Skip to content

Intl: canonicalize physical string storage before locale diagnostics #829

Description

@chrisbbreuer

Parent #473; related #475 and discovered during #825 validation. Exact pushed baseline 8e1e441 aborts in the ReleaseSafe test262 Intl subtree: StringCell reports a truncated WTF-8 sequence while constructing RangeError text for an invalid locale. The captured bytes end in the Latin-1 byte df for an en-ß input. canonicalizeLocaleList currently passes v.asStr() directly for a string input, while canonicalizeLocaleTagForIntl interpolates those bytes into an engine string. The physical Latin-1 representation must be converted through the existing StringData/WTF-8 boundary, not sanitized repaired text substituted or the assertion disabled. Preserve the original invalid locale and RangeError, add Latin-1/UTF-16/WTF-8 and lone-surrogate cases, validate individual and list inputs and supportedLocalesOf/constructor entry points, then rerun the full scored Intl corpus. The complete baseline abort transcript and stack sample are retained locally; no completed Intl denominator or regression claim is made from that aborted run.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions