Skip to content

docs: the development line is 0.9.0-dev, not 0.10.0-dev - #420

Open
aaltshuler wants to merge 1 commit into
mainfrom
docs/version-line-0.9
Open

docs: the development line is 0.9.0-dev, not 0.10.0-dev#420
aaltshuler wants to merge 1 commit into
mainfrom
docs/version-line-0.9

Conversation

@aaltshuler

@aaltshuler aaltshuler commented Aug 2, 2026

Copy link
Copy Markdown
Collaborator

What changed

  • docs/releases/: merged the rolling "v0.10.0" notes into v0.9.0.md as one draft for the actual upcoming release (everything since 0.8.1; header now says unreleased — rolling draft; a bridge note explains the former v0.10.0 heading), deleted v0.10.0.md. Corrected the old text's claims that the release stamps v9 ("advances from v4 to v9" → to the current stamp, v19 at writing) and that "a published binary only ever wrote v4 or v9" (only v4 or earlier — 0.9.x never shipped).
  • upgrade.md: the v9 table row no longer claims a published 0.9.x line ("export with the latest 0.9.x" pointed at a binary that doesn't exist); v9 gets the same source-build-at-matching-commit guidance as v5–v8, and the v9→v10 migration recipe says "final-v9 source build." All "0.10.0-dev" labels → "0.9.0-dev"; "final 0.10.0 format" → "final 0.9.0 format."
  • versioning.md: the load-bearing false sentence — "Internal schema v9 maps to OmniGraph 0.9.x — the first published release line to serve any of these formats" — replaced with the truth (release prepared at v9, never published; only source builds serve it), with an honest note that the in-code release map still says 0.9.x pending the release-prep relabel. Nine stanza labels 0.10.0-dev → 0.9.0-dev.
  • testing.md: the v9-seam paragraph stops calling 0.9.x "published" while keeping the literal pinned strings it documents accurate to the code; two 0.10.0-dev mentions relabeled.
  • firehose-path-specs.md: "0.10.0 release gate" → "0.9.0 release gate" (×2).

Why

v0.8.1 is the latest actual release. 0.9.0 was prepared (a2215f60 workspace bump + notes) and never tagged, so every development build since — including the one stamping v19 graphs today — identifies as 0.9.0. The docs' "0.10.0-dev" line named a release line that doesn't exist, and worse, the v9 guidance sent operators looking for a published 0.9.x binary that was never built.

Deliberately out of scope

The code half of the same problem: release_for_internal_schema_version still maps 9 => "0.9.x" (with a now-false "published line" comment) and 10..=19 => "0.10.0-dev", and those exact strings are pinned by the in-file grammar tests and the env-gated v9 crossversion seam. Changing them is a coordinated code+test change and is annotated as release-prep work next to the existing arm-split instruction in migrations.rs; the docs touched here describe that pending state honestly rather than pretending it's done.

Validation

  • scripts/check-agents-md.sh — OK (73 links, 69 docs)
  • residual sweep: no "0.10" mentions left outside the deliberate historical bridge note and RFC-0023's accurate historical statements
  • docs-only change; the classifier should prove it docs-only (this PR is also a live test of the --no-renames classifier fix, since it deletes a docs file)

Greptile Summary

Corrects the documented development release line from 0.10.0-dev to 0.9.0-dev and consolidates the rolling release notes under v0.9.0.

  • Relabels internal-schema development strands and release gates as 0.9.0-dev.
  • Corrects the claim that schema v9 was served by a published 0.9.x release.
  • Moves the former v0.10.0 rolling notes into the unreleased v0.9.0 draft.
  • Updates upgrade guidance for historical source-built formats.

Confidence Score: 4/5

The final-v9 upgrade instructions should identify an immutable compatible commit before merging; the remaining release-history wording issue is non-blocking.

The PR fixes the release-line history, but operators with schema-v9 graphs are still directed to a source build that neither the refusal output nor the documentation makes locatable.

Files Needing Attention: docs/user/operations/upgrade.md, docs/releases/v0.9.0.md

Important Files Changed

Filename Overview
docs/releases/v0.9.0.md Consolidates the rolling release notes correctly but overstates when development binaries began identifying as 0.9.0.
docs/user/operations/upgrade.md Corrects the nonexistent published-v9 guidance but does not provide an immutable final-v9 revision needed to perform the documented rebuild.
docs/dev/versioning.md Accurately distinguishes the untagged v9 preparation from the published release history and relabels later strands.
docs/dev/testing.md Aligns historical cross-version seam descriptions with the actual 0.9.0 development line.
docs/dev/firehose-path-specs.md Relabels the two release-gate references without changing the documented protocol.
docs/releases/v0.10.0.md Deletes the obsolete draft after moving its content into v0.9.0; no repository or related-repository references to the old path remain.

Fix All in Claude Code

Reviews (1): Last reviewed commit: "docs: the development line is 0.9.0-dev,..." | Re-trigger Greptile

Greptile also left 2 inline comments on this PR.

Context used:

v0.9.0 was prepared (workspace bump, release notes) but never tagged or
published — v0.8.1 remains the latest release. The docs nevertheless treated
0.9.0 as a completed boundary: they claimed v9 maps to 'the first published
release line', told operators to export v9 graphs 'with the latest 0.9.x
binary' (a binary that does not exist), and labeled every stamp since v10 as
a '0.10.0-dev' line whose binaries actually identify as 0.9.0.

Relabel throughout: v9 joins v5-v8 as an unpublished 0.9.0-dev format with
source-build export guidance, v10-v19 are 0.9.0-dev, and the release gate is
0.9.0. Merge the rolling 'v0.10.0' release notes into docs/releases/v0.9.0.md
as one draft for the actual upcoming release (spanning everything since
0.8.1) and delete v0.10.0.md.

Deliberately out of scope (docs describe it honestly instead): the in-code
release map still labels stamp 9 as '0.9.x' inside refusal messages, and the
grammar tests plus the env-gated v9 crossversion seam pin those exact
strings. That relabel changes pinned test strings and is annotated as
release-prep work next to the existing arm-split note in migrations.rs.
Comment thread docs/releases/v0.9.0.md
renames, substrate-native key-conflict fencing, and crash recovery for every
writer — plus the experimental streaming-profile line described in the
sections further below. This file is the rolling draft for the unreleased
0.9.0 line; every development build since v0.8.1 identifies as 0.9.0.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Overstated development version identity

The workspace remained at 0.8.1 until the later release-preparation bump, so this sentence incorrectly says the source builds that wrote earlier development formats identified as 0.9.0. That mismatch makes historical binaries harder to identify when rebuilding v5–v8 graphs.

Knowledge Base Used: Dev conventions and invariants

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

Fix in Claude Code

| internal schema v17 | unreleased (earlier 0.10.0-dev source builds) | final v17 source build at merge `41a5990d53238d63d17e139859c66613f9c25867` |
| internal schema v18 | unreleased (earlier 0.10.0-dev source builds) | final v18 source build at merge `c7c81b186bed37989fe5ce591baf0965b5102648` |
| internal schema v19 | unreleased (current 0.10.0-dev source builds) | — current development format; a later pre-release strand may supersede it |
| internal schema v9 | no published release (0.9.0-dev source builds; the refusal label still reads `0.9.x`) | a source build at the matching commit |

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Final-v9 build remains unlocatable

When an operator needs to export a schema-v9 graph, the refusal supplies only 0.9.x and this row now requires a source build at an unspecified matching commit. Because no 0.9.x binary was published and no immutable final-v9 revision is given, the operator cannot obtain the binary required to perform the documented rebuild.

Knowledge Base Used: Dev conventions and invariants

Fix in Claude Code

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