1 |
On Thu, 2004-05-13 at 00:13, Jon Portnoy wrote: |
2 |
> On Wed, May 12, 2004 at 10:55:37PM -0500, Andrew Gaffney wrote: |
3 |
> > Jon Portnoy wrote: |
4 |
> > >On Wed, May 12, 2004 at 10:27:48PM -0500, Andrew Gaffney wrote: |
5 |
> > > |
6 |
> > >>I was wondering why there was the separation between the compile-time |
7 |
> > >>depends (DEPEND) and the run-time depends (RDEPEND). Aren't all of them |
8 |
> > >>needed in order to run the program? Is there a practical reason for |
9 |
> > >>splitting them up? |
10 |
> > > |
11 |
> > >DEPEND is for *building* the package; consider binary packages. |
12 |
> > |
13 |
> > Yes, but run-time dependencies can be installed before the package itself. |
14 |
> > I still don't see a reason for splitting them up. |
15 |
> > |
16 |
> |
17 |
> Because you don't want to include build-time dependencies when you |
18 |
> distribute the binary package. |
19 |
|
20 |
Here's an example if you're not totally clear on what Jon is saying ( |
21 |
which is 100% correct BTW ). Pkgconfig is needed to build XFce, but not |
22 |
needed to run it (a DEPEND, not an RDEPEND), so there is no need to |
23 |
include or install pkgconfig from the GRP set as it would just take up |
24 |
precious space/bloat on the cd's and the binary user's computer/HD. |
25 |
This also comes into play with reverse deps and depclean. How would you |
26 |
be able to track what packages were no longer needed at runtime without |
27 |
discriminating between the two? This leads to a much cleaner secure |
28 |
system. |
29 |
-- |
30 |
Brad Cowan <bcowan@g.o> |
31 |
Developer, |
32 |
Gentoo Linux http://www.gentoo.org/~bcowan |
33 |
|
34 |
Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xB1F16A56 |
35 |
Key fingerprint = C408 75B9 E68D 26E2 EAAE 20CF 4D5E 293D B1F1 6A56 |