Skip to content

docs(release): v1.6.0 notes - #61

Open
pedromvgomes wants to merge 1 commit into
mainfrom
docs/release-v160
Open

docs(release): v1.6.0 notes#61
pedromvgomes wants to merge 1 commit into
mainfrom
docs/release-v160

Conversation

@pedromvgomes

Copy link
Copy Markdown
Owner

Release notes for v1.6.0, covering pipeline.{ci,cd}.independent_stages (#60).

Same convention as v1.5.0: the notes commit is the tagged commit, so merging this is what v1.6.0 gets cut from, then v1 moves.

What the notes cover

  • Why the fixed graph is still right, and why an edge in it is a claim — that the later stage consumes what the earlier produced — which a repository can be correct to deny.
  • The measured cost. wardnet's e2e rebuilds the daemon from source and reads nothing build uploads; serialising them takes a daemon PR from ~24m to ~35m. From run 32688637246, not estimated.
  • What it deliberately cannot do — remove a sibling edge, never add or reorder one, and never drop preflight.

Two sections carrying the weight

## Detaching a stage does not ungate it says outright that ci-gate still waits on the stage and a failing end2end still blocks the merge. That's the one way this feature could be badly misread, and a speed change that quietly weakened branch protection would be the worst outcome available — so it's stated plainly rather than left implicit.

## Upgrading is the inverse of v1.5.0: nothing to do, no fleet sync, no re-render. Verified by a byte-identical render diff against v1.5.0 (bc668c9), comments included. Worth the contrast being explicit, since the previous release did want a fleet sync.

It also closes with the one caution I'd want a reader to take away: adopt it only where the claim is actually true, because the default edge costs time but never correctness, and being wrong here means testing a build you didn't wait for.

Range

v1.5.0..main is exactly one commit (#60), so there's nothing else to fold in.

Covers pipeline.{ci,cd}.independent_stages (#60): why a fixed graph is
still right, why an edge is a claim a repository can be right to deny,
and the measured eleven minutes wardnet would otherwise pay per daemon
PR waiting on artefacts its e2e suite never reads.

Two sections carry the load. "What it deliberately cannot do" states the
constraint that keeps the graph fixed — a named stage removes a sibling
edge, never adds or reorders one. "Detaching a stage does not ungate it"
says outright that ci-gate still waits on the stage, because a speed
change that quietly weakened branch protection is the worst way this
could be misread.

Upgrading is the inverse of v1.5.0: nothing to do, no fleet sync, no
re-render, verified by a byte-identical render diff against v1.5.0.
@github-actions

Copy link
Copy Markdown

bulwark

  • scan — no findings

  • coverage — go: 76.9% (baseline 76.9%)

📦 Full bulwark output — complete scan and coverage logs.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant