Changes for version 0.45 - 2026-08-10
- Enhancements
- Added --version option to bin/extract_schemas
- Changed close_stdin default from off to on
- Arrayref testing - add an undef entry in the middle of the array
- Hashref testing - added undef in a nested hash
- Add t/path.t: explicit CFG path-coverage tests for Model::Method, BenchmarkGenerator, Analyzer::Complexity, Analyzer::SideEffect, and Planner::Isolation. Each subtest exercises one uniquely identifiable branch not covered by the module-specific unit tests. Paths added: resolve_return_type() signal branches (legacy_type×4 including the value='object'/'self'/other/undef sub-paths, context_aware, error_pattern) and tie-break cases (all-zero→constant, object=property tie→object); resolve_confidence() at all four threshold boundary points (19→low, 20→medium, 39→medium, 40→high); add_evidence() weight=0 stored as 0 not default 1; _representative_value() all 15 branches (float/integer with no/min-only/max-only/both constraints, boolean/arrayref/hashref/unknown types, undef spec); _quote_value() all 4 paths; _build_call() all 5 call-syntax combinations; generate() missing-key croaks, has_new+builtin skips constructor, empty new_spec hashref→no-arg constructor, empty transforms→default variant; Complexity::analyze() given/when branch tokens, keyword inside dquoted strings stripped, escaped dquote handled, unmatched brace depth clamp, early_returns boundaries at 1 and 2 returns; SideEffect::analyze() read/write IO keywords, self+globals→impure purity path, false-positive guards for $self->{field}= inside strings and comments; Planner::Isolation::plan() scalar-falsy time/network values omit flags, hashref-truthy env propagates. Dead code flagged: Model::Method::resolve_return_type() line 552 — the `|| 'unknown'` fallback is unreachable because %score is always initialised with 3 keys so $winner is always a non-empty string.
- Bug fixes
- Fix SchemaExtractor::_detect_accessor_methods injecting the property name as a spurious extra input parameter when a getter/setter's argument is named differently from the stored property (e.g. sub logdir { my $dir = shift; ... $self->{logdir} = $dir } generated both 'dir' and 'logdir' as position-0 parameters). The position assignment now targets the existing input key rather than creating a new key named after the property.
- Fix t/type_params.t dying with "Your vendor has not defined BSD::Resource macro RLIMIT_AS" on platforms where BSD::Resource is installed but does not export RLIMIT_AS (e.g. some BSDs and macOS builds). The RLIMIT_AS() call is now wrapped in eval; if it croaks, the test is skipped with an explanatory message instead of aborting. Also fix "Bad plan. You planned 0 tests but ran 1" on the same platform: the BEGIN { use_ok(...) } block fired test #1 before the plan(skip_all => ...) call declared 0 tests. Replaced with a plain use so no test runs before the skip checks.
- Fix _compile_signature_isolated() calling decode_json() on empty or non-JSON output when the child subprocess is killed by the kernel OOM killer (SIGKILL) before it can write to stdout or stderr: the caller now guards for empty stdout and wraps decode_json() in eval, so a silent child crash degrades gracefully (return undef → fall back to heuristics) rather than croaking with "malformed JSON string" and propagating through the Test::Builder subtest boundary to kill the test process with exit 255.
- Fix test-generator-index coverage trend chart always ending one commit behind the current run: the coverage snapshot is written to the archive AFTER the HTML is generated, so the current commit's point was never in the trend data — the last visible point was the previous commit's (which could show anomalously low coverage, e.g. a partial Devel::Cover run, making the trend appear to crash at the right-hand edge). The current run is now appended as the final trend point using the same arithmetic mean of per-file percentages as the archive-based points, guarded by a SHA check to avoid duplication if the snapshot was somehow already committed before HTML generation.
- Fix test-generator-index delta column showing the wrong value for blib/-instrumented modules (the common case under 'cover -test'). %deltas was keyed by the normalised lib/ path ($display_file) but the lookup used the raw Devel::Cover blib/ key ($file), so the exists() check always failed and the code silently fell through to a sparkline-based fallback. The fallback computes delta from the last two sparkline points; when the most-recent archive entry contained a 0.0 (from an older undef-percentage snapshot), the fallback computed 0.0 - 95.6 = -95.6%, displaying "▼ 95.6%" — the same number as the actual coverage. Fixed by: (1) using $display_file (already normalised to lib/) for both the %deltas lookup and the $prev_data lookup; (2) fixing the sparkline key-resolution order to avoid prepending "blib/" to a key already starting with "blib/" (which produced a double-prefix "blib/blib/..." path that never matched anything in historical snapshots).
- Fix t/fuzz.t invoking bin/fuzz-harness-generator via the shebang (#!/usr/bin/env perl) rather than $^X: on CPAN smokers where $PATH resolves to a different system perl, the wrong interpreter was used, causing @INC to contain only system paths and missing declared PREREQ_PM dependencies. t/fuzz.t now calls [$^X, '...fuzz-harness-generator']. bin/fuzz-harness-generator similarly derives the prove path from $^X so the generated harness is always run under the same perl installation. Fixes GitHub#8
- Fix t/cli-test-generator-mutate.t failing on systems with git older than 1.8.5: replaced all 'git -C <dir>' invocations (which require git >= 1.8.5) with a portable chdir-based helper.
- Fix _compile_signature_isolated() calling setrlimit(RLIMIT_AS) on the parent process before open3(): the parent's address space was capped at 50 MB, causing the test framework to OOM during teardown inside a subtest, which manifested as "A context appears to have been destroyed without first calling release()" from Test::Builder and an "Out of memory" crash. The memory limit is now applied only in the child subprocess (via an env var passed through local %ENV), leaving the parent's address space unrestricted.
- Fix t/type_params.t 'signature_for' and 'documented example' subtests failing when _compile_signature_isolated() falls back to heuristics (e.g. because the 50 MB RLIMIT_AS is insufficient for Type::Params + Types::Common on some machines). The heuristics extract real parameter names ('first', 'second', 'child') while the payload hardcoded 'arg0', 'arg1' — causing a key-name mismatch. Fix: (1) update the payload to use $p->name when Type::Params 2.x provides it, falling back to 'argN' only when the attribute is absent or empty; (2) rewrite the test assertions to sort parameters by position and check type/optionality rather than specific hash key names, so they pass regardless of which extraction path was used; (3) accept either 'object' or 'string' for the Types::Standard Object parameter in the 'documented example' subtest — the heuristic cannot infer 'object' without the subprocess (the body has no ref/blessed/isa checks), so both results are valid.
- Fix Makefile.PL declaring Data::Random, Data::Random::String, Data::Random::String::Matches 0.02, Data::Random::Structure, and Unicode::GCString in TEST_REQUIRES when all five are compile-time `use` statements in lib/App/Test/Generator/Template.pm. Any `cpanm --notest` install omits test-only deps, causing Template.pm to fail to load at first use. All five moved to PREREQ_PM.
Documentation
Generate Benchmark harnesses from ATG schema files
Demonstrate the schema extractor
Deploy ATG GitHub Actions workflows to a CPAN module repository
Extract test schemas from Perl modules
Generate fuzzing + corpus-based test harnesses from test schemas
Generate round-trip tests from POD code examples
Test coverage dashboard generator
Run mutation testing against a Perl test suite
Modules
Fuzz Testing, Mutation Testing, LCSAJ Metrics and Test Dashboard for Perl modules
Generate Benchmark harnesses from ATG schemas
AFL-style coverage-guided fuzzing for App::Test::Generator
Serialise a test plan to YAML
Static LCSAJ extraction for Perl
Merge LCSAJ path data with runtime hits
Evidence-based model of a single method under test
Negate boolean return expressions to expose missing assertion coverage
Replace return expressions with undef to expose missing undef-return checks in the test suite
Generate and apply mutation tests
Extract runnable code examples from a Perl module's POD
Example module for schema extraction testing
Extract test schemas from Perl modules
Template for the test files generated by App::Test::Generator
Debugger backend for LCSAJ coverage
Provides
in lib/App/Test/Generator/Analyzer/Complexity.pm
in lib/App/Test/Generator/Analyzer/Return.pm
in lib/App/Test/Generator/Analyzer/ReturnMeta.pm
in lib/App/Test/Generator/Analyzer/SideEffect.pm
in lib/App/Test/Generator/Emitter/Perl.pm
in lib/App/Test/Generator/Mutant.pm
in lib/App/Test/Generator/Mutation/Base.pm
in lib/App/Test/Generator/Mutation/ConditionalInversion.pm
in lib/App/Test/Generator/Mutation/NumericBoundary.pm
in lib/App/Test/Generator/Planner.pm
in lib/App/Test/Generator/Planner/Fixture.pm
in lib/App/Test/Generator/Planner/Grouping.pm
in lib/App/Test/Generator/Planner/Isolation.pm
in lib/App/Test/Generator/Planner/Mock.pm
in lib/App/Test/Generator/TestStrategy.pm