Skip to content

landing: the conformance figure is 24,106, and +404 of the jump is the measurement - #45

Merged
tannevaled merged 1 commit into
mainfrom
conformance-24106
Nov 20, 2026
Merged

tannevaled merged 1 commit into
mainfrom
conformance-24106

Conversation

@tannevaled

Copy link
Copy Markdown
Contributor

The conformance figure is 24,106 — and this page's own gate was red until it said so

go-embedded-ruby/ruby#961 re-recorded BASELINE on the pinned ubuntu runner; it now sums to 24,106. The conformancefigure@v0.8.0 step in deploy-pages.yml reads that file from ruby's main, so from the moment go-embedded-ruby/ruby#961 merged it failed here on 6 claims, all reading 23,672 — and every push and pull request to this repository would have failed it for a reason unrelated to what it proposed.

site change
hugo.toml:74 23,672 → 24,106; re-recorded 2026-11-03 → 2026-11-20
layouts/_default/list.html:73 the same, in the "what CI guarantees" item
layouts/_default/list.html:79 the climb from 6,000
layouts/index.html:25, :60, :69 trust bar, card, stat

⭐ One sentence the page owed itself

This page separates jumps that were fixes to the measurement from jumps in what the VM can do. This one is mostly the first kind. The shim's platform guards compared pointer_size: 64 against a size in bytes, so every such guard was shut on every machine and two whole files had been counted at zero. Fixing that is +404 of the day's +434: main alone would have read 23,702, and 24,106 − 23,702 = 404. Without saying so, the page would present a better instrument as a better interpreter — the exact confusion the sentence it extends was written to prevent. Both places that list the measurement fixes now count three.

Checked against the new BASELINE rather than assumed, and left alone because still true: 2,202 of 2,206 files produce a result (4 FILEFAIL), and core/random/bytes_spec.rb's row is still 8, its floor.

Control

The pinned gate, on this tree, against ruby's main:

6 total claim(s), 1 version claim(s) and 3 stubbed-helper mention(s) checked against the gate's 24106 at Ruby 4.0.9

⛔ The count is the evidence, not the exit status. A rewording that took a claim out of the tool's vocabulary would have checked 5 and passed just the same; it checked 6 before and 6 after.

Refs go-embedded-ruby/ruby#961, go-embedded-ruby/ruby#837 — both closed; named for where the figure comes from.

🤖 Generated with Claude Code

…e measurement

go-embedded-ruby/ruby#961 re-recorded BASELINE on the pinned ubuntu
runner, and it now sums to 24,106. This page's own gate
(conformancefigure@v0.8.0, the step in deploy-pages.yml) was RED against
ruby's main from that merge on: 6 claims, all reading 23,672. Every push
and pull request here would have failed it for a reason that had nothing
to do with the change being proposed.

Six figures and two dates move:

  hugo.toml:74            23,672 -> 24,106, re-recorded 2026-11-03 -> 11-20
  list.html:73            the same, in the "what CI guarantees" item
  list.html:79            the climb from 6,000
  index.html:25, :60, :69 the trust bar, the card, the stat

⭐ AND ONE SENTENCE THE PAGE OWED ITSELF. This page carefully separates
jumps that were fixes to the MEASUREMENT from jumps in what the VM can
do. This one is mostly the first kind: the shim's platform guards
compared `pointer_size: 64` against a size in bytes, so every such guard
was shut on every machine and two whole files had been counted at zero.
Fixing that is +404 of the day's +434 -- main alone would have read
23,702, and 24,106 - 23,702 = 404. Without saying so, the page would have
presented a better instrument as a better interpreter, which is the
exact confusion the sentence it extends was written to prevent. Both
places that list the measurement fixes now count three.

Still true and left alone, checked against the new BASELINE rather than
assumed: 2,202 of 2,206 spec files produce a result with 4 FILEFAIL, and
core/random/bytes_spec.rb's row is still 8, its floor.

CONTROL: the pinned gate, run on this tree against ruby's main, checks
6 claims and passes. The COUNT is the evidence, not the exit status: a
rewording that took a claim out of the tool's vocabulary would have
checked 5 and passed just the same.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@tannevaled
tannevaled merged commit c220ecc into main Nov 20, 2026
4 checks passed
@tannevaled
tannevaled deleted the conformance-24106 branch November 20, 2026 18:04
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