Make it build with ghc-9.6 #5286

erikd · 2023-05-03T01:16:15Z

No description provided.

cabal.project

michaelpj · 2023-05-03T08:35:01Z

doc/read-the-docs-site/plutus-doc.cabal

@@ -70,7 +70,7 @@ executable doc-doctests
    , base               >=4.9      && <5
    , bytestring
    , containers
-    , flat               <0.5
+    , flat               ^>=0.6


We need to be careful with this, we have to check carefully that upstream hasn't broken anything. Is this change necessary for 9.6?

The old version does not compile with 9.6 (~~probably due to base bounds~~ nope, much worse than that).

If we can get away with allow-newering the base bounds then I think I'd mildly prefer this, since this is a sensitive dependency and this will change the bound for all GHC versions atm.

Would it not make sense to do one or both of the following?

Write tests for flat and submit them upstream.

Write tests for flat and keep them in the Plutus repo.

I tried removing the version bump for flat and added allow-newer: flat:*. The result was:

src/Flat/Decoder/Strict.hs:240:19: error: [GHC-31891] • Illegal term-level use of the type constructor or class ‘TA.Array’ • imported qualified from ‘Data.Text.Array’ at src/Flat/Decoder/Strict.hs:45:1-53 • Perhaps use variable ‘array’ (line 239) • In the first argument of ‘T.Text’, namely ‘(TA.Array array)’ In the first argument of ‘return’, namely ‘(T.Text (TA.Array array) 0 (lengthInBytes `div` 2))’ In a stmt of a 'do' block: return (T.Text (TA.Array array) 0 (lengthInBytes `div` 2)) | 240 | return (T.Text (TA.Array array) 0 (lengthInBytes `div` 2)) | ^^^^^^^^

and

src/Flat/Encoder/Prim.hs:243:25: error: [GHC-76037] Not in scope: data constructor ‘TA.Array’ NB: the module ‘Data.Text.Array’ does not export ‘Array’. | 243 | eUTF16F_ !(TI.Text (TA.Array array) w16Off w16Len) s = | ^^^^^^^^

It supposedly is possible to constrain text < 2.0 but that may cause issues further up the stack.

We have tests. Nonetheless I'm cautious :)

@bezirg could you look into upgrading to flat >= 0.6 on master and check that the script dump test continues to pass?

Ok I am on it.

The test infrastructure this needs got broken a while ago so it might be a little while until we can test this. I am tentatively in favour of merging since I think it's very likely not broken.

The tests are finished locally.
Both plutus-master and plutus-master+flat0.6.patch pass the current mainnet script tests

michaelpj · 2023-05-05T11:20:03Z

Now it appears to fail to build with 8.10 and fail to even plan with 9.2

erikd · 2023-05-08T01:49:06Z

@michaelpj I rebased this against master, then made sure it (other than plutus-tx-plugin) builds here with ghc-8.10, ghc-9.2 and ghc-9.6.

I think the failure to create a build plan was due to the line:

    build-depends: ghc >=9.6 && <9.8

I have been adjusting those bounds for each compiler.

Would be happy to hear any solutions to that issue.

michaelpj · 2023-05-09T08:36:19Z

I think we discussed this a bit on Slack, but:

plutus-tx-plugin depends on ghc, and is finicky.
So we're only supporting a single version of GHC and it's non-trivial to switch which one.
That version is 9.2, so the bounds on ghc should remain the same.
We prevent plutus-tx-plugin and its downstream packages from building on other GHCs with buildable: False conditionals. That prevents cabal from trying to put it in the plan for other GHC versions and getting confused.
Those buildable conditionals need to be updated to exclude GHCs newer than 9.2.
That should leave us with all those packages building only on 9.2, which is what we want for now.

erikd · 2023-05-09T08:46:48Z

Ah, this is what I need:

diff --git a/plutus-tx-plugin/plutus-tx-plugin.cabal b/plutus-tx-plugin/plutus-tx-plugin.cabal
index a40d9ccde..6635d44b6 100644
--- a/plutus-tx-plugin/plutus-tx-plugin.cabal
+++ b/plutus-tx-plugin/plutus-tx-plugin.cabal
@@ -49,7 +49,7 @@ flag use-ghc-stub
 library
   import:          lang
 
-  if impl(ghc <9.0)
+  if impl(ghc < 9.2) || impl(ghc >= 9.3)
     buildable: False

and in a number of other places.

plutus-core/index-envs/src/Data/RandomAccessList/SkewBinarySlab.hs

michaelpj

Looks pretty much there. We need to update the GHC versions section of CONTRIBUTING, but I can do that when I add 9.6 CI.

michaelpj · 2023-05-10T11:34:49Z

cabal.project

+    , th-compat:*
+
+  constraints:
+    , microlens >= 0.2.0.0


Please add a comment explaining these (I would expect us cabal to pick the compatible versions of these libraries automatically) and when we can remove them.

michaelpj · 2023-05-10T11:35:26Z

cabal.project

+    , dependent-map:*
+    , hedgehog:*
+    , protolude:*
+    , th-compat:*


I'm assuming we have upstream issues for everything that's jailbroken by this?

michaelpj · 2023-05-10T11:36:03Z

doc/read-the-docs-site/plutus-doc.cabal

@@ -70,7 +70,7 @@ executable doc-doctests
    , base               >=4.9      && <5
    , bytestring
    , containers
-    , flat               <0.5
+    , flat               ^>=0.6


The test infrastructure this needs got broken a while ago so it might be a little while until we can test this. I am tentatively in favour of merging since I think it's very likely not broken.

michaelpj · 2023-05-10T11:36:40Z

plutus-core/plutus-core/src/PlutusCore/Core/Type.hs

-    deriving anyclass (NFData)
+    deriving stock (Functor, Generic)
+
+deriving stock instance (Show tyname, Show name, GShow uni, Everywhere uni Show, Show fun, Show ann, Closed uni)


Yeah, it really sucks. But perhaps we could introduce some aliases in order not to write this whole thing our every time. Maybe some generic one that works for every constraint (we'll only need to special-case things like GShow). Not sure if it's worth it, though.

plutus-core/plutus-core/stdlib/PlutusCore/StdLib/Meta/Data/Tuple.hs

plutus-core/untyped-plutus-core/test/Evaluation/FreeVars.hs

prettyprinter-configurable/prettyprinter-configurable.cabal

michaelpj · 2023-05-11T14:58:47Z

I pushed a commit attempting to make the 9.6 CI work, and also merged master to get a haskell.nix update.

michaelpj · 2023-05-11T15:59:44Z

Sigh: input-output-hk/haskell.nix#1943

michaelpj · 2023-05-11T16:06:28Z

Okay, it seems to find a build plan now, let's see how far it gets.

michaelpj · 2023-05-12T09:56:18Z

src/Codec/CBOR/Magic.hs:260:20: error: [GHC-83865]
    • Couldn't match expected type ‘Word64#’ with actual type ‘Word#’
    • In the first argument of ‘W64#’, namely ‘(toWord w#)’
      In the expression: W64# (toWord w#)
      In an equation for ‘w64’: w64 w# = W64# (toWord w#)
    |
260 |     w64 w# = W64# (toWord w#)
    |                    ^^^^^^^^^

maybe well-typed/cborg#311 ?

Looks like we need to poke our WT colleagues or something.

michaelpj · 2023-05-12T09:57:54Z

Non-darwin fails because of cardano-crypto-class using -Werror :| I think they fixed that on master but we need a release.


src/Cardano/Crypto/Hash/Short.hs:23:35: error: [GHC-58520] [-Wtype-equality-requires-operators, Werror=type-equality-requires-operators]
    The use of ‘~’ without TypeOperators
    will become an error in a future GHC release.
    Suggested fix: Perhaps you intended to use TypeOperators
   |
23 | instance (KnownNat n, CmpNat n 33 ~ 'LT) => HashAlgorithm (Blake2bPrefix n) where
   |                                   ^

michaelpj · 2023-05-15T14:22:59Z

There was a cardano-crypto-class release that should have 9.6 support, bumped our lower bound
I've enabled the 9.6 CI on linux only for now, to get around the cborg issue. That'll be a problem in the long run but apparently WT are aware of it and we can probably just ignore it for now.
I simplified the allow-newers a lot
I added a s-r-p for inline-r, since it actually needs some code changes to build :(

I still seem to be getting some warnings locally, I'll fix those next, and then I think this might be ready.

michaelpj · 2023-05-16T09:51:16Z

An actual test failure! And an interesting one.

michaelpj · 2023-05-16T10:33:18Z

plutus-core/plutus-core/src/PlutusCore/Evaluation/Machine/MachineParameters.hs

-    deriving anyclass (NFData, NoThunks)
+    deriving anyclass (NFData)
+
+-- For some reason the generic instance gives incorrect nothunk errors,


FYI @effectfully

angerman · 2023-05-17T01:23:36Z

force-pushed to re-excite ci.

Co-Authored-By: effectfully <effectfully@gmail.com>

Add GHC 9.6 to our CI.

Co-Authored-By: effectfully <effectfully@gmail.com>

erikd · 2023-05-18T07:16:54Z

I think all that is missing is #5309

erikd force-pushed the erikd/ghc-9.6 branch 5 times, most recently from 39846a1 to 5c2dab1 Compare May 3, 2023 03:51

michaelpj reviewed May 3, 2023

View reviewed changes

erikd force-pushed the erikd/ghc-9.6 branch 2 times, most recently from c40ea93 to 7c27753 Compare May 5, 2023 01:16

erikd mentioned this pull request May 5, 2023

Upgrade plutus to build with ghc-9.6 #5301

Closed

angerman linked an issue May 5, 2023 that may be closed by this pull request

Upgrade plutus to build with ghc-9.6 #5301

Closed

erikd force-pushed the erikd/ghc-9.6 branch from 7c27753 to 6a47300 Compare May 7, 2023 23:25

erikd force-pushed the erikd/ghc-9.6 branch from 6a47300 to 90e1088 Compare May 8, 2023 01:58

erikd mentioned this pull request May 9, 2023

Look into upgrading to flat ^>= 0.6 on master and check that the script dump test continues to pass #5309

Closed

erikd force-pushed the erikd/ghc-9.6 branch from 90e1088 to 2d45d17 Compare May 9, 2023 09:21

michaelpj reviewed May 9, 2023

View reviewed changes

plutus-core/index-envs/src/Data/RandomAccessList/SkewBinarySlab.hs Outdated Show resolved Hide resolved

erikd force-pushed the erikd/ghc-9.6 branch 6 times, most recently from b30089f to fd2e4e6 Compare May 10, 2023 07:06

michaelpj reviewed May 10, 2023

View reviewed changes

michaelpj added the No Changelog Required Add this to skip the Changelog Check label May 11, 2023

michaelpj reviewed May 16, 2023

View reviewed changes

angerman force-pushed the erikd/ghc-9.6 branch from b82157a to f2c5942 Compare May 17, 2023 01:23

erikd force-pushed the erikd/ghc-9.6 branch 3 times, most recently from a7bcba9 to d3d6088 Compare May 17, 2023 07:44

erikd changed the title ~~WIP Make it build with ghc-9.6~~ Make it build with ghc-9.6 May 17, 2023

erikd force-pushed the erikd/ghc-9.6 branch from d3d6088 to 8e9d118 Compare May 17, 2023 22:03

angerman force-pushed the erikd/ghc-9.6 branch from 8e9d118 to cfe3946 Compare May 18, 2023 01:14

erikd and others added 13 commits May 18, 2023 13:58

WIP Make it build with ghc-9.6

2f00e81

Co-Authored-By: effectfully <effectfully@gmail.com>

9.6 CI

f2845fb

Add GHC 9.6 to our CI.

Bump cabal-install to support newer GHCs

b94faeb

Relax bound for inline-r

5550cbe

Un-conditionalize s-r-p, setup sha256map

4a24b5c

Bump cardano-crypto-class bound

d5c920d

Only build 9.6 on linux for now

8598a9a

Simplify allow-newers

471aebe

Fix warnings

4844665

more

fdd8f34

Don't build the 9.6 shell on aarch64-darwin either

d11e12f

Fix nothunks for MachineParameters

42edf8b

Make it build with ghc-9.6

4ae5473

Co-Authored-By: effectfully <effectfully@gmail.com>

angerman force-pushed the erikd/ghc-9.6 branch from cfe3946 to 4ae5473 Compare May 18, 2023 05:58

michaelpj merged commit ab4d2cc into master May 18, 2023

michaelpj deleted the erikd/ghc-9.6 branch May 18, 2023 08:59

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make it build with ghc-9.6 #5286

Make it build with ghc-9.6 #5286

erikd commented May 3, 2023

michaelpj May 3, 2023

erikd May 3, 2023 •

edited

Loading

michaelpj May 3, 2023

erikd May 3, 2023 •

edited

Loading

erikd May 3, 2023 •

edited

Loading

michaelpj May 4, 2023

michaelpj May 4, 2023

bezirg May 9, 2023

michaelpj May 10, 2023

bezirg May 11, 2023

michaelpj commented May 5, 2023

erikd commented May 8, 2023 •

edited

Loading

michaelpj commented May 9, 2023

erikd commented May 9, 2023 •

edited

Loading

michaelpj left a comment

michaelpj May 10, 2023

michaelpj May 10, 2023

michaelpj May 10, 2023

michaelpj May 10, 2023

effectfully May 10, 2023

michaelpj commented May 11, 2023

michaelpj commented May 11, 2023

michaelpj commented May 11, 2023

michaelpj commented May 12, 2023

michaelpj commented May 12, 2023

michaelpj commented May 15, 2023

michaelpj commented May 16, 2023

michaelpj May 16, 2023

angerman commented May 17, 2023

erikd commented May 18, 2023

Make it build with ghc-9.6 #5286

Make it build with ghc-9.6 #5286

Conversation

erikd commented May 3, 2023

Choose a reason for hiding this comment

erikd May 3, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

erikd May 3, 2023 • edited Loading

Choose a reason for hiding this comment

erikd May 3, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

michaelpj commented May 5, 2023

erikd commented May 8, 2023 • edited Loading

michaelpj commented May 9, 2023

erikd commented May 9, 2023 • edited Loading

michaelpj left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

michaelpj commented May 11, 2023

michaelpj commented May 11, 2023

michaelpj commented May 11, 2023

michaelpj commented May 12, 2023

michaelpj commented May 12, 2023

michaelpj commented May 15, 2023

michaelpj commented May 16, 2023

Choose a reason for hiding this comment

angerman commented May 17, 2023

erikd commented May 18, 2023

erikd May 3, 2023 •

edited

Loading

erikd May 3, 2023 •

edited

Loading

erikd May 3, 2023 •

edited

Loading

erikd commented May 8, 2023 •

edited

Loading

erikd commented May 9, 2023 •

edited

Loading