[average-web-dev]

On 1.9.2, reading stack of an error whose prototype was replaced throws a
Java ClassCastException, which scripts cannot catch:

function E() {}
E.prototype = Object.create(Error.prototype);
var e = new Error("m");
Object.setPrototypeOf(e, E.prototype);
e.stack; // java.lang.ClassCastException: NativeObject cannot be cast to NativeError

This is exactly what transpiled class MyError extends Error {} does (Babel's
wrapNativeSuper), so reading .stack of any transpiled custom error fails.

Cause and fix

NativeError.getStackDelegated() casts the error's prototype to NativeError
to look up the stack trace settings (stackTraceLimit, prepareStackTrace).
On master this cast was changed to ScriptableObject in #2295 ("Convert
NativeError to descriptor"), but that PR can't be backported on its own.

This PR only checks whether the prototype is a ScriptableObject and otherwise
uses the default settings. This also covers a null prototype and a Proxy as
prototype. A null prototype still throws a NullPointerException on
master; I can open a separate PR for that.

Tests

New test ErrorPropertiesTest.stackWithChangedPrototype for a plain object,
null and a Proxy as prototype, in all modes. It fails without the fix. The
full test suite passes, except NativeDateTest.toTimeStringEST, which also
fails on an unmodified rhino_1_9_backports on my machine (time zone data).
test262.properties doesn't change.

A note on how this was developed

I developed this PR with the help of an AI coding agent (Claude Code), and I
reviewed, tested and take responsibility for all of the changes. If the project
has concerns about or rules for AI-assisted contributions, please let me know.

I'm dyslexic, so I usually let AI help me write or proofread text like this
description. If anything reads unclear, please ask and I'll clarify.

🤖 Generated with Claude Code

[gbrail]

Here, are you sure that this is a 1.9 bug or one in the master branch?

[average-web-dev]

Partly both. The ClassCastException from the description is 1.9 only: on master, #2295 changed the cast to ScriptableObject. But with a null prototype, master still throws a Java NullPointerException that a script's try/catch can't catch:

var e = new Error("m");
Object.setPrototypeOf(e, null);
e.stack; // java.lang.NullPointerException in NativeError.getStackDelegated()

I'll open a PR for master with the same fix and tests first, and then update this PR to be a backport of it.

[average-web-dev]

I opened #2550 with the fix for master. This PR is its backport: the same
fix (a plain instanceof check and cast instead of pattern matching, since 1.9
targets Java 11) and the same test. On 1.9 the test also covers the
ClassCastException case from the description.