From ebfcc4375a182a2332f58f79848ddf07439d89ba Mon Sep 17 00:00:00 2001 From: bboonnee Date: Mon, 3 Aug 2026 12:59:06 +0800 Subject: [PATCH] U-013 follow-up: drop the exclude pathspec, it was itself a hard error --- docs/development-plan.md | 1 + src/test_work_node_run.py | 7 ++++++- src/work_node_run.py | 24 +++++++++++++++--------- 3 files changed, 22 insertions(+), 10 deletions(-) diff --git a/docs/development-plan.md b/docs/development-plan.md index 47500f2..7537576 100644 --- a/docs/development-plan.md +++ b/docs/development-plan.md @@ -137,6 +137,7 @@ - 覆盖：01-M `work_node_execution`；03-C `c_work_node_run`；REQ-03（boundary） - **真实 bug（2026-08-02/03，U-042 第一次真实自举压测时发现）**：`RealGitWorktreeIsolationTest` 从 U-013 建成那天起，`run_fn` 一直是测试自己写的、写完文件后主动 `git add` + `git commit` 的假实现——这个盲区从没被真实 Coder 会话踩过，因为 PH1~PH4 所有真实调用都没启用 `needs_worktree=True`（development-plan.md U-031 的实现补充里已经记过这条），直到 U-035 让 self-target 场景无条件强制 worktree，U-042 第一次真实自举跑起来后才第一次真正踩中。真实现象：Coder 在 worktree 里确实编辑了文件，自己也验证过（`git status`/`git diff --stat` 都在它自己的 `test_report` 里回显了），但从未 `git commit`——Coder 人设里没有任何一处要求它提交自己的改动。`merge_worktree()` 只会 `git merge` 已提交的分支，worktree 分支上如果压根没有新提交，merge 就是个真正的空操作（"Already up to date"，不产生新 commit），紧接着 `remove_worktree()` 把这个还带着未提交改动的目录直接删掉——一次真实、已验证过的 Coder 改动就这样被无声吞掉，`implement_code()` 算出来的 `code_diff` 自然是空的。修复：新增 `work_node_run.ensure_worktree_committed()`，在 `merge_worktree()` 之前检查 worktree 是否有未提交改动（`git status --porcelain`），有则自动 `git add -A` + 提交一条 `graphlemon: auto-commit uncommitted worktree changes for `，worktree 本来就干净（比如纯排查、没有真的改代码）时什么也不做。新增两个回归测试：`test_uncommitted_worktree_changes_are_auto_committed_before_merge`（复现"改了但没提交"场景，断言改动最终真的落到主分支）、`test_clean_worktree_with_no_changes_is_still_a_valid_merge`（确认"确实没什么好改"这种合法结果不会被误伤）。这是"判断用 Agent，机制用代码"原则的又一次体现——不指望靠给人设加一条"记得 commit"的提示词来保证正确性，而是在系统层面兜底，人设漂移/遗忘不应该导致真实工作成果被无声丢弃。 - **真实 bug 的第二层（2026-08-03，U-042 第八/九次真实压测发现，比上面这条更深一层）**：`ensure_worktree_committed()` 只兜住了 worktree 自己的未提交改动，但 git worktree 的设计本身是**和主仓库共享同一个项目身份**（同一个 `.git`，只是另一份检出）——一次真实压测发现，`claude -p` 的 `memory: project` 写入并不总是遵守调用时传入的 `cwd`，而是可能直接落在**主仓库自己的检出目录**上，和当时具体是哪个节点、哪次调用完全无关（不是 `_check_prd_consistency` 一个函数的问题，把它单独包一层 worktree 只是挪动了"这次踩中的是哪次调用"，没有从根上解决）。后果：主仓库工作区多出一处未提交的 memory 文件改动，下一个节点自己的、原本完全正常的 worktree merge 会因为"主分支有本地改动会被覆盖"直接报错退出（`git merge` exit 2）。修复：`ensure_worktree_committed()` 从只处理 worktree 自身，扩展成同时也对 `repo_path`（主仓库检出目录）本身跑一遍同样的"有脏改动就自动提交"逻辑，在每次 `merge_worktree()` 之前调用两次——不管是哪次真实调用造成的漂移、原因是什么，统一兜底，而不是针对某一个特定调用点单独打补丁。修复过程中顺带发现并解决了这条兜底自己引入的一个新问题：对主仓库跑 `git add -A` 时，如果某个节点的 worktree 检出目录（`.graphlemon-worktrees//`）当时还没被 `remove_worktree()` 清理，会被当成一个类似 git link 的条目一并加进去，紧接着目录被删掉后 `git status` 就会显示成一条诡异的"D"（删除）——两处修复：`.graphlemon-worktrees/` 加进 `.gitignore`（这是长期正确性），`ensure_worktree_committed()` 自己的 `git add -A`/`git status --porcelain` 也显式排除这个路径（`-- . ':!.graphlemon-worktrees'`），不依赖调用方仓库一定配置了对的 `.gitignore`。新增回归测试 `test_drift_on_main_checkout_is_auto_committed_before_merge`，模拟"改动落在主仓库而不是 worktree"这个具体场景，并验证接下来一个节点的合并不受影响。 + - **上面这条修复自己的一个真 bug（2026-08-03，U-042 第十次真实压测立刻发现）**：`-- . ':!.graphlemon-worktrees'` 这个显式排除 pathspec，一旦该路径**同时**已经在 `.gitignore` 里（这次修复自己刚加的），`git add -A` 会直接报错退出（"paths are ignored by one of your .gitignore files"，即使配置 `advice.addIgnoredFile=false` 抑制提示文案，退出码依然是 1，不是单纯的提示）——本地复现过，是真的会失败，不是巧合。既然 `.gitignore` 已经能让普通的 `git add -A`（不带任何显式 pathspec）正确跳过这个目录，显式排除反而是多余且互相冲突的，直接删掉，`ensure_worktree_committed()` 恢复成最简单的 `git add -A`。同时把 `test_work_node_run.py` 的 `_init_git_repo()` 测试夹具也补上同一条 `.gitignore` 规则，让所有真实用到 `git worktree add` 的测试仓库和生产环境配置保持一致，而不是靠巧合"测试仓库正好没触发这条冲突"。 ### PH1-S3 端到端验收（收口） diff --git a/src/test_work_node_run.py b/src/test_work_node_run.py index 040fc73..14d8881 100644 --- a/src/test_work_node_run.py +++ b/src/test_work_node_run.py @@ -23,7 +23,12 @@ def _init_git_repo(path): subprocess.run(["git", "init"], cwd=path, check=True, capture_output=True) with open(os.path.join(path, "README.md"), "w") as f: f.write("seed\n") - subprocess.run(["git", "add", "README.md"], cwd=path, check=True, capture_output=True) + # matches graphlemon's own .gitignore (2026-08-03) -- without this, + # ensure_worktree_committed's plain `git add -A` on repo_path picks up + # a still-present worktree checkout as a tracked gitlink-like entry. + with open(os.path.join(path, ".gitignore"), "w") as f: + f.write(".graphlemon-worktrees/\n") + subprocess.run(["git", "add", "README.md", ".gitignore"], cwd=path, check=True, capture_output=True) subprocess.run( ["git"] + work_node_run._GIT_IDENTITY + ["commit", "-m", "seed"], cwd=path, check=True, capture_output=True, diff --git a/src/work_node_run.py b/src/work_node_run.py index 9183d6c..5854973 100644 --- a/src/work_node_run.py +++ b/src/work_node_run.py @@ -144,19 +144,25 @@ def ensure_worktree_committed(path, node_exec_id): why -- `path` being a worktree or the main checkout doesn't change what "auto-commit any dirty state here" means.""" # `.graphlemon-worktrees/` (create_worktree's own checkout location) - # lives INSIDE repo_path while a node is running -- excluded from both - # the dirty-check and the add, not just relying on the caller repo - # having a matching .gitignore (2026-08-03: without this, `git add -A` - # on repo_path picked up the still-present worktree checkout itself as - # a tracked gitlink-like entry, which then showed as "deleted" the - # moment remove_worktree() ran right after). - exclude_worktrees = ["--", ".", ":!.graphlemon-worktrees"] + # lives INSIDE repo_path while a node is running -- without excluding + # it, `git add -A` on repo_path picked up the still-present worktree + # checkout itself as a tracked gitlink-like entry, which then showed as + # "deleted" the moment remove_worktree() ran right after (2026-08-03). + # An explicit `:!.graphlemon-worktrees` pathspec exclude was tried + # first but is itself a hard error ("paths are ignored by one of your + # .gitignore files", exit 1, even with the advice hint suppressed) the + # moment the path is ALSO gitignored -- which it now always is (see + # .gitignore). Plain `git add -A` with no explicit pathspec already + # skips gitignored paths correctly on its own; the caller's repo just + # needs `.graphlemon-worktrees/` in its own `.gitignore` for this to + # hold (graphlemon's own repo has it; test fixtures that use a real + # worktree checkout add the same entry in their own setUp). status = subprocess.run( - ["git", "status", "--porcelain"] + exclude_worktrees, cwd=path, check=True, capture_output=True, text=True, + ["git", "status", "--porcelain"], cwd=path, check=True, capture_output=True, text=True, ).stdout if not status.strip(): return - subprocess.run(["git", "add", "-A"] + exclude_worktrees, cwd=path, check=True, capture_output=True) + subprocess.run(["git", "add", "-A"], cwd=path, check=True, capture_output=True) subprocess.run( _prefix_git(["commit", "-m", "graphlemon: auto-commit uncommitted changes for " + node_exec_id]), cwd=path, check=True, capture_output=True, text=True, -- libgit2 0.21.4