Skip to content

Commit 060799d

Browse files
Alezander9Alezander9
authored
Merge pull request #19 from browser-use/tooling/check-harness-diff
tooling: script/check-harness-diff.sh
2 parents 1f64e1c + 791de51 commit 060799d

2 files changed

Lines changed: 107 additions & 2 deletions

File tree

‎harness-sync.md‎

Lines changed: 5 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -35,10 +35,13 @@ If the divergences table is empty (initial vendor state), every difference betwe
3535
### 3. Check drift
3636

3737
```sh
38-
script/check-upstream.sh
38+
script/check-upstream.sh # commit count: how many commits behind
39+
script/check-harness-diff.sh # file-level diff vs harness/main, with known-divergence filter
3940
```
4041

41-
Reports how many commits `harness/main` is ahead of our recorded `To SHA`. Then inspect what changed:
42+
`check-upstream.sh` reports how many commits `harness/main` is ahead of our recorded `To SHA`. `check-harness-diff.sh` shows per-file differences between our vendored tree and `harness/main`, splitting them into "known divergences (UPSTREAM.md §3)" and "unexpected drift" — the latter should always be either (a) commits we haven't synced yet, or (b) a Yellow-zone modification we forgot to record. Anything else is a bug.
43+
44+
Then inspect what changed:
4245

4346
```sh
4447
git fetch harness main

‎script/check-harness-diff.sh‎

Lines changed: 102 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,102 @@
1+
#!/usr/bin/env bash
2+
# check-harness-diff.sh — report exact file-level diff between our vendored
3+
# harness and upstream `harness/main`.
4+
#
5+
# Companion to `check-upstream.sh` (which reports commit count). This one
6+
# reports per-file changes, so you can see at a glance whether we're drifting
7+
# on files we shouldn't be.
8+
#
9+
# Usage: script/check-harness-diff.sh
10+
# Exits 0 regardless of diff size — informational.
11+
#
12+
# Output: three sections
13+
# 1. Files changed (per-file list, post-filter for noise + known divergences)
14+
# 2. Summary (insertions/deletions vs upstream HEAD)
15+
# 3. Verdict (clean / known divergences / unexpected drift)
16+
17+
set -euo pipefail
18+
19+
REPO_ROOT="$(git rev-parse --show-toplevel)"
20+
VENDORED="$REPO_ROOT/packages/bcode-browser/harness"
21+
TMP="$(mktemp -d)"
22+
trap 'rm -rf "$TMP"' EXIT
23+
24+
if ! git remote | grep -qx "harness"; then
25+
echo "error: 'harness' remote not configured. Add it:" >&2
26+
echo " git remote add harness https://github.com/browser-use/browser-harness.git" >&2
27+
exit 1
28+
fi
29+
30+
git fetch --quiet harness main
31+
32+
UPSTREAM_HEAD="$(git rev-parse harness/main)"
33+
UPSTREAM_SHORT="$(git rev-parse --short harness/main)"
34+
35+
# Extract upstream HEAD into a temp dir without touching index/worktree.
36+
# `git archive | tar -x` is the right tool here — `git --work-tree=X checkout`
37+
# mutates the active index relative to the current branch, which is wrong.
38+
git archive --format=tar "$UPSTREAM_HEAD" | tar -xf - -C "$TMP"
39+
40+
# Known-divergence filter:
41+
# - .gitignore (we add .venv/ — see UPSTREAM.md §3 divergences table)
42+
# Build artifacts the vendored side might generate during smoke tests:
43+
# - uv.lock, .venv/, __pycache__/, *.egg-info/, *.pyc
44+
# These are gitignored on our side, but `diff -rq` doesn't read .gitignore.
45+
# Files in the divergences table go in EXPECTED; everything else is drift.
46+
EXPECTED_REGEX='/(\.gitignore)( |$)'
47+
# `diff -rq` emits two line shapes:
48+
# "Files A and B differ"
49+
# "Only in <dir>: <name>"
50+
# Match noise in either shape.
51+
NOISE_REGEX='(uv\.lock|\.venv|__pycache__|\.egg-info|\.pyc|\.cache)'
52+
53+
DIFF_OUT="$(diff -rq "$VENDORED/" "$TMP/" 2>&1 | grep -Ev "$NOISE_REGEX" || true)"
54+
55+
echo "=== vendored vs harness/main ($UPSTREAM_SHORT) ==="
56+
echo
57+
58+
if [[ -z "$DIFF_OUT" ]]; then
59+
echo " No differences. Vendored harness matches upstream HEAD exactly."
60+
echo
61+
exit 0
62+
fi
63+
64+
EXPECTED="$(echo "$DIFF_OUT" | grep -E "$EXPECTED_REGEX" || true)"
65+
UNEXPECTED="$(echo "$DIFF_OUT" | grep -Ev "$EXPECTED_REGEX" || true)"
66+
67+
if [[ -n "$EXPECTED" ]]; then
68+
echo "Known divergences (UPSTREAM.md §3):"
69+
echo "$EXPECTED" | sed 's|^| |'
70+
echo
71+
fi
72+
73+
if [[ -n "$UNEXPECTED" ]]; then
74+
echo "Unexpected drift:"
75+
echo "$UNEXPECTED" | sed 's|^| |'
76+
echo
77+
echo "Each line is one of:"
78+
echo " - upstream commit we haven't synced yet (run script/check-upstream.sh to see commits behind), or"
79+
echo " - a Yellow-zone modification we forgot to record in UPSTREAM.md §3."
80+
echo
81+
fi
82+
83+
# Per-file line stats vs upstream, excluding noise + known-divergence files.
84+
# `diff -ruN` emits per-hunk patch text; we want body lines only (skip
85+
# +++/--- headers and @@ hunk markers), and we exclude any hunk whose header
86+
# matched the noise regex.
87+
echo "Line stats vs upstream (added on our side, removed on our side):"
88+
diff -ruN \
89+
--exclude='.venv' --exclude='__pycache__' --exclude='*.egg-info' \
90+
--exclude='*.pyc' --exclude='uv.lock' --exclude='.cache' \
91+
"$TMP" "$VENDORED" 2>/dev/null \
92+
| awk '
93+
/^\+\+\+ |^--- / { in_header=1; next }
94+
/^@@ / { in_header=0; next }
95+
in_header { next }
96+
/^\+/ { added++ }
97+
/^-/ { removed++ }
98+
END { printf " +%d / -%d lines (excluding headers/hunk markers)\n", added+0, removed+0 }
99+
'
100+
101+
echo
102+
echo "Source of truth: UPSTREAM.md §3. Update the divergences table when adding/removing intentional differences."

0 commit comments

Comments
 (0)