1 |
On Sat, 2 Feb 2013 23:33:26 +0000 |
2 |
"Aaron W. Swenson" <titanofold@g.o> wrote: |
3 |
|
4 |
> After years of "if use test ; then ..." just working when |
5 |
> FEATURES="test" is declared, it isn't working with EAPI5. |
6 |
|
7 |
You shouldn't admit that for years you didn't knew that this was |
8 |
incorrect and you should have been using IUSE=test. |
9 |
|
10 |
> I think we could save some bytes and headaches if we just add "test" |
11 |
> to IUSE_IMPLICIT. |
12 |
|
13 |
First of all, you should note that you will still need to add IUSE=test |
14 |
to pre-EAPI 5 ebuilds. |
15 |
|
16 |
Secondly, what about all the ebuilds which declare IUSE=test in EAPI 5? |
17 |
Shall we remove that value from IUSE? Keep it? |
18 |
|
19 |
What will be the impact on metadata? It seems that the PMS allows |
20 |
dependencies on IUSE_EFFECTIVE, so we can basically have dependencies |
21 |
with flags which are valid only on some of the profiles... |
22 |
|
23 |
> Portage's emerge's "--newuse" option won't be affected by this. From |
24 |
> `man emerge`: |
25 |
> > NOTE: This option ignores the state of the "test" USE flag, since |
26 |
> > that flag has a special binding to FEATURES="test" (see make.conf(5) |
27 |
> > for more information about FEATURES settings). |
28 |
> |
29 |
> What say you? |
30 |
|
31 |
I say that putting semi-automatic, dynamically changing flag |
32 |
into IUSE_IMPLICIT is a bad idea. Right now, we just keep the things |
33 |
that are really not supposed to change there. |
34 |
|
35 |
-- |
36 |
Best regards, |
37 |
Michał Górny |