det: batch file I/O holes in estates without an application JCL job (no CobolFiles), and WRITE ... ADVANCING
メンテナーはふだん 1 日以内に返信
関連するプルリクエストがすでにマージされています。
- #6111 @squid-protocol による — マージ済み
評価
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 初心者へのやさしさ
- 35/100
調査の方向性
Start with cobol_to_java_batch_forge.py and det/program.py to trace when batch runtime files and unscheduled main programs are generated, then inspect det/stmt.py around _write() for the trailing phrase handling. Use unit and synthetic COBOL programs to check WRITE ... ADVANCING against the IBM reference, and verify the named holes distinguish unsupported phrases. Done means the specified batch regression checks pass and the translation only survey reports before and after counts.
索引モデルが issue の本文から書いたものです。
説明
Why
The 2026-10-22 non-burned translation-only survey (#6064 comment) has two gaps that cover about two-thirds of the hole statements in the local non-burned corpora (dsf, zopeneditor-sample: 124 hole statements over 10 programs):
| hole reason | statements |
|---|---|
<OPEN/READ/WRITE/CLOSE> <file>: the generated project has no batch package (CobolFiles) to run files through |
~50 |
WRITE ADVANCING / other phrases |
33 |
Both are generic batch COBOL, not estate quirks, so they are likely to recur on any unseen batch estate.
1. No batch package unless the estate has an application JCL job
cobol_to_java_batch_forge.py BatchForge.sources() returns {} when self.applications is empty, so the batch runtime (CobolFiles, Sysout, CobolAbend, …) is generated only when the estate holds at least one application JCL job. det/program.py:has_batch() then finds no batch/CobolFiles.java, and every FD gets fd.why = "the generated project has no batch package…" (program.py ~L1029). Every file I/O statement becomes a hole.
dsf has no JCL at all. The unscheduled main-program path (the CardDemo CBTRN01C case: a main program with files, no USING, no COMMAREA) only counts when self.applications is already non-empty (bool(self.applications) and key in self.unscheduled, ~L343/L348).
Fix: emit the batch runtime when the estate has any batch program: an application job or an unscheduled main program with files. Treat a no-JCL estate's file-owning main programs as unscheduled batch steps, as CardDemo's CBTRN01C already is. A CICS-only estate (no FDs in a main program) must still get no batch package, as today.
2. WRITE … ADVANCING
det/stmt.py _write() (~L1000) holes anything after WRITE rec [FROM x]: "WRITE ADVANCING / other phrases". Model BEFORE/AFTER ADVANCING n LINE(S) | mnemonic | PAGE (IBM Enterprise COBOL Language Reference, WRITE statement, sequential files) for line-sequential and printer files: carriage control as IBM defines it for the record (ASA control byte when the file is ADV; the line/page skip otherwise). Refuse by name any phrase still not modelled (AT END-OF-PAGE/EOP with LINAGE, mnemonic names not in SPECIAL-NAMES), as a named hole, never a guess.
Sweep the class in the same PR: _write currently holes every trailing phrase, so check which others the surveys hit (INVALID KEY / NOT INVALID KEY on indexed/relative WRITE, END-WRITE) and model or name them separately, so the hole reason says which phrase.
Evidence constraints
- No burned estate uses
ADVANCING(grep over the 6 burned corpora: 0 files). Part 2 therefore can't reach L2+ on a burned estate: test it with unit and synthetic programs (language-crucible) against the IBM reference, and report it as translation-only until a burned program uses it. Don't build cases ondsf/zopeneditor/census repos (burned-only rule). - Part 1 regression: CardDemo/CBSA/GenApp batch must be unchanged (byte-identical generated estate where they have application jobs); the async credit-card estate (CICS-only, 0 JCL) must still get no batch package.
- Afterwards: re-run the non-burned translation-only survey (
det_survey.py --no-compileoverdsf,zopeneditor-sample) and post the before/after counts here. Baseline: dsf 0 / 7 whole, 231 / 318 statements; zopeneditor-sample 2 / 5 whole, 595 / 632.
- 主要言語
- Python
- スター
- 57
- フォーク
- 11
- 平均マージ
- 1時間 5分
- マージ済み PR（30日）
- 976
環境構築
- Dockerfile・Docker Compose ファイルなし
- プルリクエストのテンプレートあり
- コントリビューションガイドを読む
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
squid-protocol/gitgalaxy のほかの issue
-
claim:melek priority: high
squid-protocol/gitgalaxy#6534 · コメント 2 件 ·
メンテナーはふだん 1 日以内に返信
-
claim:melek priority: high
squid-protocol/gitgalaxy#6501 · コメント 2 件 ·
メンテナーはふだん 1 日以内に返信
-
claim:nyx enhancement legacy-modernization scope:core status: blocked testing
squid-protocol/gitgalaxy#6492 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
-
bug claim:nyx scope:sweep
squid-protocol/gitgalaxy#6454 · コメント 8 件 ·
メンテナーはふだん 1 日以内に返信
-
bug claim:melek scope:frontier
squid-protocol/gitgalaxy#6433 · コメント 2 件 ·
メンテナーはふだん 1 日以内に返信
squid-protocol/gitgalaxy の issue をすべて見る
似ている issue
-
Project submission:オープン
難易度 1/5 1時間未満 初心者へのやさしさ 72/100
slavakurilyak/awesome-ai-agents#845 ·
メンテナーはふだん 1 日以内に返信
-
tests: pin that the atomic writers in utils/paths.py stage the temporary file in the target's directory対応中かも このイシューにリンクされたプルリクエストがオープン中、またはマージ済みです。 オープンdifficulty:2 enhancement good first issue
難易度 1/5 1〜3時間 初心者へのやさしさ 92/100
MakazhanAlpamys/Soup#2178 ·
メンテナーはふだん 1 日以内に返信
-
[BUG] dcg.waggr permutation p-values are folded twice, so near-zero scores get the smallest p-valuesオープン
難易度 2/5 1〜3時間 初心者へのやさしさ 75/100
scverse/rapids-singlecell#854 ·
メンテナーはふだん 1 日以内に返信
-
Slang input: int8_t4_packed and uint8_t4_packed are copied into CrossGL対応中かも @ninglinLiu が今日担当しました。 オープンarea:frontend backend:slang bug direction:input
難易度 2/5 1〜3時間 初心者へのやさしさ 70/100
メンテナーはふだん 1 日以内に返信
-
random_elements(unique=True) on an OrderedDict: ZeroDivisionError once only zero weights remain, AssertionError with use_weighting=False対応中かも このイシューにリンクされたプルリクエストがオープン中、またはマージ済みです。 オープン
難易度 2/5 1〜3時間 初心者へのやさしさ 85/100
メンテナーはふだん 13 日以内に返信