GHC 7.8 cabal restrictive constraints

[snoyberg]

I'm superseding #43 with this. As the GHC 7.8 release is becoming more imminent, these issues are now becoming more important.

BlogLiterately-diagrams-0.1.2.3 (Brent Yorgey @byorgey byorgey@gmail.com) cannot use:

• base-4.7.0.0 -- >=4.3 && <4.7

ChasingBottoms-1.3.0.6 (Neil Mitchell @ndmitchell) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5
• base-4.7.0.0 -- >=4.0 && <4.7

accelerate-0.13.0.5 (Ryan Newton ryan.newton@alum.mit.edu @rrnewton @AccelerateHS) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

active-0.1.0.7 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

ad-3.4 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.2 && <0.5
• template-haskell-2.9.0.0 -- >=2.5 && <2.9

adjunctions-3.2.1 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

arithmoi-0.4.0.3 (Brent Yorgey byorgey@gmail.com) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

async-2.0.1.4 (FP Complete michael@fpcomplete.com @simonmar) cannot use:

• base-4.7.0.0 -- >=4.3 && <4.7

cassava-0.2.2.0 (FP Complete michael@fpcomplete.com @tibbe) cannot use:

• array-0.5.0.0 -- <0.5

charset-0.3.5 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.2 && <0.5

circle-packing-0.1.0.2 (Brent Yorgey byorgey@gmail.com) cannot use:

• base-4.7.0.0 -- >=4.5 && <4.7

comonad-extras-3.0.1 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

data-lens-template-2.1.7 (Adam Bergmark adam@edea.se @roconnor) cannot use:

• template-haskell-2.9.0.0 -- >=2.4 && <2.9

diagrams-builder-0.4.0.6 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• base-4.7.0.0 -- >=4.2 && <4.7

diagrams-cairo-0.7 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• base-4.7.0.0 -- >=4.2 && <4.7
• unix-2.7.0.0 -- >=2.4 && <2.7

diagrams-contrib-0.7 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• base-4.7.0.0 -- >=4.2 && <4.7

diagrams-haddock-0.2.1 (Brent Yorgey byorgey@gmail.com) cannot use:

• base-4.7.0.0 -- >=4.4 && <4.7

diagrams-lib-0.7.1.1 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

encoding-0.6.7.2 (FP Complete michael@fpcomplete.com) cannot use:

• binary-0.7.1.0 -- <0.6

esqueleto-1.3.4.2 (Felipe Lessa felipe.lessa@gmail.com @meteficha) cannot use:

• base-4.7.0.0 -- >=4.5 && <4.7

fclabels-1.1.7.1 (Ryan Newton ryan.newton@alum.mit.edu @sebastiaanvisser) cannot use:

• template-haskell-2.9.0.0 -- >=2.2 && <2.9

force-layout-0.2 (Brent Yorgey byorgey@gmail.com) cannot use:

• base-4.7.0.0 -- >=4.2 && <4.7

graphs-0.5 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

haskeline-0.7.0.3 (Adam Bergmark adam@edea.se @bergmark) cannot use:

• base-4.7.0.0 -- >=4.1 && <4.7
• unix-2.7.0.0 -- >=2.0 && <2.7

heist-0.13.0.1 (FP Complete michael@fpcomplete.com @snapframework @mightybyte) cannot use:

• process-1.2.0.0 -- ==1.1.*

hscurses-1.4.1.2 (Stefan Wehr wehr@factisresearch.com @skogsbaer) cannot use:

• unix-2.7.0.0 -- >=2.4 && <2.7

http-types-0.8.1 (michael@snoyman.com @aristidb) cannot use:

• array-0.5.0.0 -- >=0.2 && <0.5

intervals-0.2.2 (Brent Yorgey byorgey@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

kan-extensions-3.7 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

keter-1.0.1.1 (michael@snoyman.com @snoyberg) cannot use:

• unix-2.7.0.0 -- >=2.5 && <2.7

keys-3.0.3 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

language-javascript-0.5.8 (Alan Zimmerman @alanz) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

lens-3.9.2 (Brent Yorgey byorgey@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

monadic-arrays-0.2.1.2 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

optparse-applicative-0.5.2.1 (Adam Bergmark adam@edea.se @pcapriotti) cannot use:

• process-1.2.0.0 -- ==1.1.*

parallel-3.2.0.3 (Brent Yorgey byorgey@gmail.com) cannot use:

• array-0.5.0.0 -- >=0.1 && <0.5

reducers-3.0.2 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

repa-3.2.3.3 (FP Complete michael@fpcomplete.com) cannot use:

• base-4.7.0.0 -- ==4.6.*
• template-haskell-2.9.0.0 -- ==2.8.*

representable-functors-3.2.0.1 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

snap-0.13.0.3 (FP Complete michael@fpcomplete.com @snapframework) cannot use:

• template-haskell-2.9.0.0 -- >=2.2 && <2.9

stringsearch-0.3.6.4 (michael@snoyman.com) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

temporary-1.1.2.4 (Brent Yorgey byorgey@gmail.com) cannot use:

• unix-2.7.0.0 -- >=2.3 && <2.7

th-lift-0.5.5 (Felipe Lessa felipe.lessa@gmail.com) cannot use:

• template-haskell-2.9.0.0 -- >=2.4 && <2.9

type-eq-0.3 (Adam Bergmark adam@edea.se @glehel) cannot use:

• base-4.7.0.0 -- >=3.0 && <4.7

unbounded-delays-0.1.0.5 (FP Complete michael@fpcomplete.com @basvandijk) cannot use:

• base-4.7.0.0 -- >=4 && <4.7

vault-0.3.0.2 (michael@snoyman.com @HeinrichApfelmus) cannot use:

• base-4.7.0.0 -- >=4.5 && <4.7

[ndmitchell]

I have never heard of ChasingBottoms before today, so suspect you are after NilsAndersDanielsson rather than me for that.

[ekmett]

So far I've closed out:

• reducers
• monadic-arrays
• charset
• trifecta
• kan-extensions
• comonad-extras
• intervals
• graphs

I'll keep chipping away as time permits.

[ekmett]

Fixed:

• representable-functors
• keys
• adjunctions

This leaves two big projects that fall under my purview that have large open change sets and bigger issues: lens and ad.

Each of those will likely take a fair bit more time to fix.

[snoyberg]

@ndmitchell ChasingBottoms is used by the unordered-containers test suite. It's complete chance who that package would end up being assigned to.

[ndmitchell]

I have had zero success contacting Nils - if anyone has a working email address for him, please do share.

[snoyberg]

Nils just merged and released a patch I sent him, so ChasingBottoms is now fixed: http://hdiff.luite.com/cgit/ChasingBottoms/commit/?id=079e8ca9cdb6e6d9c75e4a4e89921fdb35555577

[snoyberg]

OK, here's the newest set of issues:

FenwickTree-0.1 (Michal J. Gajda @mgajda) cannot use:

• base-4.7.0.0 -- <4.7

Octree-0.5.1 (Michal J. Gajda @mgajda) cannot use:

• base-4.7.0.0 -- >=4.0 && <4.7

accelerate-0.14.0.0 (Ryan Newton ryan.newton@alum.mit.edu @AccelerateHS) cannot use:

• base-4.7.0.0 -- ==4.6.*

ad-3.4 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.2 && <0.5
• template-haskell-2.9.0.0 -- >=2.5 && <2.9

arithmoi-0.4.0.3 (Brent Yorgey byorgey@gmail.com) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

async-2.0.1.4 (FP Complete michael@fpcomplete.com @simonmar) cannot use:

• base-4.7.0.0 -- >=4.3 && <4.7

cassava-0.2.2.0 (FP Complete michael@fpcomplete.com @tibbe) cannot use:

• array-0.5.0.0 -- <0.5

circle-packing-0.1.0.2 (Brent Yorgey byorgey@gmail.com) cannot use:

• base-4.7.0.0 -- >=4.5 && <4.7

data-lens-template-2.1.7 (Adam Bergmark adam@bergmark.nl @roconnor) cannot use:

• template-haskell-2.9.0.0 -- >=2.4 && <2.9

diagrams-builder-0.4.2 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• base-4.7.0.0 -- >=4.2 && <4.7

diagrams-cairo-1.0 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• unix-2.7.0.0 -- >=2.4 && <2.7

diagrams-lib-1.0.0.1 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

esqueleto-1.3.4.2 (Felipe Lessa felipe.lessa@gmail.com @meteficha) cannot use:

• base-4.7.0.0 -- >=4.5 && <4.7

fay-0.18.1.2 (Adam Bergmark adam@bergmark.nl @faylang) cannot use:

• process-1.2.0.0 -- <1.2

fclabels-2.0.0.5 (Ryan Newton ryan.newton@alum.mit.edu @sebastiaanvisser) cannot use:

• template-haskell-2.9.0.0 -- >=2.2 && <2.9

force-layout-0.2 (Brent Yorgey byorgey@gmail.com) cannot use:

• base-4.7.0.0 -- >=4.2 && <4.7

hPDB-examples-1.0 (Michal J. Gajda @mgajda) cannot use:

• base-4.7.0.0 -- <4.7

heist-0.13.0.3 (Adam Bergmark adam@bergmark.nl @snapframework) cannot use:

• process-1.2.0.0 -- ==1.1.*

hscurses-1.4.1.2 (Stefan Wehr wehr@factisresearch.com @skogsbaer) cannot use:

• unix-2.7.0.0 -- >=2.4 && <2.7

kan-extensions-3.7 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

language-javascript-0.5.8 (Alan Zimmerman @alanz) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

lens-3.9.2 (Brent Yorgey byorgey@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

lifted-async-0.1.0 (michael@snoyman.com @maoe) cannot use:

• base-4.7.0.0 -- >=4.3 && <4.7

quickcheck-instances-0.3.3 (Mario Blazevic blamario@yahoo.com @aslatter) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

repa-3.2.3.3 (FP Complete michael@fpcomplete.com) cannot use:

• base-4.7.0.0 -- ==4.6.*
• template-haskell-2.9.0.0 -- ==2.8.*

snap-0.13.1.2 (FP Complete michael@fpcomplete.com @snapframework) cannot use:

• template-haskell-2.9.0.0 -- >=2.2 && <2.9

temporary-1.1.2.4 (Roman Cheplyaka roma@ro-che.info) cannot use:

• unix-2.7.0.0 -- >=2.3 && <2.7

type-eq-0.3 (Adam Bergmark adam@bergmark.nl @glehel) cannot use:

• base-4.7.0.0 -- >=3.0 && <4.7

unbounded-delays-0.1.0.5 (FP Complete michael@fpcomplete.com @basvandijk) cannot use:

• base-4.7.0.0 -- >=4 && <4.7

vault-0.3.0.2 (michael@snoyman.com @HeinrichApfelmus) cannot use:

• base-4.7.0.0 -- >=4.5 && <4.7

[mgajda]

Thanks. These three have been fixed by uploading new versions.

On 12/09/2013 12:21 PM, Michael Snoyman wrote:

FenwickTree-0.1 (Michal J. Gajda @mgajda https://github.com/mgajda) cannot use:

• base-4.7.0.0 -- <4.7

Octree-0.5.1 (Michal J. Gajda @mgajda https://github.com/mgajda) cannot use:

• base-4.7.0.0 -- >=4.0 && <4.7

hPDB-examples-1.0 (Michal J. Gajda @mgajda https://github.com/mgajda) cannot use:

• base-4.7.0.0 -- <4.7

[glaebhoerl]

• Probably a copypasteo: I'm not Adam Bergmark
• I suspect type-eq is not compatible with GHC 7.8 in its current form due to the Typeable changes. At the same time, most of the package's functionality now exists directly in base, in particular, the only reverse dependency that I know of (haskell-names) only uses dynamicEq, which exists as eqT in the new base. Given this, is it worthwhile to make type-eq compatible with GHC 7.8,or should haskell-names just be moved to using the equivalent functionality from base instead? (Does stackage intend to support multiple versions of GHC simultaneously? I don't know how any of this works.:) cc @feuerbach
• How urgent is this?

[sebastiaanvisser]

Is there an API-freeze on template-haskell already? How can I know?

At first fclabels package was broken with the new TH, then it got fixed automatically because they reversed the API changes (I guess). If TH is still in flux it doesn't make sense for me to test my package and increase the upper bounds for now.

[snoyberg]

@glehel I do intend to support versions 7.4, 7.6, and 7.8 of GHC simultaneously. It's possible for some packages to be selectively excluded based on the GHC version. So whatever you and @feuerbach decide to do will be fine with me. It would be nice if these changes happened sooner rather than later so we can get some serious build testing with GHC 7.8.

@sebastiaanvisser I think TH is pretty stable at this point, though there aren't any guarantees.

[UnkindPartition]

@glehel I know this functionality is in the new base, but I (or, rather, my users) would like to keep compatibility with older GHCs (7.4 and 7.6).

It would be nice if type-eq could provide this compatibility, but if you decide to discontinue the package, I can hack around it in haskell-names itself.

@sebastiaanvisser nice pic :)

[glaebhoerl]

@snoyberg @feuerbach OK, I'll see what I can do (and when).

[maoe]

I just relaxed the upper bound for base in lifted-async and uploaded it to Hackage. It still fails to build with GHC HEAD because async fails, though.

[bergmark]

process shouldn't be a problem for fay, but i'll wait until haskell-names is ok.

[glaebhoerl]

If anyone could point me to an x86-32 linux build of a recent GHC 7.8 nightly/snapshot I can download, that would be a big help.

Edit: not relevant any more

[glaebhoerl]

Sorry for taking so long, I finally had some time to do this. I've uploaded type-eq-0.4 which works with GHC 7.8 and base-4.7.

How to handle the Typeable changes is quite perplexing for a package like this. For now I'm just following the path of least resistance, which basically means that under GHC 7.8 plain Typeable means the new kind-polymorphic one from Data.Typeable, while the formerly provided functions using Typeable1..7 are still provided, and get them from Data.OldTypeable.

There was only one function that haskell-names actually uses though, so hopefully this should be ok.

(btw: hi, I renamed myself, I used to be glehel)

[aslatter]

I've uploaded quickcheck-instances-0.3.5 which has more permissive upper bounds for array.

[UnkindPartition]

Thanks @glaebhoerl, that was pretty fast!

[UnkindPartition]

I've uploaded haskell-names-0.3.2.4 which builds with GHC 7.8, although --allow-newer is needed for temporary (unix dependency, cc @batterseapower) and data-lens-template (template-haskell dependency, cc @roconnor)

[byorgey]

Uploaded force-layout-0.2.0.1.

[byorgey]

cirlcle-packing-0.1.0.3 has been uploaded. Currently looking into arithmoi which is affected by some primop changes in base-4.7. See https://bitbucket.org/dafis/arithmoi/issue/4/equality-primop-type-error-under-base-47 .

[sebastiaanvisser]

fclabels-2.0.1 has been uploaded. The dependency on template-haskell has been widened.

[byorgey]

The rest of my packages (diagrams-lib, diagrams-cairo, diagrams-builder) are stuck on lens. lens-3.10 defines Foldable and Traversable instances for (,), Either, and Const, but these are now provided in base-4.7.

[ekmett]

Let me fix that.

[ekmett]

Try 3.10.1

[byorgey]

@ekmett: I get a different compile error now. Should I file an issue over at the lens repo?

[ekmett]

Please do. I just fired off a fix for that one issue. I don't have 7.8 installed at the moment to test with.

[tibbe]

cassava-0.3.0.0 supports array-0.5.0.0. Do you also need new releases in the 0.2 series?

[snoyberg]

No, 0.3 should be fine. I was actually just running some Stackage builds, so I ran GHC 7.7.20131211, and got the following:

accelerate-0.14.0.0 (Ryan Newton ryan.newton@alum.mit.edu @AccelerateHS) cannot use:

• base-4.7.0.0 -- ==4.6.*

ad-3.4 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.2 && <0.5
• template-haskell-2.9.0.0 -- >=2.5 && <2.9

arithmoi-0.4.0.3 (Brent Yorgey byorgey@gmail.com) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

bytestring-show-0.3.5.5 (Ketil Malde) cannot use:

• array-0.5.0.0 -- <0.5

data-lens-template-2.1.7 (Adam Bergmark adam@bergmark.nl @roconnor) cannot use:

• template-haskell-2.9.0.0 -- >=2.4 && <2.9

diagrams-builder-0.4.2 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• base-4.7.0.0 -- >=4.2 && <4.7

diagrams-cairo-1.0 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• unix-2.7.0.0 -- >=2.4 && <2.7

diagrams-lib-1.0.0.1 (Brent Yorgey byorgey@gmail.com @diagrams) cannot use:

• array-0.5.0.0 -- >=0.3 && <0.5

esqueleto-1.3.4.2 (Felipe Lessa felipe.lessa@gmail.com @meteficha) cannot use:

• base-4.7.0.0 -- >=4.5 && <4.7

fay-0.18.1.3 (Adam Bergmark adam@bergmark.nl @faylang) cannot use:

• process-1.2.0.0 -- <1.2

heist-0.13.0.3 (Adam Bergmark adam@bergmark.nl @snapframework) cannot use:

• process-1.2.0.0 -- ==1.1.*

hscurses-1.4.1.2 (Stefan Wehr wehr@factisresearch.com @skogsbaer) cannot use:

• unix-2.7.0.0 -- >=2.4 && <2.7

kan-extensions-3.7 (Edward Kmett ekmett@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

lens-3.9.2 (Brent Yorgey byorgey@gmail.com @ekmett) cannot use:

• array-0.5.0.0 -- >=0.3.0.2 && <0.5

repa-3.2.3.3 (FP Complete michael@fpcomplete.com) cannot use:

• base-4.7.0.0 -- ==4.6.*
• template-haskell-2.9.0.0 -- ==2.8.*

snap-0.13.1.2 (FP Complete michael@fpcomplete.com @snapframework) cannot use:

• template-haskell-2.9.0.0 -- >=2.2 && <2.9

threepenny-gui-0.4.0.2 (FP Complete michael@fpcomplete.com) cannot use:

• template-haskell-2.9.0.0 -- >=2.7.0 && <2.9

unbounded-delays-0.1.0.5 (FP Complete michael@fpcomplete.com @basvandijk) cannot use:

• base-4.7.0.0 -- >=4 && <4.7

So we're getting a lot closer.

[bergmark]

fay 0.19.1 allows process 1.2.*

[ekmett]

What is capping lens at 3.9.2, diagrams?

[byorgey]

It was diagrams, but the latest diagrams release (as of earlier this week) uses lens-4.0. I don't know if there's anything else still blocking it.

[ekmett]

Fair enough. I wasn't actively doing maintenance on the 3.9 branch, so I was trying to figure out if I needed to start.

[snoyberg]

Barring a few specific packages which I've explicitly excluded for GHC 7.8, upper bounds are now resolved. I'm running a Stackage build with 7.8.0rc2 now, and (probably Sunday morning) will start providing feedback on broken packages.