cabal-install uses internal libraries.#3533
Conversation
|
LGTM, but build bot failures will need to be fixed.
Nice! |
ezyang
force-pushed
the
cabal-use-convenience
branch
from
76df543 to
416cbd5
Compare
|
Build failures were due to CI starting to build Haddocks for our internal library, and thus finding out all the places where we got the Haddock syntax wrong. I think I've nailed all of them. |
|
So I'm guessing that this means (assuming it's merged) that we could also do so for the solver-only bits? (That is, the (I'd be happy to do a PR for that, time permitting.) |
Yes.
Please do! |
ezyang
force-pushed
the
cabal-use-convenience
branch
2 times, most recently
from
a0a7118 to
6d88c67
Compare
There are lots of benefits:
* We only build solver code once, instead of repeatedly
rebuilding it for each test suite.
* Duplication in the cabal-install file is greatly
reduced.
There is one downside, which is that we don't rebuild
cabal-install with asserts turned on. I think we should
look into making this a runtime toggleable chunk of code.
Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>
A side effect of making cabal-install an internal library is that Haddock now generates documentation for it. Make sure that all of our comments are valid! Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>
Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>
…ternal libraries. Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>
ezyang
force-pushed
the
cabal-use-convenience
branch
from
6d88c67 to
13a9fe1
Compare
|
I guess even when I get this green, we can't do this for 1.26, because the need for internal library support means that you can't "cabal install cabal-install" from 1.24 (which doesn't understand internal libraries.) So we need one intervening release before we can put this in. |
|
Ah, true. Guess it'll have to wait for post-1.26. |
|
Well, in principle we can release cabal-install 1.26 faster than originally planned (e.g. in September). |
|
Won't this need a |
|
Yes, but I don't see this as a problem. |
|
Well, I'm going to close this for now, no way to merge. |
|
We should revive this after the next release. |
4 participants
There are lots of benefits:
There is one downside, which is that we don't rebuild
cabal-install with asserts turned on. I think we should
look into making this a runtime toggleable chunk of code.
Signed-off-by: Edward Z. Yang ezyang@cs.stanford.edu