# SPDX-FileCopyrightText: Copyright Hewlett Packard Enterprise Development LP

# SPDX-License-Identifier: MIT

#

# PiG's command interface. Developers, CI, and coding agents run make targets

# instead of ad hoc shell commands. `make help` lists the targets by group.

#

# Every script behind these targets lives in automation/ (see its README):

# dev/ for setup and local tools, ci/ for gates and test runners, gen/ for

# generators and the upstream mirror, release/ for release checks, porter/ for

# Pig Porter, images/ for CI images, and make/ for the parity targets included

# at the end of this file.

#

# Reward function: `make coverage` should print three numbers going up

# (parity_score, runtime_score, sdk_score) and zero going down (untested,

# orphans, divergences).

.DEFAULT_GOAL := help

# Caches follow XDG_CACHE_HOME; temporary files follow TMPDIR. Sandboxes,

# devcontainers, and hosted runners often forbid writes elsewhere.

PIG_CACHE_HOME ?= $(or $(XDG_CACHE_HOME),$(HOME)/.cache)

# golangci-lint caches absolute package paths. Isolate each checkout so parallel worktrees retain warm results without serving findings from another tree.

GOLANGCI_LINT_CACHE ?= $(PIG_CACHE_HOME)/golangci-lint/$(notdir $(CURDIR))

export GOLANGCI_LINT_CACHE

PIG_TMP ?= $(patsubst %/,%,$(or $(TMPDIR),/tmp))

# macOS mktemp ignores TMPDIR without a template, so recipes always pass one.

MKTEMP = mktemp "$(PIG_TMP)/pig.XXXXXXXX"

MKTEMP_DIR = mktemp -d "$(PIG_TMP)/pig.XXXXXXXX"

PIG_DEV_HOME ?= $(PIG_CACHE_HOME)/pig-dev

# Machine-local toolchain and module-proxy settings written by `make setup`.

# Absent on machines that need neither.

-include $(PIG_DEV_HOME)/env.mk

# Development gates select the local SDK through go.work, regardless of env.mk. Standalone-module and publication recipes set GOWORK=off on their own commands.

ifeq ($(GOWORK),off)

unexport GOWORK

endif

# Workspace mode accepts only -mod=readonly or -mod=vendor.

ifneq ($(filter -mod=mod,$(GOFLAGS)),)

export GOFLAGS := $(strip $(filter-out -mod=mod,$(GOFLAGS)))

endif

PIG_BIN ?= $(HOME)/.local/bin/pig

# Gate-only pig binary. It is built through go.work to exercise the developing SDK, then runs with GOWORK=off so the extensions it validates build as standalone modules. module-publication separately verifies the published dependency pins.

CHECK_PIG_BIN := $(CURDIR)/tmp/check-bin/pig

# Default parity to a freshly built in-tree binary. If callers explicitly set

# PIG_BIN in the environment/command line, honor it for compatibility with the

# older parity workflow.

PARITY_PIG_BIN ?= $(if $(filter command line environment,$(origin PIG_BIN)),$(PIG_BIN),$(CURDIR)/bin/pig-parity)

RESULTS ?= $(PIG_TMP)/parity-results.json

EXTENSION_CORPUS_RESULTS ?= $(PIG_TMP)/pig-typescript-extension-corpus.json

PARITY_TIMEOUT ?= 20m

PARITY_STRESS_TIMEOUT ?= 30m

PARITY_DURABLE_TIMEOUT ?= 45m

# Probed on first use only, so targets that never schedule parity skip the probe.

PARITY_RESOURCE_LIMITS = $(eval PARITY_RESOURCE_LIMITS := $$(shell ./automation/ci/parity-resource-limits.sh))$(PARITY_RESOURCE_LIMITS)

PARITY_PARALLEL ?= $(word 1,$(PARITY_RESOURCE_LIMITS))

PARITY_GROUP_LIMITS ?= $(word 2,$(PARITY_RESOURCE_LIMITS))

UPSTREAM_VERSION := $(shell awk -F'"' '/^const UpstreamVersion = "/ { print $$2; exit }' internal/coding/pigversion/pigversion.go)

UPSTREAM_REVIEWED_VERSION := $(shell awk -F'"' '/^const UpstreamReviewedVersion = "/ { print $$2; exit }' coding/upstream.go)

PI_PACKAGE_ROOT := $(CURDIR)/extensions/sdk-ts/node_modules/@earendil-works/pi-coding-agent

PIG_PARITY_PI_BIN := $(CURDIR)/extensions/sdk-ts/node_modules/.bin/pi

CARGO_TARGET_DIR ?= $(CURDIR)/tmp/test-fixtures/rust-target

export PI_PACKAGE_ROOT PIG_PARITY_PI_BIN CARGO_TARGET_DIR

SETUP_ARGS ?=

COMPLIANCE_ARGS ?=

HARNESSES ?= pig,pi

RUNS ?= 10

MODEL ?=

EVAL_RESULTS ?= tmp/evals/overhead.json

LIVE_RUNS ?= 3

EVAL_TASKS_DIR ?=

MUTATIONS ?= 30

SEED ?= 1

PROFILE ?= cpu,heap

BENCH ?= .

BENCH_PKGS ?= ./agent/... ./tui/... ./internal/codingagent/...

BENCH_COUNT ?= 6

SLOP_MAX_VERBOSITY = $(shell sed -n 's/^max_verbosity *= *//p' test/evals/budgets.toml)

SLOP_MAX_EROSION = $(shell sed -n 's/^max_erosion *= *//p' test/evals/budgets.toml)

PIGEVAL := PYTHONPATH=$(CURDIR)/test/evals python3 -m pigeval

HELP_AWK := BEGIN {FS = ":.*\#\# "} /^\#\#@ / {printf "\n%s\n", substr($$0, 5); next} /^[a-zA-Z0-9_%-]+:.*\#\# / {printf " %-24s %s\n", $$1, $$2}

##@ Start here

setup: ## Install pinned Go, a module-proxy fallback when needed, and bin/pig (SETUP_ARGS=--all adds Node and eval harnesses)

@./automation/dev/setup.sh $(SETUP_ARGS)

doctor: ## Report every development prerequisite without changing anything

@./automation/dev/setup.sh --check $(SETUP_ARGS)

help: ## List targets by group

@awk '$(HELP_AWK)' Makefile

@echo

@echo "Parity and porting targets: make help-parity. Scripts: automation/README.md."

help-parity: ## List the parity and porting targets (automation/make/parity.mk)

@awk '$(HELP_AWK)' automation/make/parity.mk

##@ Release maintenance

.PHONY: set-version module-publication

set-version: ## Update the release identity; preserve dependencies unless SET_VERSION_ARGS=--release-modules

@./automation/ci/with-isolated-pig-home.sh go run ./automation/release/set-version -version "$(VERSION)" $(SET_VERSION_ARGS)

module-publication: ## Verify nested Go dependency tags and checksums against their public source

@./automation/ci/with-isolated-pig-home.sh python3 -B -m unittest automation/ci/test_module_publication.py

@python3 automation/ci/check-module-publication.py

##@ Build

pig: node-runtime ## Build bin/pig from this checkout, with symbols for profiling

@mkdir -p bin

CGO_ENABLED=0 go build -buildvcs=false -trimpath -ldflags "-X main.Build=$$(git rev-parse --short HEAD 2>/dev/null || echo dev)" -o bin/pig ./cmd/pig

clean: ## Remove bin/ and tmp/ (builds, eval results, profiles, benchmarks)

rm -rf bin tmp

node-runtime: ## Regenerate the embedded Node runtime archive and content digest

go generate ./coding/extension/host/subprocess

build: node-runtime ## Compile every package (go build ./...)

go build -buildvcs=false ./...

# install: build pig binary → ~/.local/bin/pig (or PIG_BIN override)

# with embedded git sha + macOS ad-hoc codesign.

install: node-runtime ## Install a stripped pig to PIG_BIN (default ~/.local/bin/pig)

@mkdir -p $(dir $(PIG_BIN))

CGO_ENABLED=0 go build -buildvcs=false -trimpath -ldflags "-s -w -X main.Build=$$(git rev-parse --short HEAD 2>/dev/null || echo dev)" -o $(PIG_BIN) ./cmd/pig

@[ "$$(uname -s)" = "Darwin" ] && command -v codesign >/dev/null 2>&1 && codesign --force --sign - $(PIG_BIN) >/dev/null 2>&1 || true

@echo "installed: $$($(PIG_BIN) --version) → $(PIG_BIN)"

##@ Generated files

# Keep producer/consumer order even under make -j. Reviewed mappings and evidence ledgers are inputs, not generated dispositions. Coverage deliberately ignores machine-local parity results so every checkout writes the same bytes.

generate: export LC_ALL := C

generate help-text: export FORCE_COLOR := 0

generate: ## Regenerate committed inventories, coverage, catalogs, help, and docs mirrors

@$(MAKE) model-catalogs

@$(MAKE) node-runtime

@$(MAKE) highlight-grammars

@$(MAKE) help-text

@$(MAKE) knowledge-graph

@$(MAKE) interface-proposals

@$(MAKE) test-inventory-generate

@$(MAKE) known-gaps

@$(MAKE) custom-factory-ledger

@$(MAKE) coverage RESULTS=

@echo "Generated files refreshed. Review and commit the diff; CI compares these files."

help-text: parity-deps ## Regenerate the pinned upstream CLI help with PiG's identity

@./automation/gen/gen-help.sh

highlight-grammars: parity-deps ## Snapshot Pi's highlight.js grammars into tui/internal/hljs and re-record Pi's highlighting oracle

@node automation/gen/generate-highlight-grammars.mjs "$(PI_PACKAGE_ROOT)" tui/internal/hljs

@node automation/gen/generate-highlight-oracle.mjs "$(PI_PACKAGE_ROOT)" tui/internal/hljs/testdata/corpus.json tui/internal/hljs/testdata/oracle.json.gz

.PHONY: generate help-text highlight-grammars

##@ Test and lint

vet: ## Run go vet on every package, natively and for every shipped Windows architecture and for Android (Termux)

go vet ./...

GOOS=windows GOARCH=amd64 go vet ./...

GOOS=windows GOARCH=arm64 go vet ./...

CGO_ENABLED=0 GOOS=android GOARCH=arm64 go vet ./...

LINT_BASE ?= main

lint: ## Run golangci-lint over the repository (part of make check)

@echo "Running golangci-lint..."

go tool golangci-lint run --allow-parallel-runners --build-tags=integration,live,parity ./...

# Like go list ./..., exclude testdata directories, including nested fixture modules.

lint-changed: ## Lint whole changed Go files in changed packages against LINT_BASE (default main)

@base=$$(git merge-base HEAD "$(LINT_BASE)") || { echo "lint-changed: cannot find merge base with $(LINT_BASE)" >&2; exit 2; }; \

changed_go=$$({ git diff --name-only "$$base"...HEAD; git diff --name-only; git diff --name-only --cached; git ls-files --others --exclude-standard; } | awk '/\.go$$/ && !/(^|\/)testdata\//' | sort -u); \

if [ -z "$$changed_go" ]; then echo "lint-changed: no changed Go files"; exit 0; fi; \

packages=$$(printf '%s\n' "$$changed_go" | while IFS= read -r file; do dir=$${file%/*}; [ "$$dir" = "$$file" ] && dir=.; [ -d "$$dir" ] && printf './%s\n' "$$dir"; done | sort -u); \

if [ -z "$$packages" ]; then echo "lint-changed: no changed Go packages"; exit 0; fi; \

echo "Running golangci-lint on changed packages: $$(printf '%s' "$$packages" | tr '\n' ' ')"; \

go tool golangci-lint run --allow-parallel-runners --build-tags=integration,live,parity --new-from-rev="$$base" --whole-files $$packages

# Fast inner-loop gate (~90s). Build + vet + focused unit tests + QC smoke.

# Use between edits. `make check` is the pre-commit gate.

dev: build vet qc ## Fast inner loop: build, vet, and the QC smoke (about 90 s)

test: test-prereqs interface-deps parity-deps ## Grouped Go tests with locked interface tooling, fixtures, and safe concurrency

@./automation/ci/test-grouped.sh

# Every maintained nested Go module must pass without workspace assistance.

# Fixture modules are included because they exercise the same clean-clone

# dependency boundaries as independently distributed modules.

test-sdk-go: ## Test the Go extension SDK module on its own

@cd extensions/sdk && GOWORK=off go test ./... -count=1

test-go-modules: test-sdk-go ## Test every maintained nested Go module with GOWORK=off

@set -eu; \

for module in \

cmd/pig/testdata/login-preview \

coding/extension/host/subprocess/testdata/sdk-fixture \

examples/extensions/go-factory \

piglets/porter/extensions/pig-porter \

piglets/standard; do \

echo "testing Go module: $$module"; \

(cd "$$module" && GOWORK=off go test ./... -count=1); \

done

# The Rust SDK's own unit tests. `test` builds the Rust conformance fixtures,

# which compiles the SDK library but not its test targets, so a test-only break

# such as a struct literal missing a newly added field compiles clean and ships.

# That is exactly how ReadyMsg.state reached main red.

test-sdk-rs: ## Run the Rust extension SDK's unit tests

@cd extensions/sdk-rs && cargo test --quiet

test-sdk-ts: parity-deps ## Check the pinned TypeScript extension declarations and runtime helpers

@cd extensions/sdk-ts && npm test

# Resolve the selected Rust toolchain before validators replace HOME so tool-manager shims do not select a different compiler.

examples-check: test-prereqs parity-bin ## Build and validate every example extension

@set -eu; tmp=$$($(MKTEMP_DIR)); trap 'rm -rf "$$tmp"' EXIT; \

rust_root=$$(rustc --print sysroot); \

case "$${OSTYPE:-}" in msys*|cygwin*) rust_root=$$(cygpath -u "$$rust_root");; esac; \

export PATH="$$rust_root/bin:$$PATH"; \

cd examples/extensions/rust-factory && cargo build --release --quiet; \

cd "$(CURDIR)"; \

for extension in \

examples/extensions/go-factory \

examples/extensions/rust-factory \

examples/extensions/python-factory; do \

name=$$(basename "$$extension"); \

echo "validating example extension: $$extension"; \

./automation/ci/with-isolated-pig-home.sh "$(PARITY_PIG_BIN)" install --validate-only --json "$$extension" >"$$tmp/$$name.json"; \

python3 -c 'import json,sys; data=json.load(open(sys.argv[1])); assert data["valid"], data; assert not data.get("diagnostics"), data.get("diagnostics")' "$$tmp/$$name.json"; \

done

test-prereqs: ## Require every toolchain the core suite exercises

@missing=""; \

for tool in go git node npm python3 cargo rustc tmux; do \

command -v "$$tool" >/dev/null 2>&1 || missing="$$missing $$tool"; \

done; \

if [ -n "$$missing" ]; then \

echo "missing required Pig test toolchains:$$missing" >&2; \

exit 1; \

fi

@echo "test prerequisites: go, git, node, npm, python3, cargo, rustc, tmux"

test-fixtures: ## Build the Go and Rust extension fixtures that tests load

@./automation/ci/test-fixtures.sh

test-stress: ## Stress the grouped Go test scheduler under higher contention

@./automation/ci/test-grouped.sh stress

# Deterministic tmux-driven integration tier: real pig in a real pty, faux

# provider, no network. Carries the signal guards, which cover shutdown paths

# no unit test reaches (extension cleanup on SIGTERM, SIGINT aborting the

# operation rather than the session). Tests needing external extension sources

# skip themselves.

test-integration: ## Deterministic tmux tier: real pig, faux provider, no network

@./automation/ci/integration-tests.sh fast

# Race + goroutineleak gate for the interactive TUI concurrency model. Locks in

# the single-main-loop ownership invariant: a future off-loop UI touch fails

# here instead of shipping a latent data race. See automation/ci/test-race.sh.

test-race: ## Race and goroutine-leak gate for the TUI concurrency model

@./automation/ci/test-race.sh

# go-fix-clean enforces AGENTS.md rule #3 (Go 1.27 idiom compliance).

# `go fix -diff ./...` must produce zero output. If this target fails,

# run `go fix ./...`, manually validate any suggestions that would

# change observable behavior (e.g. type changes like

# `slices.Contains` returning bool vs a string flag), then commit.

go-fix-clean: ## Fail when go fix would change any file

@# Judge the diff on stdout; stderr (tool warnings) passes through. go fix

@# exits non-zero when it has a diff, so a failure without one is an error.

@status=0; diff_output=$$(go fix -diff ./...) || status=$$?; \

if [ -n "$$diff_output" ]; then \

echo "FAIL: go fix -diff produced output (AGENTS.md rule #3)."; \

echo "Run \`go fix ./...\` and validate suggestions before committing."; \

echo "First 40 lines of diff:"; \

echo "$$diff_output" | head -40; \

exit 1; \

fi; \

if [ "$$status" -ne 0 ]; then echo "FAIL: go fix -diff exited $$status without a diff."; exit 1; fi

@echo "go fix: clean (Go 1.27 idioms compliant)"

# compliance checks the evidence behind the README badges and the OpenSSF

# Best Practices criteria (docs/project/compliance.md): required files,

# CITATION.cff, SHA-pinned actions, read-only workflow tokens, one source for

# every version pin, reuse lint, and govulncheck. It fails on any miss.

compliance: ## Check badge and OpenSSF evidence: files, pins, workflow hardening, REUSE, govulncheck

@./automation/ci/check-compliance.py $(COMPLIANCE_ARGS)

# divergence-consistency enforces AGENTS.md rule #4. Every divergence

# numbered in docs/parity/DIVERGENCES.md must have a matching `// pig divergence (DN):`

# call-site comment in a .go file, and vice versa. Catches silent drift.

divergence-consistency: ## Match every docs/parity/DIVERGENCES.md entry to its call-site marker

@./automation/ci/check-divergence-consistency.sh

divergence-quality: ## Validate divergence records as enforceable contracts

@./automation/ci/check-divergence-quality.py

# divergence-guard fails on hidden divergences no docs/parity/DIVERGENCES.md entry

# records: invented limits and timeouts, dropped events, swallowed errors,

# success on an unknown stop reason. Current hits are ratcheted in

# automation/ci/divguard/baseline.toml, which may only shrink.

divergence-guard: ## Fail on unrecorded invented limits, dropped events and swallowed errors (ratchet)

@go run ./automation/ci/divguard

# docs-drift gates the shipped docs bundle against the code it describes: every

# documented command must exist, every listable command must be documented, and

# every page must be reachable from the index.

docs-drift: ## Fail when generated documentation no longer matches its source

@python3 automation/gen/gen-knowledge-graph.py --check

@go test ./test/docs-drift/ -count=1

npm-dist-test: ## Unit-test the npm package generator and launcher

@python3 -m unittest automation/release/npm/test_pack_npm.py

@node --test automation/release/npm/launcher.test.js

npm-dist-e2e: ## Cross-build, npm pack, install and run pig through the npm launcher

@automation/release/npm/e2e-local.sh

standard-check: ## Verify PiG Standard requires and resolves only fused extensions

@go test ./cmd/pig -run '^TestPiGStandardRequiresAndResolvesOnlyFusedExtensions$$' -count=1

@go test ./coding/pigletbuild -run '^TestRunBuildRejectsRequiredFusedFallback$$' -count=1

@go test ./coding/extension/host/subprocess -run '^TestHost_MixedFusedPackedAndIsolatedExtensions$$' -count=1

@go build -buildvcs=false -o $(CHECK_PIG_BIN) ./cmd/pig

@GOWORK=off ./automation/ci/with-isolated-pig-home.sh "$(CHECK_PIG_BIN)" install --validate-only --json piglets/standard/extensions/piglogin >/dev/null

@GOWORK=off ./automation/ci/with-isolated-pig-home.sh "$(CHECK_PIG_BIN)" install --validate-only --json piglets/standard/extensions/pigrunner >/dev/null

@GOWORK=off ./automation/ci/with-isolated-pig-home.sh "$(CHECK_PIG_BIN)" install --validate-only --json piglets/standard/extensions/angrypigs >/dev/null

@GOWORK=off ./automation/ci/with-isolated-pig-home.sh "$(CHECK_PIG_BIN)" piglet validate piglets/standard/pig-standard.yaml >/dev/null

##@ Gates

check-core: startup-proxies build vet lint test test-go-modules test-sdk-rs test-sdk-ts typescript-extension-corpus examples-check test-race test-integration lint-scenarios port-map-drift coverage-drift standard-check go-fix-clean divergence-consistency divergence-quality divergence-guard check-contracts-fast source-hygiene docs-drift ## All deterministic build, lint, unit, race, and drift gates

check: module-publication check-core parity-fast ## Pre-commit gate: published dependencies, check-core and parity-fast

@echo

@echo "make check: all gates green."

@echo

@echo "Reminder: run 'make coverage' before signaling success to refresh"

@echo "the AGENTS.md dashboard and test/parity/coverage.md. Do not hand-edit."

# verify = deterministic gates + declared durability parity + coverage.

# The recommended one-shot for end-of-loop verification: `parity` proves every

# declared pair and regenerates the dashboard after the core gates pass.

verify: check-core check-contracts parity ## End-of-loop gate: check-core, contracts, and full parity

@echo

@echo "make verify: dashboard refreshed deterministically from"

@echo "docs/parity/PORT_MAP.md + test/parity/scenarios. Commit the diff if any."

qc: parity-bin ## Build a fresh pig, then run the smoke and focused parity QC

@PIG_BIN="$(PARITY_PIG_BIN)" ./automation/ci/qc-smoke.sh

# Release gate. Tighter than `check`: parity-live must pass, coverage

# must not regress, no new entries in docs/parity/DIVERGENCES.md without justification.

release-check: parity-live parity-perf ## Release gate: live and perf parity, no new divergences, no coverage loss

@./automation/ci/check-divergence-delta.sh

@./automation/ci/check-coverage-delta.sh

startup-proxies: ## Run deterministic startup allocation, cache, encoding, and artifact guards

@automation/ci/with-isolated-pig-home.sh python3 automation/perf/startup_proxies.py

@automation/ci/with-isolated-pig-home.sh python3 -m unittest discover -s automation/perf -p 'test_*.py'

.PHONY: startup-proxies

##@ Evals and performance

evals: pig ## Measure harness overhead against the mock model (HARNESSES=all for every installed harness)

@$(PIGEVAL) overhead --harnesses $(HARNESSES) --runs $(RUNS) --pig bin/pig --out $(EVAL_RESULTS)

@$(PIGEVAL) report $(EVAL_RESULTS) --out $(EVAL_RESULTS:.json=.md)

evals-requests: pig ## Save every request body and diff the first two HARNESSES (default: pig against pi)

@$(PIGEVAL) requests --harnesses $(HARNESSES) --pig bin/pig --diff

evals-live: pig ## Run test/evals/tasks with a real model: make evals-live MODEL=provider/model

@test -n "$(MODEL)" || { echo "evals-live: set MODEL, for example MODEL=anthropic/claude-sonnet-5" >&2; exit 2; }

@$(PIGEVAL) live --harnesses $(HARNESSES) --pig bin/pig --model "$(MODEL)" --runs $(LIVE_RUNS) $(if $(EVAL_TASKS_DIR),--tasks-dir $(EVAL_TASKS_DIR)) --out tmp/evals/live.json

@$(PIGEVAL) report tmp/evals/live.json --out tmp/evals/live.md

evals-mutate: ## Generate seeded bug-fix tasks from real Go files (MUTATIONS=30 SEED=1) into tmp/evals/mutation-tasks

@$(PIGEVAL) mutate --count $(MUTATIONS) --seed $(SEED) --out tmp/evals/mutation-tasks

evals-publish: ## Copy a reviewed EVAL_RESULTS run to test/evals/results/latest.json and regenerate docs/site/docs/evals.md

@test -f "$(EVAL_RESULTS)" || { echo "evals-publish: $(EVAL_RESULTS) not found; run make evals first" >&2; exit 2; }

@cp "$(EVAL_RESULTS)" test/evals/results/latest.json

@$(PIGEVAL) report test/evals/results/latest.json --out docs/site/docs/evals.md

evals-test: ## Unit-test pigeval, the eval harness

@PYTHONPATH=$(CURDIR)/test/evals python3 -m unittest discover -s test/evals/tests -q

perf-check: ## Apply test/evals/budgets.toml (latency, memory, and request-size ceilings) to EVAL_RESULTS

@$(PIGEVAL) check $(EVAL_RESULTS)

profile: pig ## Profile one prompt round trip (PROFILE=cpu,heap,allocs,block,mutex,trace) into tmp/profile

@$(PIGEVAL) profile --pig bin/pig --kinds $(PROFILE) --out tmp/profile

pgo: pig ## Merge CPU profiles from repeated round trips into tmp/pgo/default.pgo for a profile-guided build

@$(PIGEVAL) profile --pig bin/pig --kinds cpu --runs 20 --out tmp/pgo --merge tmp/pgo/default.pgo

slop: ## Measure erosion and clone verbosity (SlopCodeBench metrics) and list the heaviest functions

@go run ./automation/ci/slopmetrics -top 15

slop-check: ## Fail when erosion or verbosity exceed the ratchet in test/evals/budgets.toml [slop]

@go run ./automation/ci/slopmetrics -top 0 -max-verbosity $(SLOP_MAX_VERBOSITY) -max-erosion $(SLOP_MAX_EROSION) >/dev/null && echo "slop-check: within verbosity $(SLOP_MAX_VERBOSITY), erosion $(SLOP_MAX_EROSION)"

bench: ## Run Go benchmarks into tmp/bench/current.txt (BENCH=regex, BENCH_PKGS, BENCH_COUNT=6)

@mkdir -p tmp/bench

@go test -run '^$$' -bench '$(BENCH)' -benchmem -count $(BENCH_COUNT) $(BENCH_PKGS) > tmp/bench/current.txt; status=$$?; cat tmp/bench/current.txt; exit $$status

bench-base: ## Keep the last benchmark run as the base for bench-compare

@cp tmp/bench/current.txt tmp/bench/base.txt

bench-compare: ## Compare the last benchmark run with the base by median; fails above a 5% regression

@$(PIGEVAL) benchcmp tmp/bench/base.txt tmp/bench/current.txt --fail

##@ Docs

knowledge-graph: ## Regenerate the knowledge graph page, JSON-LD, and Mermaid from docs/knowledge-graph/pig.graph.json

@python3 automation/gen/gen-knowledge-graph.py

##@ Upstream and catalogs

upstream-mirror: parity-deps ## Mirror the pinned Pi source into .upstream/current

@./automation/gen/mirror-upstream.sh

model-catalogs: parity-deps ## Regenerate model catalogs from the pinned published Pi package

@./automation/gen/generate-model-catalogs.sh

##@ Dependencies

interface-deps: ## Install the locked TypeScript compiler used by parity inventories

@python3 automation/ci/npm-locked.py test/parity/interface-extractor

parity-deps: interface-deps ## Install the exact locked Pi comparator and the TypeScript compiler its scenarios import

@python3 automation/ci/npm-locked.py extensions/sdk-ts

@test -x "$(PIG_PARITY_PI_BIN)" || { echo "exact Pi comparator not found: $(PIG_PARITY_PI_BIN)" >&2; exit 1; }

@test -d "$(PI_PACKAGE_ROOT)" || { echo "exact published Pi package not found: $(PI_PACKAGE_ROOT)" >&2; exit 1; }

@test "$$(env -u HTTPS_PROXY -u HTTP_PROXY -u ALL_PROXY -u https_proxy -u http_proxy -u all_proxy "$(PIG_PARITY_PI_BIN)" --version)" = "$(UPSTREAM_VERSION)" || { echo "Pi comparator version does not match $(UPSTREAM_VERSION)" >&2; exit 1; }

.PHONY: node-runtime npm-dist-test npm-dist-e2e compliance evals-mutate slop slop-check # Deterministic tmux-driven integration tier # Release gate. Tighter than `check` # The recommended one-shot for end-of-loop verification # docs-drift gates the shipped docs bundle against the code it describes # install # numbered in docs/parity/DIVERGENCES.md must have a matching `// pig divergence (DN) # the single-main-loop ownership invariant bench bench-base bench-compare build check check-core clean dev divergence-guard divergence-quality doctor evals evals-live evals-publish evals-requests evals-test examples-check go-fix-clean help help-parity interface-deps knowledge-graph lint lint-changed model-catalogs parity-deps perf-check pgo pig profile qc setup standard-check test test-fixtures test-go-modules test-prereqs test-sdk-go test-sdk-rs test-sdk-ts test-stress upstream-mirror vet

include automation/make/parity.mk

include automation/make/ci.mk