Method lookup still finds a subroutine after Symbol::Util delete_sub
維護者通常 1 天內回覆
還沒有人認領這個 Issue。
評估
研究方向
Start with the upstream t/50delete_sub.t test, especially test 23, and reproduce it under standard Perl to confirm the expected result. Trace the method lookup and symbol-table deletion paths, then add a project-owned regression covering delete_sub on both JVM and interpreter backends. Done means the focused regression and unchanged Symbol-Util 0.0203 test pass on both backends.
由索引模型根據 Issue 內容生成。
描述
Summary
After Symbol::Util::delete_sub removes a subroutine from the main package, PerlOnJava can still dispatch that subroutine as a class method. This breaks Perl's method lookup semantics: deleting main::FOO should make main->FOO unavailable, while unrelated calls through the remaining symbol table continue to work.
This is a pure-Perl compatibility failure, reproducible on both PerlOnJava execution backends. The unchanged upstream test passes under system Perl.
Reproduction
The failure is from Symbol-Util 0.0203, t/50delete_sub.t, test 23 (main->FOO is ok [2]). The relevant upstream sequence is:
use Symbol::Util 'delete_sub';
*FOO = sub { "code" };
# The method is initially available.
my $before = eval { main->FOO };
# Remove the CODE slot for main::FOO.
delete_sub("::FOO");
# Expected: method lookup fails, so the eval result is undefined.
my $after = eval { main->FOO };
Expected behavior under Perl: defined $after is false. On PerlOnJava, the upstream assertion ok(! defined eval { main->FOO }, 'main->FOO is ok [2]') fails, indicating that method dispatch still finds a defined result after deletion.
Evidence
- CPAN random tester run:
20261002-095841-54171 - Distribution:
Symbol-Util0.0203 (pure Perl) - Failing test:
t/50delete_sub.t, assertion 23 of 24 - Full PerlOnJava suite result: 306/307 assertions passed; 1/8 test programs failed
- System Perl: all 307 assertions across 8 test programs passed in the archived oracle run; the focused
t/50delete_sub.talso passes independently - PerlOnJava JVM backend: focused test fails at assertion 23
- PerlOnJava interpreter backend: same focused assertion fails
- Regression record: last known pass date 2026-09-12 at commit
141cdae65 - Reproduced in current checkout at commit
0d3f46e4524573d58dd4ae91565f6de2d8b79435 - Environment: macOS 26.6.2, Temurin OpenJDK 24.0.2, system Perl 5.42.2
The archived failing log also prints closedir() attempted on handle $GEN4 ... during CPAN configuration, but the target completed its tests and reported one ordinary assertion failure. The system-Perl oracle passed, and the focused failure reproduces directly on both backends, so that warning does not explain this result.
Expected behavior
Once delete_sub("::FOO") removes main::FOO's CODE slot, a subsequent main->FOO lookup must fail on both execution backends. The method lookup path must reflect the current package symbol table after this mutation.
Related issues
- #1119 covers earlier
Symbol::Utilpackage-stash and typeglob-slot failures. Its report says subroutine deletion passed; this CPAN run records a later regression in that behavior. - #1523 covers method loss after a package stash is copied back. It is related to method dispatch after symbol-table mutation, but this report concerns stale successful lookup after
delete_subremoves a method.
Suggested acceptance criteria
- Add a focused project-owned regression test for method lookup after
delete_subremoves a named CODE slot. - Validate the test with standard Perl before using it as a PerlOnJava regression test.
- Confirm the test passes on both JVM and interpreter backends.
- Confirm the unchanged
Symbol-Util0.0203t/50delete_sub.ttest passes.
- 主要語言
- Perl
- 星號
- 65
- 分支
- 8
- 平均合併
- 5 小時 49 分鐘
- 30 天內合併 PR
- 177
環境準備
- 提供 Dockerfile 或 Docker Compose 檔案
- 沒有 Pull Request 範本
- 閱讀貢獻指南
從這裡開始
- 先讀完整個 Issue，再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫，在一個分支上完成修改。
- 送出 Pull Request，並在描述裡引用這個 Issue 編號。
fglock/PerlOnJava 的其他 Issue
-
area:runtime documentation enhancement
難度 2/5 1-3 小時 新手友好度 72/100
fglock/PerlOnJava#1579 ·
維護者通常 1 天內回覆
-
area:runtime bug
難度 4/5 1-2 天 新手友好度 55/100
fglock/PerlOnJava#1958 · 1 則留言 ·
維護者通常 1 天內回覆
-
Fix Java TLS socket reads for chunked HTTPS response bodies可能已有人在做 關聯的 PR 仍在進行中或已合併。 未關閉area:io bug
難度 4/5 3-5 天 新手友好度 56/100
fglock/PerlOnJava#1957 · 2 則留言 ·
維護者通常 1 天內回覆
-
area:cpan-port enhancement
難度 5/5 一週以上 新手友好度 25/100
fglock/PerlOnJava#1955 ·
維護者通常 1 天內回覆
-
area:runtime bug
難度 4/5 3-5 天 新手友好度 22/100
fglock/PerlOnJava#1951 ·
維護者通常 1 天內回覆
查看 fglock/PerlOnJava 的全部 Issue
相似的 Issue
-
難度 2/5 1-3 小時 新手友好度 72/100
libwww-perl/HTTP-Message#236 ·
-
1.severity: security
難度 2/5 1-3 小時 新手友好度 68/100
維護者通常 1 天內回覆
-
難度 2/5 1-3 小時 新手友好度 76/100
維護者通常 2 天內回覆
-
wave-followup
難度 2/5 1-3 小時 新手友好度 74/100
-
documentation
難度 2/5 1-3 小時 新手友好度 72/100
維護者通常 1 天內回覆