1 |
Hi all, |
2 |
|
3 |
I have a question that I wanted to ask here before I open a potentially |
4 |
erroneous bug about it. |
5 |
|
6 |
For the ipython nbconvert command I need pandoc, which depends on various |
7 |
Haskell packages. Yesterday ghc was upgraded to 7.6.3-r1, which triggered |
8 |
rebuilds (AFAIU due to sub-slot dependencies). The interesting thing is that |
9 |
dev-haskell/parsec is a dependency of dev-haskepp/network, but was *not* built |
10 |
before it, such that my emerge @world died at that point. I ran |
11 |
haskell-updater afterwards, which got the order right, so that ipython |
12 |
nbconvert still works. |
13 |
|
14 |
For completeness, here are the dependencies of the network-2.4.1.2, copied |
15 |
straight out of the ebuild: |
16 |
|
17 |
RDEPEND=">=dev-haskell/parsec-3.0:=[profile?] |
18 |
>=dev-lang/ghc-6.10.4:=" |
19 |
DEPEND="${RDEPEND} |
20 |
>=dev-haskell/cabal-1.8 |
21 |
test? ( dev-haskell/hunit |
22 |
dev-haskell/test-framework |
23 |
dev-haskell/test-framework-hunit |
24 |
dev-haskell/test-framework-quickcheck2 |
25 |
)" |
26 |
|
27 |
From the above I would think that, since $RDEPEND is a subset of $DEPEND, |
28 |
parsec should have been rebuilt *before* network. Am I wrong about that? |
29 |
|
30 |
I guess I'm wondering whether this is a bug in the ebuild (I don't know the |
31 |
details of the sub-slot syntax), a bug in portage, or neither. |
32 |
|
33 |
Greetings |
34 |
-- |
35 |
Marc Joliet |
36 |
-- |
37 |
"People who think they know everything really annoy those of us who know we |
38 |
don't" - Bjarne Stroustrup |