Upstream Test Mapping

Perl5 Test Suite

The perl5 source tree (perl5-upstream/t/) contains thousands of tests organized by category:

perl5-upstream/t/
  base/     - Fundamental tests
  cmd/      - Control flow
  comp/     - Compilation
  io/       - I/O operations
  op/       - Operators
  re/       - Regular expressions
  run/      - Running perl
  uni/      - Unicode
  lib/      - Library/module tests
  ...

Mapping to PetaPerl Tests

PetaPerl’s test directories correspond to perl5-upstream categories, though the mapping is not 1:1:

Porting Process

When porting a test from perl5-upstream:

1. Convert to raw TAP: Remove use Test::More, use test.pl, and other test framework dependencies. Replace with inline ok sub and print "1..N\n" plan.

2. Suffix with -perl5.t: Indicates the test originated from perl5-upstream.

3. Add source comment: Document where the test came from:

   # Ported from perl5-upstream/t/op/arith.t
   

4. Verify under both runtimes: Run with t/TEST and confirm both perl5 and pperl produce identical output.

5. Handle unsupported features: If the test uses features pperl doesn’t support, either:

   • Skip those specific subtests with comments explaining why
   • Split the test to separate supported from unsupported portions

re_tests

The regex engine has a specialized test suite based on perl5’s re_tests format. PetaPerl passes 1959/1972 tests (99.3% of perl5-passing tests). The 13 remaining failures are in:

• Self-referential captures (3 tests)
• local in code blocks (2 tests)
• Multi-character case folding (2 tests)
• Branch reset references (5 tests)
• String interpolation edge case (1 test)

Coverage Philosophy

PetaPerl does not aim to run the entire perl5 test suite verbatim. Instead:

• Port selectively: Focus on tests that cover features pperl implements
• Write new tests: For pperl-specific features (JIT, parallelization, daemon mode)
• Conformance over coverage: A smaller set of passing tests is better than a large set with many expected failures