You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Proposal: Add optional threadId to retryAttempts entries #74
Entries in retryAttempts have their own start and stop, but no threadId. The test object's threadId describes the final attempt (§9.3), so a consumer knows when an earlier attempt ran, but not on which thread or worker.
Timeline views need this. Today they have to guess, usually the final attempt's thread, and the guess is often wrong:
Gradle test-retry runs retries in a second round, with new test JVMs.
Producers can't add it themselves: §11 forbids unknown fields outside extra, and the schema sets additionalProperties: false on attempt items.
Proposal
Add an OPTIONAL threadId (string) to the Attempt History Entry Object (§11): "Identifies the thread or worker on which this attempt executed." Same format as the test object's threadId.
Reword §9.25 to "on which the final attempt executed".
If an attempt entry has no threadId, its thread is unknown; it is not implied to be the test object's threadId. (Inheriting would make every existing report claim that earlier attempts ran on the final attempt's thread.)
Within a document, a threadId value SHOULD identify one worker, so executions that share it ran one at a time. Producers that run tests in several processes or on several hosts SHOULD qualify it, for example fork-2/Test worker.
Schema, in retryAttempts.items.properties:
"threadId": {
"description": "Thread or worker on which this attempt executed",
"type": "string"
}
Example: a test that failed on one Gradle fork and passed on retry on another.
All of this is additive; every valid 0.2.0 document stays valid.
Context
I ran into this while building a timeline report for github-test-reporter (ctrf-io/github-test-reporter#279). Related: #58 (retry model, settled in #62), #57 and #73 (attempt identity). Happy to open the spec and schema PR.
Thanks @alexshamrai, this makes sense, happy to add optional threadId to attempt history and clarify that the Test-level field describes the final attempt.
Let’s follow the existing optional identity-field pattern without adding a separate omission rule at this stage. Qualifying worker identifiers across processes and hosts makes sense too, but I’d avoid guaranteeing serial execution, since some frameworks allow concurrent tests within a worker.
Feel free to open a PR with changes, specification version would increase to 0.3.0
Activity
Ma11hewThomas commentedon Oct 24, 2026
Thanks @alexshamrai, this makes sense, happy to add optional threadId to attempt history and clarify that the Test-level field describes the final attempt.
Let’s follow the existing optional identity-field pattern without adding a separate omission rule at this stage. Qualifying worker identifiers across processes and hosts makes sense too, but I’d avoid guaranteeing serial execution, since some frameworks allow concurrent tests within a worker.
Feel free to open a PR with changes, specification version would increase to 0.3.0