1 |
On Sun, 2 Oct 2016 22:06:19 -0400 |
2 |
Ian Stakenvicius <axs@g.o> wrote: |
3 |
|
4 |
> On 02/10/16 04:59 PM, James Le Cuirot wrote: |
5 |
> > On Sun, 2 Oct 2016 21:48:04 +0100 |
6 |
> > James Le Cuirot <chewi@g.o> wrote: |
7 |
> > |
8 |
> >> SRC_URI="gogdownloader://tomb_raider_1/en1installer1 -> |
9 |
> >> setup_tomb_raider_${PV}.exe" IUSE="gogdownloader" |
10 |
> >> RESTRICT="!gogdownloader? ( fetch ) mirror" |
11 |
> >> DEPEND="games-util/lgogdownloader" |
12 |
> > |
13 |
> > Probably goes without saying but that should have been: |
14 |
> > DEPEND="gogdownloader? ( games-util/lgogdownloader )" |
15 |
> > |
16 |
> |
17 |
> So I think the idea of supporting this is great, but I think the best |
18 |
> way to go about it is to support it in the same manner that the VCS |
19 |
> eclasses do -- that is, DONT use SRC_URI but instead use GOG_URI , and |
20 |
> have the gog eclass within src_unpack handle the actual fetching. |
21 |
> |
22 |
> If in a future EAPI we get src_fetch() and the ability to specify |
23 |
> custom fetching, then this can be improved, but for now I'd say do it |
24 |
> the same as the VCS eclasses do. |
25 |
|
26 |
src_fetch would be nice for VCS but I think it's overkill here as what |
27 |
we have already works fine. If Portage were as flexible with its |
28 |
configuration as Paludis, there would be no issue, save for doing away |
29 |
with the static list of protocols in the PMS. |
30 |
|
31 |
The simplest solution I can see would be to make make.globals a |
32 |
directory and then packages like lgogdownloader could just drop files |
33 |
in there. This could have other uses too. Can't think of any right now |
34 |
though. ;) |
35 |
|
36 |
Instead of dictating the list of supported protocols, PMS could then |
37 |
say something to the effect of: |
38 |
|
39 |
"As well as supporting the http://, https://, ftp:// protocols out of |
40 |
the box, the package manager must be sufficiently configurable to allow |
41 |
packages to extend this list by installing additional files." |
42 |
|
43 |
Different files would need to be installed to support both Portage |
44 |
(plus pkgcore) and Paludis but I could live with that. Otherwise |
45 |
Portage would probably need to implement fetchers the same way that |
46 |
Paludis does. The end result would be nice but could we seamlessly |
47 |
migrate existing user configuration? How many users actually change |
48 |
FETCHCOMMAND? |
49 |
|
50 |
http://paludis.exherbo.org/configuration/fetchers.html |
51 |
|
52 |
I understand that a new EAPI would still be required as there is |
53 |
otherwise no guarantee that the user will have a version of Portage |
54 |
with this fix, short of depending on sys-apps/portage, which is bad! In |
55 |
the meantime, I'll have to make do with src_unpack. I'll also file a |
56 |
bug so we can further discuss which avenue to take. |
57 |
|
58 |
-- |
59 |
James Le Cuirot (chewi) |
60 |
Gentoo Linux Developer |