Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
55 lines (41 sloc) 2.19 KB

Goolge Cloud Storage

Note: to apply the all permissive CORS policy to a new bucket, run: gsutil cors set util/gs-cors.json gs://[BUCKET]

Here are the buckets used by this project. Most of them have a *-staging counterpart for the wptdashboard-staging project.

wptd

This is a publicly readable bucket (with an all-permissive CORS policy) that serves the test results on wpt.fyi. The bucket has wpt.fyi-internal result summary files and results for individual tests for each test run. Generally, you shouldn't directly consume this bucket. Please see the wptd-results bucket below if you want to fetch the raw results.

The directory schema is roughly:

  • [SHA] (full or short)
    • [product-id]-summary.json.gz
    • [product-id]
      • Split JSONs with the same directory hierarchy as tests in WPT...

(See wptd-results for the definition of [product-id].)

wptd-results

This is a publicly readable bucket (with an all-permissive CORS policy) that contains full raw results (wptreport.json generated by wpt run). The wpt.fyi frontend currently doesn't serve files from this bucket, but various APIs return raw_results_url pointing to this bucket.

The directory schema is roughly:

  • [Full SHA]
    • [product-id]
      • report.json

where

  • [product-id] := [browser]-[version]-[os]{-[os_version]}{-[checksum]}
  • [browser] := {chrome, edge, firefox, safari, …}
  • [version], a version string, preferably output by browser-binary --version, with anything other than [A-Za-z0-9_.-] replaced by underscores (e.g. "61.0a1", "68.0.3409.2_dev")
  • [os] := {linux, windows, macos, android, ios}
  • [os_version], a version string similarly sanitized as [version]
  • [checksum], some checksum of the wptreport.json served as a UID to differentiate multiple runs of the same version of a browser at the same sha; currently, it's a prefix of the SHA1 of the report.

Users who want to fetch the raw history results should not try to construct the URL themselves, as [product-id] has many optional parts and even a checksum. Instead, please use the APIs to query test runs and download the raw_results_url of the test runs.