abhigyanpatwari/GitNexus · error · ValueError

plan artifact changed while hashing: {path}

Error message

plan artifact changed while hashing: {path}

What it means

Thrown by snapshot_plan_docs after it has streamed the whole plan file through sha256. Once the digest is computed the descriptor is fstat'd again; if st_size or st_mtime_ns changed between the pre-read fstat and this post-read fstat, the file was modified while being hashed, so the digest does not describe a stable artifact. The harness rejects it to keep evidence reproducible.

Source

Thrown at eval/workflow_bench/runner_artifacts.py:281

    snapshot: dict[Path, str] = {}
    for path in sorted(plans.iterdir()):
        if path.suffix.lower() not in {".md", ".html"}:
            continue
        metadata = path.lstat()
        if stat.S_ISLNK(metadata.st_mode):
            raise ValueError(f"plan artifact cannot be a symlink: {path}")
        if not stat.S_ISREG(metadata.st_mode):
            raise ValueError(f"plan artifact must be a regular file: {path}")
        descriptor = os.open(path, os.O_RDONLY | getattr(os, "O_NOFOLLOW", 0))
        try:
            opened = os.fstat(descriptor)
            if not stat.S_ISREG(opened.st_mode) or opened.st_dev != metadata.st_dev or opened.st_ino != metadata.st_ino:
                raise ValueError(f"plan artifact changed while opening: {path}")
            with os.fdopen(descriptor, "rb", closefd=False) as handle:
                snapshot[path] = hashlib.file_digest(handle, "sha256").hexdigest()
            after = os.fstat(descriptor)
            if (opened.st_size, opened.st_mtime_ns) != (after.st_size, after.st_mtime_ns):
                raise ValueError(f"plan artifact changed while hashing: {path}")
        finally:
            os.close(descriptor)
    return snapshot


def new_plan_doc(worktree: Path, before: dict[Path, str]) -> Path:
    """Return the sole new or modified plan, rejecting ambiguous evidence."""

    after = snapshot_plan_docs(worktree)
    deleted = sorted(path for path in before if path not in after)
    if deleted:
        raise ValueError("planning deleted existing plan artifact(s): " + ", ".join(str(path) for path in deleted))
    changed = sorted(path for path, digest in after.items() if before.get(path) != digest)
    if len(changed) != 1:
        raise ValueError(f"planning must create or modify exactly one plan artifact; observed {len(changed)}")
    return changed[0]

View on GitHub (pinned to d540b00184)

Solutions

  1. Guarantee the agent process is fully terminated before snapshot_plan_docs runs.
  2. Disable any file-sync or auto-format watcher on the worktree's docs/plans directory.
  3. Move the worktree off any FUSE/network filesystem onto local disk.
  4. Retry the arm once the writing process is confirmed stopped (check no process holds the file open with lsof).

Example fix

// before
agent_proc.terminate()  # snapshot may race with flush
snapshot = snapshot_plan_docs(worktree)

// after
agent_proc.wait()  # ensure all writes flushed and fd closed
snapshot = snapshot_plan_docs(worktree)
Defensive patterns

Strategy: validation

Validate before calling

import os, stat, time
from pathlib import Path

def is_stable_during_read(path: Path) -> bool:
    st_before = path.stat()
    with open(path, 'rb') as fh:
        fh.read()
    st_after = path.stat()
    return (st_before.st_size, st_before.st_mtime_ns) == (st_after.st_size, st_after.st_mtime_ns)

Type guard

null

Try / catch

try:
    snap = snapshot_plan_docs(worktree)
except ValueError as e:
    if 'changed while hashing' in str(e):
        # file modified during read; ensure writer is stopped and retry once
        raise SystemExit('plan modified during hashing; stop the writer and retry')
    raise

Prevention

When it happens

Trigger: (opened.st_size, opened.st_mtime_ns) != (after.st_size, after.st_mtime_ns) after hashlib.file_digest completes. The plan file's size or nanosecond mtime shifted during the read loop.

Common situations: An agent still writing its plan when the snapshot runs; an editor/linter auto-formatting the file on a timer; a sync daemon (Dropbox/iCloud) touching the file; a filesystem that updates mtime on read (rare, some FUSE setups).

Related errors


AI-assisted analysis of abhigyanpatwari/GitNexus@d540b00184 (2026-08-12). Data as JSON: /api/errors/aac414a1de97ec55. Report an issue: GitHub.