Unsupported specialization error related to Warnings

[GregoryTravis]

I am getting this error in my Map_Error WIP:

    - [FAILED] an error thrown inside map with on_problems=Problem_Behavior.Report_Warning should be attached as a warning [52ms]
        Reason: An unexpected panic was thrown: (Unimplemented.Error Unsupported specialization: Unexpected values provided for InvokeMethodNodeGen@11675922: [org.enso.interpreter.runtime.state.State@3e0d722f, UnresolvedSymbol<is_a>, WithWarnings{11 has 0 warnings}, [Ljava.lang.Object;@435a2e1f], [State,UnresolvedSymbol,WithWarnings,Object[]])
        at <enso> Array_Like_Helpers.vector_from_function(Internal)
        at <enso> Array_Like_Helpers.vector_from_function(/Users/gmt/dev/enso/enso/built-distribution/enso-engine-0.0.0-dev-macos-aarch64/enso-0.0.0-dev/lib/Standard/Base/0.0.0-dev/src/Internal/Array_Like_Helpers.enso:87:15-68)
        at <enso> Vector.map_with_index(/Users/gmt/dev/enso/enso/built-distribution/enso-engine-0.0.0-dev-macos-aarch64/enso-0.0.0-dev/lib/Standard/Base/0.0.0-dev/src/Data/Vector.enso:704:9-100)

Ocurring in this test.

At this particular revision I have a bug in which Array.hasWarnings() returns true, but Array.getWarnings() returns an empty list. If I add logic to not add the empty list of warnings to the value, the problem goes away.

[JaroslavTulach]

What's the revision where the problem appeared? The latest revision is ee92e19 - is it still reproducible there?

[GregoryTravis]

@JaroslavTulach Yes precisely that revision.

[JaroslavTulach]

Following @GregoryTravis steps I can see there is an instance of WithWarnings wrapping value 11 and having no warnings. The question for @hubertp (as the author of current warnings implementation) is: shall we be able to construct WithWarnings instance that doesn't have any warnings? The code seems to be ready for that - the hasWarnings checks for size() > 0 - however our specializations don't seem to handle this correctly.

I can add assert warnings.size() > 0 into the constructor and then I get:

Execution finished with an error: java.lang.AssertionError
        at <java> org.enso.runtime/org.enso.interpreter.runtime.error.WithWarnings.wrap(WithWarnings.java:86)
        at <java> org.enso.runtime/org.enso.interpreter.runtime.data.vector.Array.readArrayElement(Array.java:110)
        at <java> org.enso.runtime/org.enso.interpreter.runtime.data.vector.ArrayGen$InteropLibraryExports$Cached.readArrayElementNode_AndSpecialize(ArrayGen.java:150)
        at <java> org.enso.runtime/org.enso.interpreter.runtime.data.vector.ArrayGen$InteropLibraryExports$Cached.readArrayElement(ArrayGen.java:131)
        at <java> org.graalvm.truffle/com.oracle.truffle.api.interop.InteropLibrary$Asserts.readArrayElement(InteropLibrary.java:3996)
        at <java> org.graalvm.truffle/com.oracle.truffle.api.interop.InteropLibraryGen$CachedDispatch.readArrayElement(InteropLibraryGen.java:8819)
        at <java> org.enso.runtime/org.enso.interpreter.runtime.data.vector.Vector$Generic.readArrayElement(Vector.java:228)
        at <java> org.enso.runtime/org.enso.interpreter.runtime.data.vector.ArrayLikeAtNode.vectorAt(ArrayLikeAtNode.java:63)
        at <java> org.enso.runtime/org.enso.interpreter.runtime.data.vector.ArrayLikeAtNodeGen.executeAndSpecialize(ArrayLikeAtNodeGen.java:136)
        at <java> org.enso.runtime/org.enso.interpreter.runtime.data.vector.ArrayLikeAtNodeGen.executeAt(ArrayLikeAtNodeGen.java:97)
        at <java> org.enso.runtime/org.enso.interpreter.node.expression.builtin.immutable.AtVectorNode.execute(AtVectorNode.java:23)

Or we can unwrap a value when there are no warnings. But returning WithWarnings which has no warnings just confuses the specializations. Opinions, @hubertp?

[hubertp]

The question for @hubertp (as the author of current warnings implementation) is: shall we be able to construct WithWarnings instance that doesn't have any warnings? The code seems to be ready for that - the hasWarnings checks for size() > 0 - however our specializations don't seem to handle this correctly.

I can add assert warnings.size() > 0 into the constructor and then I get:

Or we can unwrap a value when there are no warnings. But returning WithWarnings which has no warnings just confuses the specializations. Opinions, @hubertp?

If I did, then probably unconsciously. I don't think we should have WithWarnings with no warnings inside them.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2023-12-18):

Progress: - UnsupportedSpecialization is duplicate: #8551

• PR: Avoid no WithWarnings warnings #8583
• review return types: Better context info in Type_Error raised from return type checks #8566 (review)
• Truffle GIL request discussion: Let languages control multi threaded access ... is not allowed for language(s) js oracle/graal#8033 (comment)
• review duplicated self: Syntax error results in silent error #8560 (comment)
• review "simple" httpbin mock: Add tests for Enso Cloud auth + simple API mock for Enso_User #8511 (comment)
• discussing autoscoping arguments
• pair programming with Dmitry
• discussing "autoscoping"
• meetings, demos, standups It should be finished by 2023-12-20.

Next Day: Finish UnsupportedSpecialization

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2023-12-19):

Progress: - PR integrated: #8583

• discussing with Radek: https://docs.google.com/document/d/1FfFOhXQ90ygucWtqJqieZEB4lk-vYILLezjyY1EKFSc/edit
• analyzing DataflowErrors - Try to improve performance of Dataflow Errors #8137 It should be finished by 2023-12-20.

Next Day: Create a PR for DataflowErrors