docs: duration does not tell you whether a job ran #6

Merged
jlxq0 merged 1 commit from docs/duration-is-not-the-discriminator into main 2026-08-26 07:18:42 +00:00
Owner

The bullet merged in #3 an hour ago says a job ending seconds after it starts never ran. That is true of run 16913 and false as a general rule, and I wrote it before seeing the counter-examples.

Two docker jobs finished legitimately in ten and nineteen seconds the same afternoon — 17003 and 17070 — because buildkitd's local layer cache hits every Rust layer when no source changed. cargo ran in 2m22s and then in 39s on the same gates, same repository, same day.

The discriminator is whether a task exists in GET actions/tasks for that sha and job name, not the clock. That is also what separates a skipped job from a real one, so the file was already saying it once in the bullet above and should not contradict itself two paragraphs later.

Documentation only. No code, no workflow change.

The bullet merged in #3 an hour ago says a job ending seconds after it starts never ran. That is true of run 16913 and false as a general rule, and I wrote it before seeing the counter-examples. Two `docker` jobs finished legitimately in ten and nineteen seconds the same afternoon — 17003 and 17070 — because buildkitd's local layer cache hits every Rust layer when no source changed. `cargo` ran in 2m22s and then in 39s on the same gates, same repository, same day. **The discriminator is whether a task exists in `GET actions/tasks` for that sha and job name, not the clock.** That is also what separates a skipped job from a real one, so the file was already saying it once in the bullet above and should not contradict itself two paragraphs later. Documentation only. No code, no workflow change.
docs: duration does not tell you whether a job ran
All checks were successful
CI / cargo (pull_request) Successful in 2m10s
CI / docker (pull_request) Successful in 8s
388b83d107
The bullet I merged an hour ago says a job ending seconds after it starts
never ran, which is true of 16913 and false as a general rule. Two docker
jobs finished legitimately in 10 and 19 seconds the same afternoon,
because buildkitd's local layer cache hits every Rust layer when no
source changed, and cargo ran in 2m22s and then 39s on the same gates.

The discriminator is whether a task exists in GET actions/tasks for that
sha and job name. That is also what separates a skipped job from a real
one, so it is the same check twice and the file should say it once.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NCiHcHN6LrhiuLXbG9SBYG
jlxq0 merged commit 98273db449 into main 2026-08-26 07:18:42 +00:00
jlxq0 deleted branch docs/duration-is-not-the-discriminator 2026-08-26 07:18:42 +00:00
Sign in to join this conversation.
No reviewers
No labels
waiting-on-julian
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jlxq0/hevy-mcp!6
No description provided.