fix(cli): accept multiple files in upload command

[Dmitry Gozman (dgozman)]

Summary

• upload accepts multiple paths: the last positional argument absorbs the remaining argv when its schema accepts an array, rendered as <files...> in help
• bare upload reports a proper missing-argument error instead of Invalid input
• export generateReadme to unbreak utils/generate_cli_help.js --readme

Fixes #42047

[github-actions]

Test results for "tests 1"

5 flaky

⚠️ [installation tests] › playwright-cli.spec.ts:21 › cli should work `@package-installations-ubuntu-latest`
⚠️ [installation tests] › playwright-test-package-managers.spec.ts:19 › npm: @playwright/test should work `@package-installations-ubuntu-latest`
⚠️ [chromium-library] › library/screencast.spec.ts:28 › screencast.start delivers frames via onFrame callback `@chromium-ubuntu-22.04-node20`
⚠️ [chromium-library] › library/video.spec.ts:495 › screencast › should capture static page in persistent context Radoslav Kirilov (@smoke) `@chromium-ubuntu-22.04-node22`
⚠️ [chromium-library] › library/video.spec.ts:699 › screencast › should capture full viewport on hidpi `@chromium-ubuntu-22.04-node22`

51056 passed, 1188 skipped

---

Merge workflow run.

[github-actions]

Test results for "MCP"

3 failed
❌ [chrome] › mcp/run-code.spec.ts:21 › browser_run_code_unsafe @mcp-macos-latest-chrome
❌ [chromium] › mcp/cli-core.spec.ts:106 › fill numeric @mcp-macos-latest-chromium
❌ [firefox] › mcp/cli-core.spec.ts:195 › dialog-accept @mcp-windows-latest-firefox

7851 passed, 1272 skipped

---

Merge workflow run.

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the CI failures.

🟡 Nothing points at this PR, but two failures I can't fully clear

All three failures come from this PR's own run (merge sha bde85db04) and all three are timeouts — none land in code this PR changes. dialog-accept is a known pre-existing flake; the other two only show up on this run, so I can't prove they're flakes, only that the diff doesn't plausibly reach them.

Details

This PR touches the CLI daemon argument parser (command.ts), the upload command, and the help generator. The parser change only alters behavior when the last positional arg is variadic (an array) — the non-variadic path is byte-for-byte the same, so fill, dialog-accept, and browser_run_code_unsafe run exactly as before. Every failure is a timedOut, not a parse/assertion error, which reads like a slow runner rather than a logic regression.

Pre-existing flake / infra

• [firefox] › mcp/cli-core.spec.ts › dialog-accept — pre-existing flake. Across the test-results DB it failed on 15 SHAs spanning ~13 other PRs and pushes (e.g. PRs 42019, 42014, 41999, 41982, 41970, 41962, 41941, 41909, plus several main pushes), on firefox, msedge and chrome — passing ~5450 other runs. Unrelated to dialog handling, which this PR doesn't touch.

Uncertain

• [chromium] › mcp/cli-core.spec.ts › fill numeric (timedOut) — only failing occurrence in the DB is this PR's own run; passes on ~7200 other runs. The fill command is non-variadic, so the parser change leaves its path unchanged. Looks like a timeout flake, but with no failure on another SHA I can't call it proven. To be sure I'd want to see it fail (or flip) on a run this PR can't be responsible for.
• [chrome] › mcp/run-code.spec.ts:21 › browser_run_code_unsafe (timedOut, macos) — same story: sole failure is this PR's run, ~6200 clean elsewhere. The browser_run_code_unsafe tool has nothing to do with the CLI upload/parser changes, so the diff doesn't reach it. Timeout signature, but no cross-run evidence to confirm the flake.

None of the three is caused by this PR. A re-run of the MCP job should clear the two uncertain timeouts.

_{Triaged by the Playwright bot - agent run}