Support ghc-9.10 #18
Support ghc-9.10 #18
Conversation
The `base` that ships with `ghc-9.10` (base version 4.20.*) exports the function `foldl'` so we need to hide its to avoid a name collision.
| @@ -158,7 +158,7 @@ import Data.Strict.HashMap.Autogen.Internal.List (isPermutationBy, unorderedComp | |||
| import Data.Semigroup (Semigroup (..), stimesIdempotentMonoid) | |||
| import GHC.Exts (Int (..), Int#, TYPE, (==#)) | |||
| import GHC.Stack (HasCallStack) | |||
| import Prelude hiding (filter, foldl, foldr, lookup, map, | |||
| import Prelude hiding (filter, foldl, foldl', foldr, lookup, map, | |||
| null, pred) | |||
There was a problem hiding this comment.
@infinity0 Is this the best way to do this? I looked around at the scripts but could not find a better way.
There was a problem hiding this comment.
No, this is not the best way. The best way is to
- ensure the upstream libraries already support GHC 9.10.
- update the versions inside
strict-containers/regen.shto the versions that support GHC 9.10, and then re-run it. ideally this Just Works and there's nothing else to do. - if this fails, then adjust our patches in
strict-containers/patches/, and retry (2) - if (3) is too hard to do manually, then
strict-containers/update-patches-from-git.shcan help automate part of this process, but you should understand exactly what it's doing as it may mess up your git repo if run incorrectly
I just checked however, and:
- (1) already fails because the latest released versions of
containers(0.7) andunordered-containers(0.2.20) don't yet support GHC 9.10, they don'timport hiding foldl'. containersmaster branch does support GHC 9.10, howeverunordered-containersdoesn't.- edit: also should wait for
vectorto release, so we can do Drop Vector patches #16, which presumably will also include GHC 9.10 support.
I suggest you wait until all these libraries have released versions that support GHC 9.10 before proceeding with this PR. At the very least, your patch here should be applied to the unordered-containers git repo before we proceed with changes in this library.
After that (i.e. contributing to unordered-containers, so that all 3 upstream git repos have master branches that support GHC 9.10), you could optionally experiment with changing the versions in regen.sh to master, but this may result in duplicate unnecessary work if the upstream libraries make further changes that invalidate our new adjusted patches. So that's why I suggest to wait until all libraries have released (tagged) versions that support GHC 9.10.
There was a problem hiding this comment.
Ok, i will not merge this PR as is, I will try the above.
There was a problem hiding this comment.
(1) already fails because the latest released versions of
containers(0.7) andunordered-containers(0.2.20) don't yet support GHC 9.10
Both packages do support GHC 9.10.
There was a problem hiding this comment.
@Bodigrim I checked the source code and they don't hiding foldl' like this PR was proposing, how could they support GHC 9.10 without that?
There was a problem hiding this comment.
@infinity0 I'm not sure what you are looking at. I just checked, say, https://hackage.haskell.org/package/containers-0.7/docs/src/Data.IntMap.Strict.Internal.html - it does hide foldl'.
There was a problem hiding this comment.
Ah right I was too hasty with a text search, used a regex that only looked on one line.
And unordered-containers has haskell-unordered-containers/unordered-containers@259dc9e which achieves the same without mentioning foldl' at all.
Ok in that case all that's needed is to wait for vector to release, the latest version 0.13.1.0 doesn't hide foldl' but its master branch does.
Then, some of the patches will have to be updated too.
vector @ 0.13.1.0 already invalidates the existing patches
diff --git a/strict-containers/regen.sh b/strict-containers/regen.sh
index a96bec1..25ed7d0 100755
--- a/strict-containers/regen.sh
+++ b/strict-containers/regen.sh
@@ -111,9 +111,9 @@ copy_test_and_rename() {
if [ -z "$CLEAN" ]; then
VERSIONS_CABAL=versions.cabal.in
rm -f $VERSIONS_CABAL
- ensure_checkout containers v0.6.6
- ensure_checkout unordered-containers v0.2.19.1
- ensure_checkout vector vector-0.13.0.0
+ ensure_checkout containers v0.7
+ ensure_checkout unordered-containers v0.2.20
+ ensure_checkout vector vector-0.13.1.0
cat $VERSIONS_CABAL | fixup_cabal versions ""
rm -f $VERSIONS_CABAL
else[..]
patching file src/Data/Strict/Vector/Autogen/Mutable.hs
Hunk #2 succeeded at 135 (offset 2 lines).
Hunk #3 succeeded at 144 (offset 2 lines).
patching file tests/Tests/Move.hs
patching file tests/Tests/Vector.hs
patching file tests/Tests/Vector/Boxed.hs
patching file tests/Tests/Vector/UnitTests.hs
patching file tests/Utilities.hs
Hunk #1 FAILED at 5.
containers @ master (> 0.7) invalidates the existing patches
diff --git a/strict-containers/regen.sh b/strict-containers/regen.sh
index a96bec1..25ed7d0 100755
--- a/strict-containers/regen.sh
+++ b/strict-containers/regen.sh
@@ -111,9 +111,9 @@ copy_test_and_rename() {
if [ -z "$CLEAN" ]; then
VERSIONS_CABAL=versions.cabal.in
rm -f $VERSIONS_CABAL
- ensure_checkout containers v0.6.6
- ensure_checkout unordered-containers v0.2.19.1
- ensure_checkout vector vector-0.13.0.0
+ ensure_checkout containers master
+ ensure_checkout unordered-containers master
+ ensure_checkout vector master
cat $VERSIONS_CABAL | fixup_cabal versions ""
rm -f $VERSIONS_CABAL
else[..]
patching file src/Data/Strict/Map/Autogen/Internal.hs
Hunk #1 succeeded at 467 (offset 9 lines).
patching file src/Data/Strict/IntMap/Autogen/Internal.hs
Hunk #1 FAILED at 359.
1 out of 1 hunk FAILED -- saving rejects to file src/Data/Strict/IntMap/Autogen/Internal.hs.rej
There was a problem hiding this comment.
Hackage has a version that could be used, but the scripts rely on the version in git and the version is git is on v0.6.8.
I have raised an issue: haskell/containers#1012
There was a problem hiding this comment.
Also haskell/vector#496
|
This is not yet ready. |
No description provided.