Skip to content

The 29 Aug release cut, and eight things the owner asked for - #13

Merged
GLESAV merged 1 commit into
mainfrom
docs/release-cut-feedback-2026-08-29
Aug 30, 2026
Merged

The 29 Aug release cut, and eight things the owner asked for#13
GLESAV merged 1 commit into
mainfrom
docs/release-cut-feedback-2026-08-29

Conversation

@GLESAV

@GLESAV GLESAV commented Aug 30, 2026

Copy link
Copy Markdown
Owner

A manual test cut of caf10bd on macOS, and the notes it produced. No code changed — this is docs/release_cut_2026-08-29_feedback.md only.

The cut was green

Release build clean (2 benign warnings) · 631 tests, 1 skipped, 0 failures, 35 files, 23.5 s.

Verified on device: navigation direction is correct (finger up → sky), the done card is whole (verse, counts, lifetime line, no scrim), the journal's numbers match the persisted store exactly, and a source grep finds no network, analytics, StoreKit, notification or location API anywhere in Vesper/.

Five findings

Severity
F1 First-run hint returns on every new field — spec says "never returns", but restart() clears the flag it reads Major
F2 Checklist §XV is stale — the deep-backdrop rough edge was fixed by this very cut Docs
F3 No installed simulator runtime reaches the 18.5 deployment floor, so the minimum supported OS is untested Major
F4 Unused-result warning, JournalView.swift:560 Minor
F5 Version and build numbers are still v1.2's Blocker for submission

Eight requests, two of them major

R7 — the pops do not read as distinct. Measured, not asserted: of the 100 catalog entries, zero override voice, pattern or burst. All three fall through to the family, so 100 pops carry 10 sounds, 10 burst motions and at most 5 haptics, constant within a family. The channels are live and read (SceneRenderer.swift:473, GameSimulation.swift:590, HapticsEngine.swift:41) — the renderer is ready and the data is empty. An authoring gap, not an engine one.

R8 — the weather under-reads, for two separable reasons: clear is 44% of fields, and flowCarry is 0 for clear, snow and fog — half the airs move nothing at all. Storm also carries less than rain, which seems backwards.

R1 (auto-rise to sky — likely already built, needs a hands-off confirm), R2 (pointer to the next stone), R3 (piercing fireworks), R4 (chain reach too wide), R5 (visible dark edge at the sky/field join), R6 (more fireworks).

Also recorded

Three non-findings, written down so nobody re-chases them. The "inert sky" and the "unpoppable orbs" were a stale window origin in my test harness, not the app.

Every item names its file, line and constant so the change can be made deliberately — CLAUDE.md's one-knob-at-a-time rule stands.

🤖 Generated with Claude Code

https://claude.ai/code/session_01MiBreiwVPx8LF5pDdBeVff

A manual test cut of caf10bd on macOS, and the notes it produced.

The cut itself was green: Release build clean, 631 tests / 1 skipped /
0 failures across 35 files. Navigation direction is correct, the done
card is whole, the journal's numbers match the store exactly, and a
source grep finds no network, analytics, StoreKit, notification or
location API anywhere in the app.

Five findings (F1-F5). The one that matters for play is F1: the
first-run hint is specified as "never returns" but restart() clears the
flag it reads, so it comes back on every field. F3 records that no
installed simulator runtime can reach the 18.5 deployment floor, so the
minimum supported OS is currently untested. F5 records that the version
and build numbers are still v1.2's.

Eight requests (R1-R8) raised by the owner during the pass. Two are
major and both have a measurable cause rather than a taste argument:

  R7 - the pops do not read as distinct. Of the 100 catalog entries,
  zero override voice, haptic pattern or burst motion; all three fall
  through to the family, so 100 pops carry 10 sounds, 10 burst motions
  and at most 5 haptics, constant within a family. The channels are
  live and read - this is an authoring gap, not an engine one.

  R8 - the weather under-reads for two separable reasons: clear air is
  44% of fields, and flowCarry is 0 for clear, snow AND fog, so half
  the airs move nothing at all.

Also recorded: three non-findings, so nobody re-chases them. The "inert
sky" and the "unpoppable orbs" were a stale window origin in the test
harness, not the app.

No code changed. Every item names its file, line and constant so the
change can be made deliberately, one knob at a time.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MiBreiwVPx8LF5pDdBeVff
@GLESAV
GLESAV merged commit 167fa33 into main Aug 30, 2026
2 checks passed
@GLESAV
GLESAV deleted the docs/release-cut-feedback-2026-08-29 branch August 30, 2026 00:06
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