devops(bidi): fix publishing when a single csv report is uploaded

[Yury Semikhatsky (yury-s)]

Summary

• actions/download-artifact only creates per-artifact subdirectories when more than one artifact matches the pattern. Only moz-firefox-nightly uploads a csv today, so the glob matched nothing and the publish step failed on the first run.
• Name each report after its channel at upload time and merge them into one flat directory, which does not depend on how many artifacts exist.
• Drop the branch seeding, bidi-reports is created manually.

Follow up to #42760.

[github-actions]

Test results for "tests 1"

2 failed
❌ [installation tests] › playwright-cdn.spec.ts:43 › playwright cdn failover should work (https://cdn.playwright.dev/dbazure/download/playwright) @package-installations-macos-latest
❌ [installation tests] › playwright-cdn.spec.ts:43 › playwright cdn failover should work (https://playwright.download.prss.microsoft.com/dbazure/download/playwright) @package-installations-macos-latest

7 flaky

⚠️ [chromium-library] › library/chromium/chromium.spec.ts:373 › should produce network events, routing, and annotations for Service Worker `@frozen-time-library-chromium-linux`
⚠️ [chromium-library] › library/video.spec.ts:725 › screencast › should capture full viewport on hidpi `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/popup.spec.ts:260 › should not throw when click closes popup `@chromium-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:257 › third party 'Partitioned;' cookies `@firefox-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:470 › top level 'Partitioned;' cookie and same origin iframe `@firefox-ubuntu-22.04-node20`
⚠️ [firefox-page] › page/page-emulate-media.spec.ts:144 › should keep reduced motion and color emulation after reload `@firefox-ubuntu-22.04-node20`
⚠️ [firefox-page] › page/page-event-request.spec.ts:181 › should return response body when Cross-Origin-Opener-Policy is set `@firefox-ubuntu-22.04-node20`

51971 passed, 1247 skipped

---

Merge workflow run.

[github-actions]

🟢 Both failures are pre-existing CDN flakes — the PR is clear

Hi, I'm the Playwright bot and I triaged the CI failures here.

Both failures are playwright-cdn.spec.ts:43 on package-installations-macos-latest, and the same test failed on main (sha 63919674, run 35272518862) and on several unrelated PRs today. This PR only edits .github/workflows/tests_bidi.yml, which nothing in the installation suite reads.

Details

Pre-existing flake / infra

• [installation tests] › playwright-cdn.spec.ts:43 › playwright cdn failover should work (https://cdn.playwright.dev/dbazure/download/playwright) and (https://playwright.download.prss.microsoft.com/dbazure/download/playwright) — grouped, they're the same failure mode. Both fail with Expected the command to exit cleanly (0 status code), but exited with 1. Command: npx playwright install, i.e. the CDN download itself, not an assertion.

  In the aggregated CI results these two tests failed 19 of 644 runs (3%) and passed the other 625, and the failures are confined to package-installations-macos-latest — the same two tests are 0/632 failures on ubuntu-latest and 0/630 on windows-latest, which is what a mirror/network hiccup looks like rather than a code change.

  Recent failures on SHAs this PR can't be responsible for:

  when	branch / PR	sha	run
  20:44	main	63919674	35272518862
  19:40	main	ad9c8593	35266171076
  21:16	feat(screencast): style action decorations, keep cursor visible #42758	b26a1c79	35275610968
  21:05	feat(webkit): roll to r2365 #42762	fbd2504c	35274543784
  19:44	fix(trace-viewer): reject unsafe snapshot tag names #42712	eef2c8ac	35264659653

  Note 63919674 is this PR's own base commit, so the failure predates the change.

Impact of the diff

The only changed file is .github/workflows/tests_bidi.yml — artifact naming and the bidi-reports publish step. The tests 1 installation job runs from a different workflow and doesn't consume those artifacts, so there is no path from this diff to a CDN download failure.

The 7 flaky tests in the report all passed on retry and are likewise in library/page suites this workflow-only change doesn't touch.

_{Triaged by the Playwright bot - agent run}

[github-actions]

Test results for "MCP"

2 failed
❌ [firefox] › mcp/cli-session.spec.ts:54 › idle timeout shuts the session down @mcp-windows-latest-firefox
❌ [firefox] › mcp/cli-webmcp.spec.ts:35 › webmcp-list and webmcp-call @mcp-windows-latest-firefox

8588 passed, 1446 skipped

---

Merge workflow run.

[github-actions]

🟡 One known flake, one I can't pin down — but the diff can't reach either

Hi, I'm the Playwright bot and I triaged the CI failures on this PR.

Both failures are on mcp-windows-latest-firefox. cli-session.spec.ts:54 is a well-established flake on that bot (failed 29 of 85 runs, including plain main). cli-webmcp.spec.ts:35 I can only call uncertain — it has failed exactly once in 105 runs, and that once is here. The only changed file is .github/workflows/tests_bidi.yml, which the MCP workflow doesn't read, so there's no path from this diff to either.

Details

Pre-existing flake / infra

• [firefox] › mcp/cli-session.spec.ts:54 › idle timeout shuts the session down — fails with Expected substring: "(no browsers)" while the browser is still reported status: open, i.e. the idle shutdown hadn't landed yet when the assertion ran. On mcp-windows-latest-firefox it failed 29 of 85 runs (34%) and passed the other 56; on every other bot it is 0 failures (macOS/Ubuntu firefox, all chromium/chrome/msedge/webkit — 86–90 runs each). Recent failures on SHAs this PR can't be responsible for, all with the identical error:

  when	branch / PR	sha	run
  17:30	main	8bb2c097	35252655468
  16:52	main	24b69326	35248755975
  07:39	main	f1ea64bd	35195191777
  19:48	fix(mcp): keep upload modal, skip short secrets and bad headers #42713	acd1b229	35266108031
  15:55	feat(video): allow recording video at a custom fps #42752	b0447581	35242918713

Uncertain

• [firefox] › mcp/cli-webmcp.spec.ts:35 › webmcp-list and webmcp-call — fails with No WebMCP tool named "search". The page does not register any WebMCP tools., so the page's tool registration hadn't been picked up when webmcp-call ran. That reads like a registration race, and it's Windows-Firefox-only (0 failures across 106–110 runs on each of the other nine MCP bots). But I can't meet the bar for calling it a flake: this test has failed 1 of 105 runs on mcp-windows-latest-firefox, and that one failure is this PR's branch — the error string appears nowhere else in the results DB. To be sure it isn't real, I'd want the same failure on a SHA this PR doesn't own, or a re-run here that goes green.

  For what it's worth, the circumstantial case for "not this PR" is strong: the diff is a single workflow YAML.

Impact of the diff

The only changed file is .github/workflows/tests_bidi.yml — artifact naming for the per-channel csv reports and the bidi-reports publish step. The MCP suite runs from a different workflow and consumes none of those artifacts, so nothing in this change alters MCP session teardown or WebMCP tool registration.

_{Triaged by the Playwright bot - agent run}