make toolchain | Check the installed toolchains against pins.toml |
make fetch | Network: everything every target needs: crates, Go modules, Python packages, the stellar-core package and goldens, container images, driver packages and pipeline sources, all pinned and hash-verified (scripts/fetch.sh all). FETCH_ONLY="PART..." fetches only the named parts (scripts/fetch.sh lists them: deps, core, conform-pr, drivers, driver:DIR, wallet-backend, lint-sources, lint-sources-wb, docs, …); each CI job names the parts its target needs |
make demo | The one-minute demo: replays an indexer with one planted bug (go/adapters/demo) from a plain directory and prints its failures with repro commands; exits 0 when the bug is found |
make fetch-replay | Network, for indexer authors: the build dependencies (Python packages, crates, Go modules) and the pinned SeaweedFS image the replay harness serves the sep54-s3 input from |
make fetch-sdk | Network, for SDK maintainers: the build dependencies and every public conformance column’s driver packages, stellar-core’s own decoder among them (xdrcheck conform fetch --set public) |
make fetch-lint | Network, for xdrlint users: the build dependencies (xdrlint’s Go modules among them) and golangci-lint at the pinned version with the xdrlint module plugin |
make build | Build xdrcheck-corpus, xdrcheck-mutate, xdrcheck, the adapters and xdrlint, link them into ./bin, and print the export PATH=… line |
make release | scripts/release.sh all: build the release archive twice and compare, verify it offline from an empty directory, check the stable/-next channel split, pack and check the npm package, and write dist/SHA256SUMS and the release provenance (release.md); VERSION=<MAJOR+1>.0.0-rc.N builds the -next channel (refused while the corpus has no next-protocol entry) |
make repo-checks | Repository checks that need no build: schemas and their examples, corpus data, entry IDs, the field table source and guide claims, the public-tree scan and its history self-test, the documentation site’s generated pages (apps/docs/scripts/generate.py --check), workflow rules, the results data (results-check), the corpus entries the guide cites (check_guide_entries.py), the mainnet tier’s files (mainnet-tier-check), and, when xdrcheck-corpus is built, the guide’s field table, the coverage report with the golden mirror listing and the corpus index (--check of each) |
make ci-local | scripts/ci/run-local.sh: the jobs of .github/workflows/pr.yml (TIER=pr) or full.yml (TIER=full), or one job and the jobs it needs (JOB=<id>), run as on a cold hosted runner: a fresh clone of SHA (default HEAD) per job, an empty home with only the job’s declared fetch parts and artifacts, 2 CPUs and 7 GiB per step, the workflow’s run lines as written; writes summary.md (result, wall time, fetch time, peak disk per job) under CI_LOCAL_OUT (network for the fetch steps) |
make verify-static | repo-checks, then xdrcheck-corpus verify --checks schema,hashes,ids and xdrcheck pins check |
make verify-corpus | xdrcheck-corpus verify --tier curated --report reports/corpus-verify.json |
make verify-generate | xdrcheck-corpus gen --check --list corpus/scenarios/pr.list, always without network |
make test | Unit tests of every component: test-rust, test-go, test-drivers and test-cli; the full CI’s unit jobs |
make test-rust | cargo test --locked --workspace |
make test-go | go test in go/ and lint/ |
make test-cli | scripts/py/check_cli_reference.py: this page against the built tools’ help and make help |
make test-fast | The pull-request unit tier: test-rust-fast, test-go-fast, test-drivers and test-cli |
make test-rust-fast | Every test test-rust runs, the test binaries side by side (scripts/cargo-test-parallel.sh): the same build, every binary from its package directory, the documentation tests through cargo test --doc; nothing skipped or filtered |
make test-go-fast | test-go with go test -short: a test that checks testing.Short() skips (today only TestTemplateBuildsInItsOwnModule, which builds the minimal adapter’s template as a module of its own) |
make check-adapter-ts | The TypeScript minimal adapter (examples/adapter-ts): installs its pinned dependency, runs its selftest and a full replay |
make harness-selftest | Every seeded-bug indexer is replayed and must land in its documented class |
make verify-memory | The replay harness’s memory-bound tests and their negative control |
make verify-replay | The reference adapter on every pinned pre-fix and fixed wallet-backend commit, and on the current one |
make verify-replay-pr | The pull-request CI’s form of verify-replay: the same script and row-by-row assertions on one input, REPLAY_PR_INPUT (default rpc, the cheaper one); the full CI runs both inputs |
make replay-full | scripts/replay-full.sh: the whole curated corpus against wallet-backend at the pinned current commit, on the sep54-s3 and rpc inputs, into reports/replay-full/wallet-backend-<input>/; fails on a harness fault, an entry without a row or anything left behind (its unexpected rows are judged by the findings check); not part of verify, about 8 minutes |
make replay-etl-known | scripts/replay-etl-known.sh: the corpus entries that declare a stellar-etl replay divergence, against stellar-etl at its pinned commit (pins.toml), with --assert-known, into reports/replay-etl-known/stellar-etl-<commit>/; fails when the run cannot be made, a row is not of the pinned commit or an entry does not show its declared result; a run of the findings check’s required set, not part of verify |
make verify-conformance-js | The pull-request CI’s form of the conform sdk-regression assertions: xdrcheck conform build of the three JavaScript columns the assertions use, then xdrcheck conform sdk-regression, into reports/conformance/sdk-regression.json; the pr-set matrix around it runs in the full CI (verify-conformance) |
make verify-conformance-public | xdrcheck conform build, run and check on the public column set into reports/conformance-public/; not part of verify; verify-acceptance runs it before verify-findings-acceptance |
make sdk-loaders | The corpus loaders an SDK’s own tests use (packages/go, packages/rust, packages/python, packages/npm), each tested against this corpus (use-in-sdk-ci.md §1) |
make sdk-ci-examples | The Go, Rust and Python SDK CI recipes (examples/sdk-ci/<lang>) against the pinned go-stellar-sdk, stellar-xdr and stellar-sdk, offline from the caches make fetch-sdk fills; each must equal its committed expected-failure list (use-in-sdk-ci.md §2) |
make sdk-ci-recipes | examples/sdk-ci/check_recipes.py: each SDK CI recipe (Go, Rust, Python, JavaScript) end to end as an SDK maintainer uses it: a scratch SDK repository holding the pinned SDK’s tree and the recipe wired to it, the workflow’s own test step, offline; each must pass as committed and fail on a new divergence (SDKCI_LANGS="go js" selects languages; use-in-sdk-ci.md §2) |
make sdk-readiness | Over the run make verify-conformance-public wrote: the readiness table (readiness.md, readiness.json: which pinned SDK versions decode each protocol’s and CAP’s shapes) and the scoreboard page (scoreboard.html), into the same report directory |
make results | results-findings, results-known-defects, results-conformance, results-replay and results-acceptance: every file under evidence/results/ (formats in evidence/results/README.md); RESULTS_CHECK=1 compares each with its sources instead of writing it |
make results-conformance | scripts/py/results.py conformance: evidence/results/conformance-matrix.json from the runs make verify-conformance and make verify-conformance-public wrote under REPORTS |
make results-replay | scripts/py/results.py replay: evidence/results/replay-summary.json from the runs of make replay-full, verify-replay, replay-etl-known and harness-selftest under REPORTS |
make results-acceptance | scripts/py/results.py acceptance: evidence/results/acceptance.json, one row per acceptance check of docs/acceptance.md with the status and reason of the acceptance table make verify-acceptance wrote under REPORTS (acceptance.json beside acceptance.md) |
make results-known-defects | scripts/py/results.py known-defects: evidence/results/known-defects.json from the README’s known upstream defects table |
make results-prevalence | xdrcheck-corpus scan entry-tokens, then scripts/py/results.py prevalence: evidence/mainnet-scan/entry-prevalence.json, each corpus entry’s rarest shape token and how often it occurs on mainnet, from the scan’s token table (TOKEN_TABLE, not in the repository) |
make results-siblings | scripts/py/sibling_matrix.py generate: evidence/results/sibling-matrix.json and findings/SIBLINGS.md, each SDK defect class against each SDK family, from the level-resolved runs of make verify-conformance and the public set under REPORTS; RESULTS_CHECK=1 compares instead of writing |
make results-siblings-refresh | scripts/py/sibling_matrix.py refresh: the sibling matrix’s record states and its page from findings/, without a run (after a record changes state) |
make results-mainnet | xdrcheck-corpus mainnet coverage: evidence/results/mainnet-coverage.json, the tier’s transaction-level tokens held by the mainnet tier and by the curated corpus; RESULTS_CHECK=1 compares instead of writing |
make results-mainnet-check | xdrcheck-corpus mainnet coverage --check, part of verify-coverage: the statement is current with corpus/mainnet/ and the curated corpus (it moves when an entry adds a token) |
make results-siblings-check | scripts/py/sibling_matrix.py check and its tests, part of repo-checks: the sibling matrix is consistent with conformance-matrix.json and findings/, and findings/SIBLINGS.md is its rendering |
make lint-fixtures | The linter’s seeded-violation fixtures (go test -json in lint/, into reports/lint-fixtures.jsonl) |
make lint-docs | The rule pages under lint/docs/ match the rule registry and fixtures |
make lint-plugin | The golangci-lint module plugin loads under the pinned golangci-lint |
make verify-lint | lint-fixtures, lint-wallet-backend, lint-before-after, lint-stellar-etl, lint-selftest |
make verify-lint-pr | The pull-request CI’s form of verify-lint: lint-fixtures and lint-wallet-backend; the before/after pairs, stellar-etl and the self-test run in the full CI |
make verify-guide | xdrcheck-corpus guide-table --check --report reports/guide-table.json and guide-claims --check; with the network (DOCS=1, the default when not offline), the documentation site’s own checks |
make verify-coverage | xdrcheck-corpus coverage --check, and results-mainnet-check |
make mainnet-fetch | xdrcheck-corpus mainnet fetch (network): the mainnet tier’s 6,150 public-lake objects, about 1 GB, into $XDRCHECK_CACHE/mainnet/, hash-checked and resumable; a full cache makes no request (also make fetch FETCH_ONLY=mainnet; not part of make fetch) |
make verify-mainnet | xdrcheck-corpus mainnet check from the cache make mainnet-fetch filled, offline under OFFLINE=1: every part’s hash chain to its lake object, its round trip and its cut, and the tier’s 15,186 transaction-level tokens recomputed with the scan’s emitter (corpus/mainnet (corpus/mainnet/README.md)); writes reports/mainnet-check.tsv; the full CI’s P18 |
make mainnet-tier-check | scripts/py/check_mainnet_tier.py, part of repo-checks: corpus/mainnet/tier.json and every line of tier.jsonl against their schemas, the header’s sha256 and counts, one line per transaction and each token on one line, the lake keys and the cover order, without the lake objects |
make acceptance-report-pr | xdrcheck report acceptance reports --full-ci <targets>: the pull-request CI’s table, where a check met on that run whose other runs only the full CI makes (verify-replay, verify-conformance, lint-before-after, lint-stellar-etl) is proven in full CI |
make dist-bin | scripts/dist.sh bin: dist/xdrcheck-<version>-<os>-<arch>.tar.gz with xdrcheck, xdrcheck-corpus, xdrlint and the template adapters (Linux) or xdrlint (macOS), for each platform of DIST_PLATFORMS that builds completely on this machine, then SHA256SUMS; DIST_STRICT=1 fails on one that does not (release.md §9) |
make dist-image | scripts/dist.sh image: the container image xdrcheck-dist:<version> from the release assets in dist/, built with the network off and smoke-tested offline (each tool starts, the bundled corpus verifies, the demo replay finds its bug, the minimal adapter passes), then removed unless KEEP_IMAGE=1 (release.md §10) |
make release-assets | scripts/dist.sh assemble: SHA256SUMS and xdrcheck-<version>.provenance.json over every asset of the version in dist/, each asset checked against them, then the dist-image smoke test; what the release workflow runs last |
make action-selftest | action/selftest.sh: the GitHub Action (action.yml) run step by step against the release in dist/, every mode and failure path, and the workflow examples of github-action.md through actionlint |
make verify | Every check at full size, then the run summary table (acceptance-report): what the full CI runs besides make test |
make verify-pr | What the pull-request CI runs: verify-static, verify-corpus, verify-generate, test-fast, verify-replay-pr, verify-conformance-js, verify-lint-pr, verify-guide, verify-coverage, verify-findings, then acceptance-report-pr |
make fetch-nightly | Network: what only the nightly jobs need — the public columns’ driver packages (xdrcheck conform fetch --set public) and the stellar-core test image by digest |
make nightly-regenerate | Nightly: xdrcheck-corpus gen --check --all, every scenario, without network |
make verify-goldens-nightly | Nightly: golden agreement with stellar-core’s own test binary and the bulk tier across drivers |
make mutate-sensitivity | By hand: TestMutateSensitivity with XDRCHECK_MUTATE_REQUIRE=1, so xdrcheck mutate must re-find the known public findings it is tested on and cannot skip for a missing driver |
make nightly-early-warning | Nightly (N6): stellar-core’s test-lcm-next goldens at master (fetched; network) round-tripped in the oracle; the summary separates fetch (network), layout (stellar-core moved or renamed its goldens) and decode failures, and its arms line names the arms stellar-core’s XDR submodule has that the pinned definitions lack (xdrcheck xdr diff) |
make nightly-xdr-watch | By hand, network: xdrcheck xdr watch over stellar-xdr main (every #ifdef on), its other branches and open pull requests, the standing lcm_v3 item and the XDR diffs of pending CAP documents, then xdrcheck xdr candidates as a patch; writes reports-nightly/xdr-watch/ (digests per upcoming protocol, watch-log.jsonl, per-source diffs, candidates.patch). Findings never fail it |
make xdr-ceilings | xdrcheck xdr ceilings --write: rewrite the drivers’ sentinel-enum ceilings from the pinned XDR definitions; make test fails when they drift |
make nightly-refetch | Nightly, network: make fetch into an empty cache, then xdrcheck pins check, the golden tree’s hashes and xdrcheck-corpus refetch |
make watch-network | Nightly, network: xdrcheck pins check --network, the Horizon roots of mainnet and testnet and the status page’s scheduled maintenances against pins.toml (maintenance.md §5.5) |
make nightly-canary | Nightly, network: xdrcheck pins latest (pins behind the newest upstream release) and canary columns for SDK releases not yet pinned, diffed against the pinned columns; reports drift, fails only when it cannot run (maintenance.md §5.3) |
make vnext-core | By hand, network and docker: stellar-core master built with the next-protocol flag, cached by commit; VNEXT_PLAN=1 prints the plan (maintenance.md §5.4) |
make maint-selftest | Offline: renovate.json, the canary and the next-protocol build scripts against fixtures |