Move solver types to Distribution.Solver.* namespace #3406
Conversation
| -- Stability : provisional | ||
| -- Portability : portable | ||
| -- | ||
| -- Common types for dependency resolution. |
There was a problem hiding this comment.
I'm okay with removing the header, but it'd be nice to preserve the top comment about the purpose of the module.
There was a problem hiding this comment.
I actually removed it because the "purpose" isn't very clear any longer. The module is now a sort of weird mish-mash.
(I think this module should perhaps also be split, but I'm kind of waiting to see where migrating the Tests takes me... The module may become entirely redundant.)
|
Actually, perhaps we can drop the Travis check and just use an "internal library"? @ezyang Do you think "internal library" support is robust enough at this point to move the Dependency.Solver.* bits into such a library? (I'm assuming tests will have access to such a library implicitly. Right?) |
|
LGTM, but would be nice to see some of the tiny modules recombined later on. |
|
Yup, I fully intend to at some point, but I think it'll have to wait to see what happens when I try migrating the test cases, etc. I'm not sure we actually have the "full set" yet. |
|
Hm, has Travis been disabled or something? I'm not seeing any "CI status" on this PR. |
|
|
|
Actually don't do that, I found a way to fix this. |
|
Check that. Thanks! |
|
OK, for some reason I can't re-trigger the Travis hook manually (only AppVeyor), but closing and reopening this PR should work. |
|
Merged, thanks! |
|
Re adding an internal library, I think we just need to merge #3356 |
|
OK, sounds good. I'll postpone the attempt until that's merged then :). |
This completes the move of all the solver types which the modular solver requires to Distribution.Solver.*.
Like last time there are few tiny modules, but again I've decided to go for consistency; see #3383.
This means that we now have:
$ grep -r Client cabal-install/Distribution/Solver|wc -l 0which is to say that there are no longer any dependencies from the Solver namespace to the Client namespace. We should probably add a Travis check or something that ensures that none are added back accidentally.