vector-0.12.0.0 is out of bounds for:
Due to the large number of packages initially affected, there is no need to post update comments at this stage.
foldl-1.2.3 is fixed
Due to the large number of packages initially affected, there is no need to post update comments at this stage.
Since Stackage doesn't seem to record this info: the maintainer of vector-algorithms is @dolio. And since vector-algorithms doesn't use git: I have a patch which revises the upper version bounds on vector here.
(Having said that, personally as maintianer I find the update comments really useful, but of course many people will get notified via github.)
I suspect people can't update the checkboxes in my original comment?
@juhp No, at least I can't. AFAIK only collaborators can edit/update comments.
Now that I have a comment: sibe-0.2.0.4 is fixed. Waiting on JuicyPixels to fix picedit.
Update: picedit-0.1.1.2 is fixed, too.
@juhp Correct. Not sure if it's possible to allow others to update the description of the issue without making them members of the project sadly.
I believe I've made the necessary changes in foldl-statistics-0.1.4.2, and given Gabriel's update to foldl, I think these should both be fixed now. (Thanks @Gabriel439 for the quick fix).
Also, it appears vector-binary-instances has also been updated (an indirect dependency of my package).
@mdibaiee I've updated JuicyPixels' bounds on hackage (no need to make a new release), it compiles with Vector 0.12
cayley-client fixed.
Since we comment anyway: cassava-megaparsec, flac and stache fixed.
Thank you for all the Hackage updates: I have refreshed the initial comment.
hquantlib fixed but it requires packages scientific and integer-logarithms to bump their versions too.
@paulrzcz AFAIK neither scientific or integer-logarithms depend on vector? Maybe you are missinterpretting something (they depend thru criterion, which depends on cassava).
Generally, I hope that people actually test that their package is compiled against vector-0.12, e.g. there is no way to compile cassava-megaparsec against vector-0.12 without using --allow-never. Hopefully @mrkkrp did so. (cassava in fact seems to compile fine with vector-0.12).
EDIT: the dependency graph of the libraries might be helpful in such big issues, there is no much one can do as a package maintainer of some leaf package (which happens to have direct dependency as well).
@phadej After raising the vector version to 0.12.0.0 in extra-deps of stack.yaml and rebuilding the project with stack against Stackage LTS 7.15, I've got:
Error: While constructing the build plan, the following exceptions were encountered:
In the dependencies for scientific-0.3.4.9:
vector-0.12.0.0 must match >=0.7 && <0.12 (latest applicable is 0.11.0.0)
needed due to hquantlib-0.0.3.4 -> scientific-0.3.4.9
In the dependencies for vector-binary-instances-0.2.3.3:
vector-0.12.0.0 must match >=0.6 && <0.12 (latest applicable is 0.11.0.0)
needed due to hquantlib-0.0.3.4 -> vector-binary-instances-0.2.3.3
Plan construction failed.
Bumping of scientific to version 0.3.4.10 came up with integer-logarithm:
In the dependencies for scientific-0.3.4.10:
integer-logarithms must match >=1 && <1.1, but the stack configuration has no specified version
(latest applicable is 1)
needed due to hquantlib-0.0.3.4 -> scientific-0.3.4.10
The next step is to add interger-logarithms-1. So now hquantlib is buildable and has got extra-deps:
extra-deps:
- vector-0.12.0.0
- vector-binary-instances-0.2.3.4
- scientific-0.3.4.10
- integer-logarithms-1
Studying the full dependency graph shows that scientific come from aeson-0.11.2.1 package and takes integer-logarithms with it. vector-binary-instances are direct dep of statistics.
Though, it builds without additional version bumps against Stackage nightly-2017-01-10 that already includes those versions of packages, except vector-0.12.0.0.
@paulrzcz few things: use solver, e.g. stack solver, or cabal
hquantlib-0.0.3.3 master % git diff
diff --git a/hquantlib.cabal b/hquantlib.cabal
index a9deb2d..3885d56 100644
--- a/hquantlib.cabal
+++ b/hquantlib.cabal
@@ -72,7 +72,7 @@ library
parallel >= 3.2.0.0 && < 3.3.0.0,
mersenne-random >= 1.0.0.1 && < 2.0.0.0,
statistics >= 0.13.0.0 && < 0.14.0.0,
- vector >= 0.11.0.0 && < 0.12.0.0,
+ vector >= 0.11.0.0 && < 0.13.0.0,
vector-algorithms >= 0.7.0.0 && < 0.8.0.0
hs-source-dirs: src
[FL973] ~/mess/hquantlib-0.0.3.3 master % cabal new-build --only-dependencies --dry-run --constraint vector==0.12.0.0 --enable-tests|tail
- storable-complex-0.2.2 (lib:storable-complex) (requires download & build)
- vector-algorithms-0.7.0.1 (lib) (requires build)
- vector-th-unbox-0.2.1.6 (lib) (requires build)
- monad-par-extras-0.3.3 (lib) (requires build)
- hmatrix-0.18.0.0 (lib) (requires download & build)
- math-functions-0.2.1.0 (lib) (requires download & build)
- monad-par-0.3.4.8 (lib) (requires build)
- hmatrix-gsl-0.18.0.1 (lib) +onlygsl (requires download & build)
- statistics-0.13.3.0 (lib) (requires build)
- hmatrix-special-0.4.0.1 (lib:hmatrix-special) (requires download & build)
It's not visible in the output, but for vector-0.12 one need to pick scientific-0.3.4.10 and vector-binary-instances-0.2.3.4, and maybe some other ripples.
Another thing, I noticed in hquantlib:
mersenne-random >= 1.0.0.1 && < 2.0.0.0
But 2.0 < 2.0.0.0 holds. It's unfortunate legacy quirk (and hopefully cabal check will say about it, cc @hvr)
vectortiles updated.
smoothie updated.
New entry:
folds bound relaxed, also binary-tagged
These mentions were originally truncated: @NicolasT @codedmart @tmcgilchrist @acid-state @jtobin @polarina @isomorphism @klao @nilcons @bgmari @mstksg @expipiplusx @fhaust @mgajda
@bgamari *
these is updated though already ;) (the isomorphism's ping)
vector-split updated
sdl2 has been updated.
Whilst my package compiles nicely with vector-0.12.0.0, a change of behaviour in the vector package causes breakage. I hope this can be resolved within vector (as I wouldn't be surprised other code could be broken because of this change as well): https://github.com/haskell/vector/issues/152
sampling-0.3.2 is fixed.
tz and tzdata updated.
vector-sized with fix pushed to hackage
LTS 8 will be released soon, probably with vector-0.11.0.0.
After the LTS 8.0 release, the packages listed below (and any that depend on them) will continue to be included in LTS 8 builds, but will be dropped from nightly builds until a release compatible with vector-0.12.0.0 is made available.
@DanBurton we've got a release coming soon that already fixes the vector bump and can easily fix the hspec bump, we're just doing some house cleaning before-hand because it's a big change for unrelated-to-Stackage reasons.
Chart-1.8.2 uploaded.
I鈥檓 updating wavefront. Is that too late? :D
bloodhound-0.13.0.0 is released with vector-0.12.0.0 compatibility.
stb-image-redux is now vector-0.12.0.0 compatible.
I just pinged @gregorycollins again and he made a new release of hashtables that supports vector-0.12. I think this may unblock a large number of packages that are currently excluded. @DanBurton ?
(Also, I updated diagrams-cairo which was blocked on hashtables.)
Thanks a lot, @byorgey, for the update! Very helpful
Most helpful comment
Since Stackage doesn't seem to record this info: the maintainer of
vector-algorithmsis @dolio. And sincevector-algorithmsdoesn't usegit: I have a patch which revises the upper version bounds onvectorhere.