Repository navigation
Function .name and Function.prototype.toString leak compiler internals (Made$4, __anon_class_8, [native code]) #9413
Description
function outer(){ function inner(){} return inner.name } // node "inner" perry ""
// two classes both named Made, in different scopes: node "Made" perry "Made$4"
new (class extends Error{})("m").constructor.name // node "" perry "__anon_class_8"
String(class Klass{}) // node "class Klass {}" perry "function Klass() { [native code] }"
String(({ m(){ return 1 } }).m) // node "m() { return 1; }" perry "function () { [native code] }"Top-level names and NamedEvaluation are correct; nested declarations and duplicate names are not, and the mangled forms (Made$4, __anon_class_8) leak the compiler's disambiguation into user-visible strings.
71 constructor.name uses in the cc bundle. Libraries routinely dispatch on constructor.name, and Function.prototype.toString is used for source reflection — both get wrong answers here rather than approximations.
Found by a differential stress-test of claude-code under perry.
Activity
fix: static this is the constructor; class name/toString/inspect repo…
proggeramlug commentedon Sep 2, 2026
Fixed by PR #9465, landed via #9472:
.namereports the source name (not theMade$Ndisambiguation key),toString()returns real class source text via the new class-side source registry, and inspect prints[class Name]. Fixturestest_class_name_and_source_9413.ts+.ctsarm. Verified on main0a1c137d90with a fresh build: fixture byte-identical tonode --experimental-strip-types(stdout, stderr, and exit code).Follow-ups filed: #9466 (the underlying ClassId aliasing is WRONG CODE, still live), #9468 (method source text, accessor names, .cts NamedEvaluation).
test(9466): gap fixture — same-name classes at different lexical depths
fix(hir): disambiguate same-name classes per SCOPE, not per name (#9466)