
> >> Keeping it local allows individual per-pkg descriptions, so where it |
7 |
> >> means |
8 |
> >> something different, the description can say so. In both meanings, udev |
9 |
> >> defaulting to on remained best, but with the slightly different |
10 |
> >> meanings... |
11 |
> >> There was some discussion about modifying things or changing the flag |
12 |
> >> where it meant something else, but I don't know what came of that. |
13 |
> >> |
14 |
> > |
15 |
> > maybe it would be a lot of work. to even develop the tools. but it |
16 |
> > would be nice if a global use flag could have a detailed option. |
17 |
> > |
18 |
> > example. |
19 |
> > |
20 |
> > euse -i mplayer |
21 |
> > [+ C ] mplayer - Enable mplayer support for playback or encoding |
22 |
> > |
23 |
> > is what we currently get. |
24 |
> > |
25 |
> > add a -d option for --descriptive |
26 |
> > euse -id mplayer could show something like |
27 |
> > [+ C ] mplayer - Enable mplayer support for playback or encoding |
28 |
> > media-video/kmplayer - adds the ability to play back media using |
29 |
> > the mplayer engine |
30 |
> > |
31 |
> > or maybe something better... |
32 |
> |
33 |
> I don't think any specification precludes having a more descriptive |
34 |
> per-package meaning. It would just be a matter of: |
35 |
> |
36 |
> a. having devs write them in use.local.desc when necessary |
37 |
> b. Having tools look in use.local.desc first. |
38 |
> |
39 |
> But the whole point of global flags is really to consolidate the |
40 |
> description functions and keep naming consistent. So I doubt (a) will |
41 |
> ever come to pass for the majority of flags. Luckily (a) isn't a hard |
42 |
> requirement, tools don't loose functionality by looking in |
43 |
> use.local.desc first. |
44 |
|
45 |
The main problem in my eyes here is that with certain USE flags, the |
46 |
description doesn't really convey what the user will get. |
47 |
Many are in the form of "adds support for this optional thing" instead |
48 |
of "by using this optional dependency [named often the same as the USE |
49 |
flag name] you will get this and that extra functionality". |
50 |
|
51 |
|
52 |
Now what good wxGTK maintainer would I be, if I didn't pick wxwindows |
53 |
USE flag as an example: |
54 |
|
55 |
"wxwindows - Adds support for wxWindows/wxGTK GUI toolkit" |
56 |
|
57 |
Lets see what the user really gets with this in the example of a few |
58 |
packages that use this USE flag in IUSE: |
59 |
|
60 |
app-backup/bacula: a wxWidgets console, while there are other consoles |
61 |
available, such as gnome2-console and having both USE flags will result |
62 |
in two consoles. However all of these are dependent on bacule-console |
63 |
local USE flag... |
64 |
|
65 |
app-emulation-bochs: Compile a wxWidgets based GUI (other are available) |
66 |
media-gfx/zphoto: Use wxWidgets for GUI (to get ANY kind of GUI) |
67 |
media-video/gpac: Build wxOsmo4 and V4Studio |
68 |
media-video/mkvtoolnix: Build mmg and a GUI for mkvinfo |
69 |
|
70 |
All of these "support wxGTK" in the sense that they pull it in and build |
71 |
a few things against it, but it doesn't articulate what does the user |
72 |
exactly get from using that global USE flag for this particular package. |
73 |
Sometimes she gets just some little extra GUI apps, sometimes a GUI in |
74 |
the first place, sometimes an extra GUI interface in addition to others, |
75 |
and so on. |
76 |
Similar things can be observed with many other global USE flags (and |
77 |
also some local flags) - examples on request (time consuming detail |
78 |
gathering). |
79 |
|
80 |
And that's the problem - the user doesn't know what benefit will it |
81 |
bring her to use or not use a global USE flag for this particular |
82 |
package. |
83 |
|
84 |
-- |
85 |
Mart Raudsepp |
86 |
Gentoo Developer |
87 |
Mail: leio@g.o |
88 |
Weblog: http://planet.gentoo.org/developers/leio |