settled-reach/.config/hooks/pre-push

#!/usr/bin/env bash
# Pre-push hook: lint GDScript and Rust before pushing.
# Installed via: git config core.hooksPath .config/hooks
set -euo pipefail

REPO_ROOT="$(git rev-parse --show-toplevel)"
ERRORS=0
FAILED_CHECKS=""

# Record a failed check BY NAME, not just as a tally.
#
# Why (2026-07-27): this hook used to increment a bare counter at 13 sites and
# end with "N check(s) failed. Fix the errors above." — which named nothing. Six
# of those sites (fmt, clippy, cargo test, deny, ruff, tooling) print no FAIL
# line at all, so a failure was only inferable from the ABSENCE of an "— OK".
# A real push aborted on `cargo fmt` and the verdict was indistinguishable from
# any other failure; finding it meant scrolling past thousands of lines of
# unrelated test-fixture output.
#
# The consequence that matters for tooling: the final block is now
# self-contained, so `tail` on this log always shows WHAT failed. Verbosity was
# never the problem — detail is exactly what you want when something breaks;
# the problem was that the verdict didn't say anything actionable.
fail_check() {
    ERRORS=$((ERRORS + 1))
    FAILED_CHECKS="${FAILED_CHECKS}  - $1
"
}

echo "pre-push: running lint checks..."

# --- Detect which directories have changes vs remote ---
# Prefer origin/<branch> as the baseline (what the remote already has),
# but fall back to origin/main for first-push of a new branch — otherwise
# every check runs against nothing and the hook treats the whole repo as
# changed, spending tens of seconds on linters and JSON validation that
# have no diff to cover (e.g. pushing a wiki-only branch rebuilds GDScript
# and runs clippy + ruff + validates all 2762 JSON files).
BRANCH=$(git branch --show-current)
if git rev-parse --verify "origin/$BRANCH" >/dev/null 2>&1; then
    REMOTE_REF="origin/$BRANCH"
elif git rev-parse --verify "origin/main" >/dev/null 2>&1; then
    REMOTE_REF="origin/main"
else
    REMOTE_REF=""
fi

if [ -n "$REMOTE_REF" ]; then
    CLIENT_CHANGED=$(git diff --name-only "$REMOTE_REF"..HEAD -- client/ 2>/dev/null | wc -l)
    SERVER_CHANGED=$(git diff --name-only "$REMOTE_REF"..HEAD -- server/ 2>/dev/null | wc -l)
    TOOLING_CHANGED=$(git diff --name-only "$REMOTE_REF"..HEAD -- tooling/ pyproject.toml 2>/dev/null | wc -l)
else
    # No remote at all (e.g. fresh clone before first fetch) — be safe, run everything
    CLIENT_CHANGED=1
    SERVER_CHANGED=1
    TOOLING_CHANGED=1
fi

# --- GDScript parse check (headless Godot) ---
GODOT="${GODOT:-godot}"
if [ "$CLIENT_CHANGED" -eq 0 ]; then
    echo "pre-push: no client/ changes — skipping GDScript checks"
elif command -v "$GODOT" >/dev/null 2>&1 && [ -d "$REPO_ROOT/client/.godot" ]; then
    echo "pre-push: checking GDScript (parse)..."
    SCRIPT_ERRORS=$("$GODOT" --headless --path "$REPO_ROOT/client" --quit 2>&1 | grep -ci "SCRIPT ERROR" || true)
    if [ "$SCRIPT_ERRORS" -gt 0 ]; then
        echo "pre-push: FAIL — $SCRIPT_ERRORS GDScript error(s) found"
        "$GODOT" --headless --path "$REPO_ROOT/client" --quit 2>&1 | grep -i "SCRIPT ERROR"
        fail_check "GDScript parse (startup)"
    else
        echo "pre-push: GDScript parse — OK"
    fi
else
    echo "pre-push: skipping GDScript parse (no .godot/ import — run Godot once to enable)"
fi

# --- GDScript lint (gdlint static analysis) — advisory only until codebase is clean ---
if [ "$CLIENT_CHANGED" -gt 0 ] && command -v gdlint >/dev/null 2>&1 && [ -d "$REPO_ROOT/client/scripts" ]; then
    echo "pre-push: checking GDScript (gdlint — advisory)..."
    LINT_COUNT=$(gdlint "$REPO_ROOT/client/scripts/" "$REPO_ROOT/client/ui/" 2>&1 | grep -c "Error:" || true)
    if [ "$LINT_COUNT" -gt 0 ]; then
        echo "pre-push: gdlint — $LINT_COUNT issue(s) (advisory, not blocking)"
    else
        echo "pre-push: gdlint — OK"
    fi
fi

# --- GDScript format check (gdformat) — advisory only until codebase is clean ---
if [ "$CLIENT_CHANGED" -gt 0 ] && command -v gdformat >/dev/null 2>&1 && [ -d "$REPO_ROOT/client/scripts" ]; then
    echo "pre-push: checking GDScript (gdformat — advisory)..."
    FORMAT_COUNT=$(gdformat --check "$REPO_ROOT/client/scripts/" "$REPO_ROOT/client/ui/" 2>&1 | grep -c "would reformat" || true)
    if [ "$FORMAT_COUNT" -gt 0 ]; then
        echo "pre-push: gdformat — $FORMAT_COUNT file(s) need formatting (advisory, not blocking)"
    else
        echo "pre-push: gdformat — OK"
    fi
fi

# --- GDScript parse sweep — blocking, and deliberately BEFORE the suite ---
# Every project .gd must parse. ~3.7s against the suite's ~135s, and it runs
# first because that ordering makes failures CHEAPER, not the gate slower: a
# script that does not parse is now caught in 4s instead of after two minutes
# of tests that could never have covered it. A clean push pays 3.7s; a broken
# one saves over two minutes.
#
# Not redundant with the suite. gdUnit4 reports the suites that DID load as a
# clean pass, so a file that fails to parse reads as success (guarded now in
# tests/run-godot, but only for test files). This covers all 226 scripts,
# including the ~half of the codebase no test ever loads. It is also the only
# gate that covers them at all: godot-cold-parse is skill-only (never invoked
# by this hook) and sees only the startup path regardless.
if [ "$CLIENT_CHANGED" -gt 0 ] && command -v reach >/dev/null 2>&1; then
    echo "pre-push: sweeping GDScript parse (reach godot parse-sweep)..."
    if ! reach --no-input godot parse-sweep; then
        echo "pre-push: parse sweep FAILED — a script does not parse"
        fail_check "GDScript parse sweep"
    else
        echo "pre-push: parse sweep — OK"
    fi
fi

# --- Godot client test suite (gdUnit4) — blocking (T-1065) ---
# The push gate is the only automatic verification (no CI). The suite is
# ~100s, 300s-capped, and was made fully green by the T-973 debt clearance;
# binary-dependent e2e suites self-skip when the server binary is absent.
if [ "$CLIENT_CHANGED" -gt 0 ] && [ -x "$REPO_ROOT/tests/run-godot" ]; then
    echo "pre-push: running client test suite (tests/run-godot)..."
    if ! "$REPO_ROOT/tests/run-godot"; then
        echo "pre-push: client test suite FAILED"
        fail_check "client test suite (gdUnit4)"
    else
        echo "pre-push: client tests — OK"
    fi
fi

# --- Rust lint (clippy + fmt) ---
if [ "$SERVER_CHANGED" -eq 0 ]; then
    echo "pre-push: no server/ changes — skipping Rust checks"
elif command -v cargo >/dev/null 2>&1 && [ -d "$REPO_ROOT/server" ]; then
    # fmt only needs source files — always safe to run
    echo "pre-push: checking Rust (fmt)..."
    if ! (cd "$REPO_ROOT/server" && cargo fmt --check 2>&1); then
        fail_check "cargo fmt"
    else
        echo "pre-push: fmt — OK"
    fi

    # clippy needs a build — skip if target/ doesn't exist (cold worktree).
    # --all-targets lints tests + examples too (#967 closed the gap where test
    # code accumulated clippy debt unchecked).
    if [ -d "$REPO_ROOT/server/target" ]; then
        echo "pre-push: checking Rust (clippy)..."
        if ! (cd "$REPO_ROOT/server" && cargo clippy --all-targets -- -D warnings 2>&1); then
            fail_check "cargo clippy"
        else
            echo "pre-push: clippy — OK"
        fi

        # cargo test — the ONLY automatic correctness gate: nothing else (no CI
        # workflows exist) runs the suite, so without this a Rust regression
        # reaches main unverified. Gated on server/ changes; shares the target/
        # guard with clippy so a cold worktree isn't forced into a full build.
        echo "pre-push: checking Rust (cargo test)..."
        if ! (cd "$REPO_ROOT/server" && cargo test --quiet 2>&1); then
            fail_check "cargo test"
        else
            echo "pre-push: cargo test — OK"
        fi
    else
        echo "pre-push: skipping clippy + test (no target/ — run 'cargo build' once to enable)"
    fi

    # --- Rust dependency audit (cargo deny) — requires deny.toml config ---
    if command -v cargo-deny >/dev/null 2>&1 && [ -f "$REPO_ROOT/server/deny.toml" ]; then
        echo "pre-push: checking Rust (cargo deny)..."
        if ! (cd "$REPO_ROOT/server" && cargo deny check 2>&1); then
            fail_check "cargo deny"
        else
            echo "pre-push: cargo deny — OK"
        fi
    fi
else
    echo "pre-push: WARNING — cargo not found or server/ missing, skipping Rust lint"
fi

# --- Python lint (ruff) ---
if [ "$TOOLING_CHANGED" -eq 0 ]; then
    echo "pre-push: no tooling/ changes — skipping Python lint"
elif command -v ruff >/dev/null 2>&1 && [ -d "$REPO_ROOT/tooling" ]; then
    echo "pre-push: checking Python (ruff)..."
    if ! (cd "$REPO_ROOT" && ruff check tooling/ 2>&1); then
        fail_check "ruff (python lint)"
    else
        echo "pre-push: ruff — OK"
    fi
else
    echo "pre-push: skipping Python lint (ruff not found — install with: pip install 'ruff>=0.9')"
fi

# --- Tooling test gate (T-1066) ---
# make test-tooling = planet-gen determinism guard (#963) + import_economics
# --dry-run validation against the committed DB. Only worth the ~90 s when the
# push actually touches tooling/ (or pyproject.toml), same scope as ruff above.
if [ "$TOOLING_CHANGED" -eq 0 ]; then
    echo "pre-push: no tooling/ changes — skipping tooling tests"
elif command -v make >/dev/null 2>&1; then
    echo "pre-push: running tooling tests (make test-tooling)..."
    if ! (cd "$REPO_ROOT" && make test-tooling); then
        fail_check "tooling tests (make test-tooling)"
    else
        echo "pre-push: tooling tests — OK"
    fi
else
    echo "pre-push: WARNING — make not found, skipping tooling tests"
fi

# --- JSON syntax validation ---
# Use the same REMOTE_REF the directory-change detection above settled on
# (origin/<branch> preferred, origin/main fallback). Without this, a first
# push of a new branch falls through to "validate every JSON in the repo,"
# which on this repo means 2762 Python parses — tens of seconds of churn
# for a push that might not have touched any JSON at all.
if [ -n "$REMOTE_REF" ]; then
    JSON_FILES=$(git diff --name-only "$REMOTE_REF"..HEAD -- '*.json' 2>/dev/null || true)
else
    JSON_FILES=$(git ls-files '*.json')
fi
if [ -n "$JSON_FILES" ]; then
    echo "pre-push: checking JSON syntax..."
    JSON_FAIL=0
    while IFS= read -r f; do
        if [ -f "$REPO_ROOT/$f" ] && ! python3 -m json.tool "$REPO_ROOT/$f" >/dev/null 2>&1; then
            echo "  FAIL: $f"
            JSON_FAIL=$((JSON_FAIL + 1))
        fi
    done <<< "$JSON_FILES"
    if [ "$JSON_FAIL" -gt 0 ]; then
        echo "pre-push: FAIL — $JSON_FAIL JSON file(s) have syntax errors"
        fail_check "JSON syntax"
    else
        echo "pre-push: JSON — OK ($(echo "$JSON_FILES" | wc -l) file(s))"
    fi
else
    echo "pre-push: no JSON changes — skipping"
fi

# --- systems.db stamp check (#857) ---
# If the branch touches server/data/systems.db and the meta stamp does not
# match current generator sources, reject the push.  Prevents pushing a
# stale DB snapshot where generator source was modified but the DB was not
# regenerated.
#
# Runs whenever systems.db was modified in ANY branch commit vs. main —
# including on a branch's very first push (review T5: the previous version
# skipped the check for new branches because it compared against origin/$BRANCH,
# which didn't exist yet, leaving a gap where a stale DB could ship via the
# first push).  We compare against origin/main — which always exists — so the
# check covers the first-push case.
DB_IN_PUSH=$(git diff --name-only origin/main...HEAD -- server/data/systems.db 2>/dev/null | wc -l)
if [ "$DB_IN_PUSH" -gt 0 ] && command -v reach >/dev/null 2>&1; then
    echo "pre-push: checking systems.db stamp..."
    rc=0
    reach --no-input check systems-db-stamp || rc=$?
    if [ "$rc" -eq 1 ]; then
        # rc=1 means stale / unknown generator / missing source; message on stderr
        echo "  Fix: run 'make regen-db' then stage server/data/systems.db"
        echo "  Or use /pr-push — it handles regen automatically before pushing."
        fail_check "systems.db stamp (stale)"
    elif [ "$rc" -eq 2 ]; then
        # rc=2 means no meta table — treat as unstamped, warn but don't block.
        # This is legitimate immediately after the meta table is introduced;
        # the next `make regen-db` will populate it (H4).
        echo "pre-push: WARNING — systems.db has no meta stamp — run 'make regen-db' now if this DB was generated by you"
    else
        echo "pre-push: systems.db stamp — OK"
    fi
else
    echo "pre-push: systems.db not in push — skipping stamp check"
fi

# --- Client version mirror (T-1241) ---
# project.yaml is the version source of truth, but the client cannot read it at
# runtime (an exported build has no repo root), so it is mirrored into
# client/project.godot's `application/config/version`. The Atlas disk cache keys
# its ONLY invalidation signal on that value, so a stale mirror makes a build
# serve canvases generated by code it no longer runs — the T-1239 failure.
#
# Deliberately UNCONDITIONAL, unlike the systems.db check above. Drift is
# introduced by touching one file without the other, but it PERSISTS on main
# until someone notices, so gating on "was either file in this push" would let
# an existing drift ride along indefinitely. The check is two file reads.
if command -v reach >/dev/null 2>&1; then
    echo "pre-push: checking client version mirror..."
    if reach --no-input check client-version; then
        :
    else
        fail_check "client version mirror (drifted from project.yaml)"
    fi
fi

# --- Canvas-generation / version pairing (T-1242) ---
# A change to how a canvas is GENERATED is half a change; the other half is
# bumping project.yaml's version, or every warm Atlas cache keeps serving
# canvases built by code that no longer exists. That pairing broke five times
# (0.4.2 through 0.4.6), every one bumped after the fact, the last costing the
# eight-day T-1239 diagnosis. The path registry is tooling/canvas_sources.py.
#
# Self-check first: the registry fails closed on an empty glob, and a registry
# that cannot load must not be read as "nothing to enforce".
if command -v reach >/dev/null 2>&1; then
    echo "pre-push: checking canvas-generation version pairing..."
    if reach --no-input check canvas-version; then
        :
    else
        fail_check "canvas generation changed without a project.yaml version bump"
    fi
fi

# --- Data-flow diagram paths (D-262) ---
# A diagram that names files goes stale SILENTLY — nothing fails when a path
# moves, so the map keeps asserting a layout that is no longer true. This checks
# only that the paths still resolve; whether an EDGE still means what it says is
# a human check against the tool's source. Cheap (no subprocess beyond python).
if command -v reach >/dev/null 2>&1; then
    echo "pre-push: checking data-flow diagram paths..."
    if reach --no-input check dataflow-graph; then
        :
    else
        fail_check "a path named in a data-flow diagram no longer resolves"
    fi
fi

# --- Clerk review (D-221) ---
# DISABLED 2026-05-23 (#965) pending rework. Two problems made it net-negative:
#   1. Non-exhaustive — a single run reports ~the first contradiction it finds
#      and stops, so distinct real issues surfaced only on the 2nd/3rd/4th
#      re-push. An APPROVED verdict therefore can't be trusted as "clean".
#   2. Re-reviews the entire origin/main..HEAD range on every push (token burn);
#      no per-commit verdict cache.
# Off by default until reworked (exhaustive enumeration + holistic decisions/
# diff check + a SHA-keyed verdict cache invalidated on decisions/ change, with
# skip tracked in the cache rather than a commit-message trailer — see #965).
# Force-run with SR_RUN_CLERK=1.
if [ -x "$REPO_ROOT/tooling/clerk-review" ] && [ "${SR_RUN_CLERK:-0}" = "1" ]; then
    echo "pre-push: running clerk review..."
    CLERK_VERDICT=$("$REPO_ROOT/tooling/clerk-review" 2>&1 | tee /dev/stderr | tail -1)
    if [ "$CLERK_VERDICT" = "APPROVED" ]; then
        echo "pre-push: clerk — APPROVED"
    elif [ "$CLERK_VERDICT" = "REJECTED" ]; then
        echo "pre-push: clerk — REJECTED (see .cache/pre-push-review.md)"
        fail_check "clerk review (REJECTED)"
    elif [ "$CLERK_VERDICT" = "INCOMPLETE" ]; then
        echo "pre-push: clerk — INCOMPLETE (some reviews didn't finish; NOT blocking)"
        echo "  See .cache/pre-push-review.md. For a full verdict: raise SR_CLERK_MAX_TURNS / SR_CLERK_TIMEOUT,"
        echo "  or add a 'Clerk-Skip:' trailer to bulk-content commits."
    else
        echo "pre-push: clerk — '$CLERK_VERDICT' unrecognized; treating as block (see .cache/pre-push-review.md)"
        fail_check "clerk review (unrecognized verdict)"
    fi
else
    echo "pre-push: clerk review — DISABLED (#965; force-run with SR_RUN_CLERK=1)"
fi

if [ "$ERRORS" -gt 0 ]; then
    echo ""
    echo "=============================================================="
    echo "pre-push: PUSH ABORTED — $ERRORS check(s) failed:"
    printf '%s' "$FAILED_CHECKS"
    echo ""
    echo "  Search the log above for the failing check's own output, e.g."
    echo "    grep -A20 'checking Rust (fmt)' <logfile>"
    echo "  cargo fmt is auto-fixable: cd server && cargo fmt"
    echo "=============================================================="
    exit 1
fi

echo "pre-push: all checks passed."