Encode canonical charset names differ from Perl, breaking IO::HTML

从 PerlOnJava 的 Encode 实现及其现有兼容性测试入手，然后在 JVM 和解释器后端上重现 canonical-name 失败。为 ISO-8859-15 和 cp1252 映射添加针对性的回归测试覆盖，并验证 IO::HTML 1.004 测试套件在两个后端上都能通过。

Summary

IO::HTML 1.004 passes its complete upstream test suite under system Perl but fails 23 of 165 tests under PerlOnJava because Encode::find_encoding(...)->name returns Java-style charset names instead of Perl's canonical names.

Reproduction

CPAN random tester run: 20260918-141920-96054

Distribution: IO::HTML 1.004

System Perl 5.42.2:

All tests successful.
Files=5, Tests=165
Result: PASS

PerlOnJava JVM backend:

Files=5, Tests=165
Result: FAIL
Failed 2/5 test programs. 23/165 subtests failed.

The interpreter backend reproduces the same 23 failures.

Failing tests

• t/10-find.t: 14 failures
• t/20-open.t: 9 failures
• t/00-all_prereqs.t, t/00-load.t, and t/30-outfile.t: pass

The failures are all charset-name mismatches:

got:      ISO-8859-15
expected: iso-8859-15

got:      windows-1252
expected: cp1252

The affected inputs include HTML declarations for ISO-8859-15 and ISO-8859-1, including declarations near the 1024-byte scan boundary.

Likely cause

IO::HTML calls Encode::find_encoding($charset) and then uses the returned object's name method. Perl returns Encode's canonical names (iso-8859-15 and cp1252), while PerlOnJava returns Java canonical charset names (ISO-8859-15 and windows-1252).

The PerlOnJava Encode implementation already maps US-ASCII to ascii and UTF-8 to utf-8-strict, but otherwise returns the Java charset name unchanged. The canonical-name mapping needs to cover the relevant Perl aliases and should preserve the expected behavior for both backends.

Expected behavior

For equivalent Encode::find_encoding calls, PerlOnJava should return an Encode::Encoding object whose name matches Perl's canonical name, including at least:

• ISO-8859-15 -> iso-8859-15
• windows-1252 / ISO-8859-1 -> cp1252

Please add focused regression coverage for the canonical names and verify IO::HTML 1.004 on both the JVM and interpreter backends after the fix.

Related project context

The project module-compatibility notes already identify encoding-name case/alias normalization as an outstanding Encode compatibility gap affecting IO::HTML. This report provides a complete upstream reproduction and confirms the defect on both PerlOnJava execution backends.