1 |
On Sun, Oct 24, 2004 at 06:10:12PM +0100, Ciaran McCreesh wrote: |
2 |
> On Sun, 24 Oct 2004 18:31:19 +0200 Maurice van der Pot |
3 |
> | Whether a use flag is specifically turned off for a group or it is |
4 |
> | just not present in the group should make no difference for the |
5 |
> | meaning of GROUP. The same holds for -@GROUP. In other words, -use and |
6 |
> | -@GROUP should only be used in the definition of a group to filter |
7 |
> | flags out. |
8 |
> | |
9 |
> | Assume MYGNOME="@GNOME -gtk", then USE="... -@MYGNOME ..." should |
10 |
> | *not* be equivalent to USE="... -@GNOME gtk ...", instead it should |
11 |
> | not influence the state of gtk at all. |
12 |
> |
13 |
> Interesting idea. This then brings us back to one of the original |
14 |
> objections, which was that there should be some way to turn off flags |
15 |
> from within a group. Only way to do that would be to introduce yet |
16 |
> another modifier, which is inconsistent with the rest of portage and |
17 |
> will confuse the heck out of anyone who hasn't studied set theory... |
18 |
|
19 |
I just now read the comments to the previous edition. I should've done |
20 |
that sooner. |
21 |
|
22 |
What I see is that people want to be able to use -@GROUP, which is perfectly |
23 |
understandable. The actual behaviour of -@GROUP in my proposal is imho quite |
24 |
a bit more intuitive. It wouldn't have to be discouraged. |
25 |
|
26 |
I also read that some people only saw the need for negative use flags in |
27 |
case the use flags themselves are expressed as a negation (nojoystick). |
28 |
I haven't been involved enough with these things to see if there really |
29 |
is a good reason to have those kinds of use flags at all. |
30 |
If there isn't, then the definition that I prefer for -flag is "exclude flag |
31 |
from this group" and that is both sufficient and easily understandable. |
32 |
|
33 |
The people who do not understand the exact meaning of all this would |
34 |
still be able to use -@GROUP in their use flags and it would work as |
35 |
they would expect. |
36 |
|
37 |
> | I would specify it like this: |
38 |
> | |
39 |
> | KDE $X kde qt -gtk -gnome |
40 |
> | GNOME $X gtk gtk2 gnome -kde -qt |
41 |
> | |
42 |
> | USE="@KDE @GNOME" |
43 |
> | |
44 |
> | That would result in: |
45 |
> | USE="X kde qt gtk gtk2 gnome" |
46 |
> | |
47 |
> | Exactly what the user wants. |
48 |
> |
49 |
> Hm. Although, I don't think "I want KDE" should imply "I don't want |
50 |
> GNOME" anyway... |
51 |
|
52 |
Right. It's just that even with screwy group definitions, there's still |
53 |
a good chance that the result is fine. It's easier to debug if the -flags |
54 |
don't carry over to other groups. |
55 |
|
56 |
It should be reasonably simple to write an interpreter for these flags |
57 |
and groups. Unfortunately I don't know python, but I'd be willing to |
58 |
make an implementation in C or something if that would help. |
59 |
|
60 |
Can anyone see any disadvantages of this method? |
61 |
Let me know what you think. |
62 |
|
63 |
Maurice. |
64 |
|
65 |
-- |
66 |
Maurice van der Pot |
67 |
|
68 |
Gentoo Linux Developer griffon26@g.o http://www.gentoo.org |
69 |
Creator of BiteMe! griffon26@××××××××.com http://www.kfk4ever.com |