Support for 9.10 #4233
Support for 9.10 #4233
Conversation
This includes supports for all plugins, other than formatters and hlint. We need ghc-exactprint and retrie release before merging this.
wz1000
requested review from
michaelpj,
fendor,
pepeiborra,
konn,
santiweight,
joyfulmantis,
July541 and
ozkutuk
as code owners
|
If anyone wants to help, chasing up any of the allow-newers would be helpful (get upstream to release/make a revision). |
|
I've revised the |
| #if MIN_VERSION_ghc(9,9,0) | ||
| in L (noAnnSrcSpanDP dp <> l) e | ||
| newLine (L _ e) = | ||
| let dp = deltaPos 1 (defaultIndent + 1) {- TODO why is this +1 needed? -} |
There was a problem hiding this comment.
@alanz this is curious:
This code is responsible for adding typeclass method placeholders to typeclass instance declarations.
In ghc 9.10, in order to indent the method declarations by 2 spaces I effectively have to create DeltaPos like this:
dp :: DeltaPos
dp = DifferentLine 1 3where the previous code was using DifferentLine 1 2 (but also the previous code called makeDeltaAst on the whole module whereas I removed this call for GHC 9.10)
ditto in gadt plugin tests below where
delta = EpaDelta (DifferentLine 1 3) []
fixed all test failures.
Could this be an exactprint bug?
I have trouble reasoning about the expected behavior when you have
bunch of declarations with absolute positions and you insert some nodes with DeltaPos positions.
There was a problem hiding this comment.
There were changes in ghc-exactprint, to harmonise the treatment of delta values at the top level and inside the AST. So they are all coherent now, but I guess the column is now off by one, consistently for DifferentLine.
plugins/hls-eval-plugin/test/Main.hs
Outdated
| @@ -85,7 +85,8 @@ tests = | |||
| , goldenWithEval "Shows a kind with :kind" "T12" "hs" | |||
| , goldenWithEval "Reports an error for an incorrect type with :kind" "T13" "hs" | |||
| , goldenWithEval' "Returns a fully-instantiated type for :type" "T14" "hs" (if ghcVersion >= GHC98 then "ghc98.expected" else "expected") -- See https://gitlab.haskell.org/ghc/ghc/-/issues/24069 | |||
| , knownBrokenForGhcVersions [GHC92, GHC94, GHC96, GHC98] "type +v does not work anymore with 9.2" $ goldenWithEval "Returns an uninstantiated type for :type +v, admitting multiple whitespaces around arguments" "T15" "hs" | |||
| , knownBrokenForGhcVersions [GHC92, GHC94, GHC96, GHC98, GHC910] "type +v does not work anymore with 9.2" $ | |||
There was a problem hiding this comment.
It seems to me that we should just remove this testcase, since all supported GHCs now say:
:type +v 123
`:type +v' has gone; use `:type' instead
This includes supports for all plugins, other than formatters and hlint.
We need ghc-exactprint and retrie release before landing this.