show how long each build has been taking for each repo by build config #49

2026-09-10T08:56:35+02:00

mi commented

2026-09-10 08:56:35 +02:00

The duration of a single build has been recorded since the bash
predecessor, but never compared with the other runs of the same build.
The History page now carries one card per build definition: the window's
successful runs as a line, the median as a reference, and the latest run
measured against it — amber once it needs a quarter longer.

Grouped by definition with the branches pooled, not per branch: a branch
leaves two or three measurements and disappears. The median rather than
the average is what makes that pooling affordable, since one cold cache
or one branch that redefined its build no longer colours a month.

Revises Scope Decision 2 of plan step 20, which decided the opposite.

Co-Authored-By: Claude Opus 5 noreply@anthropic.com

The duration of a single build has been recorded since the bash predecessor, but never compared with the other runs of the same build. The History page now carries one card per build definition: the window's successful runs as a line, the median as a reference, and the latest run measured against it — amber once it needs a quarter longer. Grouped by definition with the branches pooled, not per branch: a branch leaves two or three measurements and disappears. The median rather than the average is what makes that pooling affordable, since one cold cache or one branch that redefined its build no longer colours a month. Revises Scope Decision 2 of plan step 20, which decided the opposite. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

mi added 1 commit 2026-09-10 08:56:35 +02:00

server: show how long each build has been taking 39e9e9bf5c

The duration of a single build has been recorded since the bash
predecessor, but never compared with the other runs of the same build.
The History page now carries one card per build definition: the window's
successful runs as a line, the median as a reference, and the latest run
measured against it — amber once it needs a quarter longer.

Grouped by definition with the branches pooled, not per branch: a branch
leaves two or three measurements and disappears. The median rather than
the average is what makes that pooling affordable, since one cold cache
or one branch that redefined its build no longer colours a month.

Revises Scope Decision 2 of plan step 20, which decided the opposite.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

mi added 1 commit 2026-09-10 08:58:14 +02:00

Merge origin/main into build-duration-trend 38db36ed00

Keeps 1.7.0 and both release-notes entries, and lets the trend leave out
the timed-out builds PR#48 introduced: their duration measures the
timeout, never the work.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

mi added 1 commit 2026-09-10 10:26:03 +02:00

docs: record the duration trend, and let its figures wrap 46f8f227d5

The architecture skill carries the subsystem facts, so it names where
the trend comes from, what it is grouped by, and why a private
repository gets no trend at all.

The figures inside a card wrap instead of widening it: on a phone the
cards must never push the build table sideways.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

mi merged commit 634801467f into main

2026-09-10 10:26:58 +02:00

mi deleted branch build-duration-trend

2026-09-10 10:26:58 +02:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: mi/werkator#49