Support GHC 9.4 #69
Support GHC 9.4 #69
Conversation
|
@parsonsmatt Is is possible to remove the use of If you do not get to it in the next couple of days, I will have a go at doing it. |
|
Once I've got the work codebase on GHC 9.4 then I'll have time to come back around and apply fixups. I'd be thrilled to have less CPP for this, but I'm unsure how other approaches would potentially affect performance. Removing |
|
Would you be willing to do a squash+merge on GitHub side of things? I'm not great at squashing on the CLI. |
|
@parsonsmatt Never mind, I will sort this out (I am really good at squashing things from the CLI). As for |
|
@erikd I could try and squash it up if you'd like. If you have any suggestions on reducing CPP then I can take a stab at it. |
|
Any update? |
|
Sorry, no. At work I am still using |
|
Alas :( |
|
I'd be happy to help push this forward. You requested earlier that I remove I saw a trick to |
|
@parsonsmatt Absolutely. Would merge and release the solution was a little nicer than the original. Had even meant to do it myself, but then life got in the way. |
| @@ -65,6 +64,14 @@ import Data.Primitive.Types (Prim (..), defaultSetByteArray#, defaultSetOffAddr# | |||
|
|
|||
| import Data.Hashable (Hashable,hashWithSalt) | |||
|
|
|||
| import Data.WideWord.Compat | |||
There was a problem hiding this comment.
@erikd I've consolidated the compat stuff to minimize noise here. Thoughts?
There was a problem hiding this comment.
I like it. That is a significant improvement.
There was a problem hiding this comment.
The only thing I would still like to address is the CPP and ViewPatterns use in things like say Data.WideWord.Word256.signum256. I am sure there is a neater way to do that (in say that Compat module).
There was a problem hiding this comment.
Maybe something like:
signum256 :: Word256 -> Word256
signum256 (Word256 a b c d) =
if isZeroWord64 a && isZeroWord64 b && isZeroWord64 c && isZeroWord64 d
then zeroWord256
else oneWord256
With the isZeroWord64 being defined in the Compat module.
There was a problem hiding this comment.
Also wondering if some of the other CPPisms like this:
#if MIN_VERSION_base(4,17,0)
Word256 (W64# (or64# a3 b3)) (W64# (or64# a2 b2))
(W64# (or64# a1 b1)) (W64# (or64# a0 b0))
#else
Word256 (W64# (or# a3 b3)) (W64# (or# a2 b2))
(W64# (or# a1 b1)) (W64# (or# a0 b0))
#endif
could be lifted into the Compat module.
There was a problem hiding this comment.
This is all made more difficult because ghc-9.4 insists on being built with the Hadrian build system which complete breaks the way I used to manage installing multiple GHC versions.
There was a problem hiding this comment.
Oh, totally - I only did the Int128 to start since I didn't want to do a bunch of work if it wasn't to your tastes 😅 I'll port the rest over.
There was a problem hiding this comment.
also tbh ghcup does work very well, you may consider adopting it? i was pretty skeptical but it is legitimately very good
There was a problem hiding this comment.
Can ghcup produce a package I can install on several machines? Can I specify where it installs stuff like /usr/lib/ghc-X.Y ?
src/Data/WideWord/Int128.hs
Outdated
| #if MIN_VERSION_base(4,17,0) | ||
| #define ONE (wordToWord64# 1##) | ||
| #else | ||
| #define ONE (1##) | ||
| #endif |
There was a problem hiding this comment.
This is very silly! I cannot define:
one# :: Word64#
one# = wordToWord64# 1##
|
Thanks @parsonsmatt . I have a local fix for the CI failures so I will finish this off in the next couple of days. I will almost certainly pull all these changes into my own branch, raise a new PR and then close this one. |
|
Closed in favor of #72 (which is getting closer and closer). |
|
@parsonsmatt @spacekitteh New version on Hackage. |
|
Thank you 🙏🏻 🙌🏻 |
|
Hopefully the GHC people do not break it again too soon. |
|
No description provided.