search

commercialhaskell / stackage

Stable Haskell package sets: vetted consistent packages from Hackage
https://www.stackage.org/
MIT License
530 stars 805 forks source link

haskell-src-exts 1.18 #1734

Closed DanBurton closed 8 years ago

DanBurton commented 8 years ago

haskell-src-exts-1.18.0 is out of bounds for:

ndmitchell commented 8 years ago

hlint-1.9.36 released which fixes one HSE issue.

ndmitchell commented 8 years ago

hoogle-5.0.2 released which fixes another HSE issue.

DanBurton commented 8 years ago

Also note that the haskell-src-exts test suite can be re-enabled as well once constraint is lifted.

mgajda commented 8 years ago

Homplexity has been updated. Michal

Pozdrawiam Michał

RyanGlScott commented 8 years ago

Be aware that pointful-1.0.8 doesn't build correctly with haskell-src-exts-1.18: https://github.com/23Skidoo/pointful/issues/10

ndmitchell commented 8 years ago

Be aware that due to the changes in haskell-src-exts nothing on this list is likely to build correctly without some changes.

RyanGlScott commented 8 years ago

Right, I only mention pointful since it's not on the list at the moment. (The haskell-src-exts upper bound for pointful-1.0.8 has since been revised, however.)

ndmitchell commented 8 years ago

The page at https://www.stackage.org/lts-6.9/package/haskell-src-exts-1.17.1 shows there are 137 packages. However, not all of those are themselves in Stackage. Any way to get the list of packages that are both in Stackage and depend on haskell-src-exts? Perhaps the Stackage website should list the depended upon in two separate sections. Essentially all 137 packages are broken by this release, the question is how many also impact stackage.

snoyberg commented 8 years ago

Here you go (copied from https://raw.githubusercontent.com/fpco/stackage-nightly/master/nightly-2016-07-27.yaml):

haskell-src-exts: users:

ndmitchell commented 8 years ago

That suggests the following authors need to at least test with HSE, and almost certainly need to make code modifications:

yamadapc commented 8 years ago

file-modules compiles and works with 1.17.1.

Daniel-Diaz commented 8 years ago

Yeah, a user just reported me that haskintex was failing to build. It was because of haskell-src-exts. I uploaded a new version (0.7.0.0) that works with the newest haskell-src-exts.

k-bx commented 8 years ago

Thanks, will get back with hprotoc fixes in 2-3 years estimated https://twitter.com/ko_bx/status/759717177381515264 !

mgajda commented 8 years ago

Hi Konstantin It is just parametrizing all AST types by new SrcLoc/SrcSpan etc Took me just a weekend to fix - please Gitter me privately if You need help. Cheers Michal On Mon, 1 Aug 2016 at 4:09 AM, Kostiantyn Rybnikov notifications@github.com wrote:

Thanks, will get back with hprotoc fixes in 2-3 years estimated https://twitter.com/ko_bx/status/759717177381515264 !

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/fpco/stackage/issues/1734#issuecomment-236453618, or mute the thread https://github.com/notifications/unsubscribe-auth/AAp12VZREfVqLr6YO7hbplY7uyjwdtkMks5qbQDugaJpZM4JTsh1 .

jaspervdj commented 8 years ago

stylish-haskell has been updated.

phadej commented 8 years ago

@jaspervdj sometimes it's worth to look on pull requests tab: https://github.com/jaspervdj/stylish-haskell/pull/121 :(

jaspervdj commented 8 years ago

@phadej Ah, somehow I missed that mail, sorry :-(

ndmitchell commented 8 years ago

I've requested removal of derive from Stackage, which solves the upper bound issue: fpco/stackage#1772

I note we are starting to drift towards the 1 month ideal upper bound on how long such upgrades take.

phadej commented 8 years ago

@yamadapc the question is about 1.18!

bergmark commented 8 years ago

I don't think I'll have time to fix fay so feel free to disable it if need be.

chrisdone commented 8 years ago

Next version of hindent (5.0.0) when published will work with 1.18.

hvr commented 8 years ago

@chrisdone jfyi, hindent-0.4.6 works fine with GHC 8 after fixing the upper bounds. Unfortunately you locked the issue at https://github.com/chrisdone/hindent/issues/203 (why?). So I can't paste the new build-matrix report demonstrating this... :-/

chrisdone commented 8 years ago

hindent 5.0.0 published, works with 1.18.

yamadapc commented 8 years ago

Sorry, I think missed the version nº. file-modules@0.1.2.4 supports 1.18

bergmark commented 8 years ago

Turns out it was pretty easy so fay has been fixed!

bergmark commented 8 years ago

New entries:

bergmark commented 8 years ago

With all the hindent buzz flying around I think we should try to get the latest version in nightly ASAP. That depends on closing this issue.

phischu commented 8 years ago

I have published haskell-names-0.8.0 which is compatible with haskell-src-exts-1.18.*, sorry for my ignorance what do I have to do now?

bergmark commented 8 years ago

thanks @phischu! That's all AFAIK, we'll let you know if anything else pops up.

bergmark commented 8 years ago

Unfortunately removing this upper bound is disabling packages of unseen proportions, granted some of this might be because of benchmarks/test suites. But still, read and weep. I'l try to figure out if there's anything we can or should have done to make this more transparent...

bergmark commented 8 years ago

It is in some sense much better than it looks ;-) It just turns out that there are a lot of transitive dependencies on haskell-src-exts. And some very popular packages are included here.

@ndmitchell: Worth noting, is that derive is still an issue. We removed it from build-constraints, but it is still included since other packages depend on it. It is obvious in hindsight that we should have put in a "derive < 0" constraint to test the removal, and I suppose the reason we missed this is that very few packages are explicitly removed from stackage. As I understand it you just removed derive because of lack of time, correct?

I'm sure Neil would appreciate if someone in this list could help out!

derive depended on by:

haskell-src-meta doesn't have a stackage maintainer so what we should have done is check the transitive dependencies at once to inform these people that they together are responsible for it. Upstream issue: https://github.com/bmillwood/haskell-src-meta/issues/57

haskell-src-meta depended on by:

@stepcut: Pretty much your whole package set is affected because of hsx2hs which you also maintain, so that one is on you unfortunately :)

hsx2hs depended on by:

bitemyapp commented 8 years ago

@bergmark Ah, yes, hello. I've been summoned. What's the CTA here? What do you want me to do?

tmhedberg commented 8 years ago

What action needs to be taken for packages which depend on haskell-src-meta?

bergmark commented 8 years ago

@bitemyapp bloodhound among others depend on derive which doesn't allow haskell-src-exts 1.18.*, so conjointly the maintainers with packages depending on it should make sure a fix gets released, or remove the dependency (which is probably impractical in this case).

@tmhedberg Likewise for haskell-src-meta dependent maintainers, someone needs to write a patch to support haskell-src-exts 1.18.*.

As a last resort; if the maintainer is not accessible hackage trustees can help out. For that, create an issue in the trustee repo

JoeyEremondi commented 8 years ago

I'm not the maintainer of QuasiText, I'd tried to add it to stackage in an attempt to get Elm to build on the FPComplete online editor. That's obviously not going to happen now, so if it needs to be removed it won't break anything for me.

In short, I don't have the knowledge or use-case to maintain QuasiText or haskell-src-exts.

ndmitchell commented 8 years ago

I ported 80% of derive, then got bored - the code is at https://github.com/ndmitchell/derive/issues/19. I suspect it's quite feasible to finish the port.

I also advertised for an alternative maintainer at https://github.com/ndmitchell/derive/issues/6, to which @tomberek and @mgajda responded with positive noises.

My main reason for removing the package from Stackage is because I don't use it, and I'd like to trim my package set. I would encourage others to take it forward, and might be persuaded to give it one last shove over this line, but probably not the next line or the one after that.

dorchard commented 8 years ago

FYI, I just realised that fortran-src doesn't actually need the derive dependency anymore, so I have removed that. Re my haskell-src-meta dependency for codo-notation- I can look into that soon.

MichaelXavier commented 8 years ago

Its worth noting that I'm responsible for adding the derive dependency for ua-parser and probably bloodhound since I work on that also. Its only used for the test suite and I use it to derive quickcheck instances. I do this so that when new constructors and fields get added to types that the tests fail if they are not accounted for. There have been efforts to add Generic-based Arbitrary derivations that would allow for dropping derive in my use cases but they have stalled out.

MichaelXavier commented 8 years ago

I've gone ahead and submitted a mergeable PR on QuickCheck. If that PR gets merged, I'd be willing to update bloodhound, ua-parser, and uri-bytestring test suites to use the new QuickCheck rather than derive. Do note that because it would be a brand new QuickCheck, I'd have to put an extra-deps in the stack.yaml until nightlies got caught up. I'm not sure how the logistics on the stackage side would work out there. I'm either going to use derive or new QuickCheck, but I'm certainly not going to write out all those instances by hand :)

ndmitchell commented 8 years ago

@MichaelXavier - that seems a much better way to go forward long term. You are using about 0.5% of derive, so it's cracking quite a small nut with a huge hammer I don't want to maintain :)

If anyone else is blocked because of derive, let me know, and I can see what can be done - otherwise I've paused my efforts and left the https://github.com/ndmitchell/derive/tree/hse-1.18 branch around if anyone wants to pick up the work.