Repository navigation
Conversation
commit: |
|
| mean | stddev | min | max | |
|---|---|---|---|---|
| PR | 228.1 ms | 2.6 ms | 224.2 ms | 233.1 ms |
| base | 230.6 ms | 3.8 ms | 225.5 ms | 237.6 ms |
Δ (PR vs base): ↓ -2.5 ms (-1.1%)
Measured with hyperfine on ubuntu-latest (3 warmup runs, 20 timed runs). CI numbers carry ±a few ms of runner jitter; treat small deltas as noise.
|
🤖 I've done a thorough review of your PR. Adds a parallel MSW/Vitest snapshot suite that records request behavior for the generated and hand-written CLI command surface.
|
|
🤖 I've done a thorough review of your PR. Adds HTTP request regression snapshots across the CLI surface.
|
|
The workflow request guard has two gaps that let required hand-written API calls disappear when baselines are updated.
|
|
Review: 2 findings. 🤖 I've done a thorough review of your PR. Adds HTTP request regression snapshots and CI coverage.
Posted both fixes as GitHub suggestion comments. |
|
The workflow guard needs to check the expected HTTP method as well as the path. The preview snapshot records |
|
🤖 I've done a thorough review of your PR. Adds parallel MSW request snapshots and Linux CI coverage.
|
|
No new concerns since the last review |
There was a problem hiding this comment.
The suite claims alias coverage, but every included invocation uses canonical option names. In particular, cf containers images delete is exercised with --force but never its public -f alias, so a regression in that yargs alias or telemetry alias mapping can be accepted by updating this snapshot. Generate one equivalent case for each OptionMeta.alias (all alias forms are represented, and the current included surface has -f).
|
🤖 I've done a thorough review of your PR. Adds HTTP request regression snapshots and CI coverage.
|
|
No new concerns since the last review |
|
LGTM |
26d71fc to
a4127eb
Compare
|
LGTM! |
|
🤖 I've done a thorough review of your PR. Adds generated-command HTTP request snapshots and sharded CI coverage.
|
|
LGTM! |
|
I am not convinced that these additional tests add much value on top of the effective snapshot which is the generated code. Do we really need to run 69,000 tests on every change? |
Adds a dedicated Vitest + MSW regression suite that runs the real
cfCLI for every generated API command and snapshots its HTTP requests. Each command must reach its declared HTTP method and endpoint before its baseline can be accepted.The current catalogue covers 2,952 generated commands and 1,169 command groups across 69,280 argument cases. Cases cover no arguments, required values, individual flags, enum and boolean values, pairwise flag presence, positional alternatives, command and option aliases, global options, local mode, body inputs, and uploads. Snapshots preserve method, URL, headers, JSON/binary bodies, and ordered multipart parts, with deterministic normalization for runtime and fixture details.
All commands listed in
hand-written-commands.jsonand their exclusive groups remain excluded: local projects, external tools, and multi-request workflows need separate fixtures outside this suite's scope. Mock responses are empty success envelopes apart from zone-name lookup. Local request rewriting stays real, with Miniflare dispatch routed through MSW.The regression task regenerates commands before testing. Twelve balanced test files run in twelve Linux GitHub Actions jobs; the small harness suite runs with the normal cross-platform CLI tests. Snapshots are marked as generated, and obsolete baselines have been removed after rebasing onto current main.
Validation:
pnpm checkpnpm --filter cf test:regression --maxWorkers=4— 4,121 testsTest and CI changes only; no changeset is required.