--check causes <<loop>> for basic expressions
#1016
Comments
|
Thanks for the report. The initial hunch is that I kind of went wild along here & swapped cons into snocs in some places, which looking back probably shouldn't, but I had some ideas to find optimizations in mind, thou not finished that. Those changes may cause infinite loops. But also there are many other things that may cause it. A bunch of reports need bisecting. Thankfully commits are good for that & https://github.com/Anton-Latukha/git-bisect-master-pr. Would look into it when would make a release from The code that processes the Lines 119 to 128 in d3d4356 |
|
Thank you @Anton-Latukha, I will look into it |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
|
Currently seems it was "always" present. (at least, an old bug). It is present in Currently waiting for when Hackage would update with the new snapshot. |
|
Well, it goes beyond |
|
Before |
|
It is pre Before, |
|
Syndrome is: Works great: Referencing set - nope: |
|
@Anton-Latukha thank you for looking into this. I did a bit of debugging. The The following diff against 4321e7d diff --git a/main/Main.hs b/main/Main.hs
index 378fa468..00f487bd 100644
--- a/main/Main.hs
+++ b/main/Main.hs
@@ -32,6 +32,7 @@ import Prettyprinter hiding ( list )
import Prettyprinter.Render.Text ( renderIO )
import qualified Repl
import Nix.Eval
+import qualified Debug.Trace
main :: IO ()
main =
@@ -124,7 +125,7 @@ main' opts@Options{..} = runWithBasicEffectsIO opts execContentsFilesOrRepl
(liftIO . putStrLn . (<>) "Type of expression: " .
ppShow . maybeToMonoid . Map.lookup @VarName @[Scheme] "it" . coerce
)
- $ HM.inferTop mempty $ curry one "it" $ stripAnnotation expr'
+ $ HM.inferTop mempty $ (\expr'' -> ("inferTop: " <> show expr) `Debug.Trace.trace` expr'') $ curry one "it" $ stripAnnotation expr'
-- liftIO $ putStrLn $ runST $
-- runLintM opts . renderSymbolic =<< lint opts expr
diff --git a/src/Nix/Scope.hs b/src/Nix/Scope.hs
index d9de14a3..70248725 100644
--- a/src/Nix/Scope.hs
+++ b/src/Nix/Scope.hs
@@ -10,6 +10,7 @@ import qualified Data.HashMap.Lazy as M
import qualified Text.Show
import Lens.Family2
import Nix.Expr.Types
+import qualified Debug.Trace
-- 2021-07-19: NOTE: Scopes can gain from sequentiality, HashMap (aka AttrSet) may not be proper to it.
newtype Scope a = Scope (AttrSet a)
@@ -110,6 +111,11 @@ lookupVarReader
-> m (Maybe a)
lookupVarReader k =
do
+ Debug.Trace.traceM $ "lookupVarReader: " <> show k
+ lexicalScopes' <- asks $ lexicalScopes @m . view hasLens
+ Debug.Trace.traceM $ "lexicalScopes length: " <> show (length lexicalScopes')
+ Debug.Trace.traceM $ "lexicalScopes: " <> show (lexicalScopes' :: [Scope a])
+
mres <- asks $ scopeLookup k . lexicalScopes @m . view hasLens
maybewill produce this output: So first a pass runs prior to |
|
Looked into the So far not found the cause, but the process definitely approaches closer. If to use only the dynamic scope search - the process terminates: cabal run hnix -- --check --expr '
let {
x = "foo";
y = "bar";
body = x + y;
}
'
...
Type of expression: [ Forall [ TV "a" ] (TVar (TV "a")) ]
rec {
x = "foo";
y = "bar";
body = x + y;
}.bodySoon would ship the simplified version of the code & it would be easier to look further. |
I get a
<<loop>>when trying to evaluate many simple Nix expressions using--check, e.g. data/nix/tests/lang/eval-okay-let.nix:This applies to all of the following expressions:
Is
--checksupposed to work?I'm on
master, building hnix usingnix-buildusing the "Full debug info" options on macOS.The text was updated successfully, but these errors were encountered: