wehub-resource-sync
26f897c1ec
release / release-please (push) Failing after 1m49s
docs / build (push) Failing after 6m34s
release / build-and-upload (arm64, linux) (push) Has been cancelled
release / build-and-upload (arm64, windows) (push) Has been cancelled
release / build-darwin (amd64, darwin) (push) Has been cancelled
release / checksums (push) Has been cancelled
release / finalize (push) Has been cancelled
release / build-darwin (arm64, darwin) (push) Has been cancelled
release / build-and-upload (amd64, linux) (push) Has been cancelled
release / build-and-upload (amd64, windows) (push) Has been cancelled
docs / deploy (push) Has been cancelled
19 lines
1.1 KiB
Go
19 lines
1.1 KiB
Go
// Package e2e holds end-to-end tests that drive the real no-mistakes
|
|
// binary against a temporary git repo and a fake agent. Tests live behind
|
|
// the `e2e` build tag so they are excluded from `go test ./...` and only
|
|
// run via `make e2e` (or `go test -tags=e2e ./internal/e2e/...`).
|
|
//
|
|
// The fake agent (cmd/fakeagent) is symlinked under each agent's binary
|
|
// name (claude, codex, opencode) into a temp PATH directory, and replies
|
|
// with deterministic canned responses defined by Scenario YAML or by the
|
|
// built-in "everything is clean" default. Every invocation is appended to
|
|
// $FAKEAGENT_LOG so tests can assert on which prompts the pipeline made.
|
|
//
|
|
// Why e2e at all: see the audit in the PR description. Unit tests for
|
|
// pipeline orchestration, executor branching, and CLI wiring give weak
|
|
// signal compared to actually running the binary against a real git
|
|
// repo. The e2e suite consolidates the "user pushes a branch and the
|
|
// pipeline runs to completion" journey into a single high-coverage test
|
|
// that is meant to grow rather than fan out into many small e2e files.
|
|
package e2e
|