Undefined CLONES array dereference aborts DBIx::Abstract test on both backends
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
调研方向
Start by reproducing the failure with the upstream t/2-dbix-abstract.t from DBIx-Abstract-1.04, which fails at line 55 on the is( @{$dbh->{'CLONES'}}, 0, ...) assertion. The issue says the standalone expression does not reproduce it, so reduce the disconnect, reconnect and explicit DESTROY sequence before looking into the runtime. Done means the minimized case passes on both JVM and interpreter backends, and t/2-dbix-abstract.t completes all 27 tests.
由索引模型根据 Issue 内容生成。
描述
Summary
DBIx-Abstract-1.04 passes its complete upstream test suite under system Perl but aborts under PerlOnJava on both execution backends. The failure occurs at the CLONES count assertion after a DBIx::Abstract object has been disconnected, reconnected, and another wrapper around its DBI handle has been explicitly destroyed.
Reproduction
Run the upstream t/2-dbix-abstract.t test from distribution DBIx-Abstract-1.04 with Check::ISA-0.09, DBI, and DBD::SQLite available. After the first nine assertions pass, PerlOnJava reports:
Can't use an undefined value as an ARRAY reference at t/2-dbix-abstract.t line 55.
The failing assertion is:
is( @{$dbh->{'CLONES'}}, 0, "no clones yet");
Immediately before it, the test has disconnected and reconnected $dbh, created a second DBIx::Abstract object around $dbh->{'dbh'}, explicitly called $dbh2->DESTROY(), then verified that $dbh is still connected. The CLONES key is absent at this point. System Perl evaluates the array dereference in the is argument list as an empty list and the assertion passes; PerlOnJava aborts while evaluating that expression.
Evidence
- CPAN tester run
20261013-145839-84917, distributionDBIx-Abstract-1.04. - System Perl 5.42.2: full upstream suite passes, 27/27 tests.
Check::ISA-0.09also passes its available tests; its optionalasatest is skipped becauseasais not installed. - PerlOnJava 5.44.1, Java 24.0.2, macOS arm64: the archived JVM run fails at test 10. Re-running the focused upstream file with the archived executable reproduces the same failure on both JVM and interpreter backends.
DBIandDBD::SQLiteload successfully in PerlOnJava, and the first nine DBIx::Abstract assertions pass, ruling out missing DBI/SQLite prerequisites as the cause.- The standalone expression on a fresh hash does not reproduce the failure. The trigger requires the test's DBIx::Abstract connection/reconnect and explicit-destruction sequence.
- GitHub duplicate search found no issue for the
CLONESfailure or this array-dereference behavior. Issue #1419 concerns Valiemon test failures and is unrelated.
Root cause
The confirmed compatibility failure is in PerlOnJava's evaluation of the array dereference in this list-valued Test::More::is argument after the DBIx::Abstract lifecycle sequence. The missing CLONES field should contribute zero list elements, as it does under system Perl; instead, PerlOnJava treats it as an invalid array reference and throws before the assertion can run. The failure is shared by both execution backends, so the defect is in shared runtime semantics rather than backend-specific execution or DBIx::Abstract itself.
The minimal trigger has not yet been reduced below the upstream lifecycle sequence. Investigation should trace how the undefined CLONES hash value and its context are represented after the explicit DESTROY call, then identify why the array dereference is rejected here while ordinary cases pass.
Proposed fix
Correct the shared runtime's handling of an undefined hash-element value dereferenced as an array in list context so that it follows Perl's behavior and yields an empty list in this case. Preserve an integration regression covering the DBIx::Abstract disconnect/reconnect and explicit-DESTROY sequence, and add a smaller unit regression once the trigger is isolated. Validate the new unit test with system Perl first, then require the focused regression and upstream test to pass on both PerlOnJava backends.
Acceptance criteria
- A minimized test that passes with system Perl also passes on JVM and interpreter backends.
t/2-dbix-abstract.tcompletes all 27 tests on both PerlOnJava backends.- The standard-Perl oracle remains green for any new project-owned unit regression.
- 主要语言
- Perl
- 星标
- 64
- 派生
- 7
- 平均合并
- 5 小时 31 分钟
- 30 天内合并 PR
- 214
环境准备
- 提供 Dockerfile 或 Docker Compose 文件
- 没有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
fglock/PerlOnJava 的其他 Issue
-
bug
难度 2/5 1-3 小时 新手友好度 72/100
fglock/PerlOnJava#1579 ·
维护者通常 1 天内回复
-
难度 3/5 1-2 天 新手友好度 48/100
fglock/PerlOnJava#1835 ·
维护者通常 1 天内回复
-
area:cpan-port bug
难度 3/5 1-2 天 新手友好度 68/100
fglock/PerlOnJava#1820 ·
维护者通常 1 天内回复
-
难度 3/5 半天 新手友好度 45/100
fglock/PerlOnJava#1808 ·
维护者通常 1 天内回复
-
area:cpan-port area:runtime bug
难度 4/5 3-5 天 新手友好度 28/100
fglock/PerlOnJava#1802 · 2 条评论 ·
维护者通常 1 天内回复
查看 fglock/PerlOnJava 的全部 Issue
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 80/100
jimsalterjrs/sanoid#1116 · 1 个 reaction ·
-
Add .gitattributes so share/skel keeps LF on a Windows git checkout可能已有人在做 @sadhana-s-codez 于 2 天前认领。 未关闭
难度 1/5 1 小时以内 新手友好度 85/100
PerlDancer/Dancer2#1854 · 1 条评论 ·
维护者通常 1 天内回复
-
bug
难度 2/5 1-3 小时 新手友好度 78/100
major/MySQLTuner-perl#1007 ·
维护者通常 1 天内回复
-
难度 1/5 1 小时以内 新手友好度 88/100
-
难度 1/5 1 小时以内 新手友好度 88/100