1.3.0-release
What's new
tag a8070094d3b1324948e51f339a10832d25acc7fb 1.3.0-release
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 9 19:26:45 2026 +0530
Release 1.3.0
commit 7e6473dadf847491b6fc494703c11483afec3b82
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 9 19:23:37 2026 +0530
Release 1.3.0: version bump, release notes and the gated release run
Version 1.2.0 -> 1.3.0 (owner-approved minor).
Release gate, read from docs/automated-tests/20261009-191730/manifest.json
(run with --release 1.3.0 on the clean tree at 0fcb643):
- Lint: PASS, 0 warnings / 0 errors in 71 files
- Tests: PASS, 1063 passed, 0 failed, 1 skipped of 1064 (the kit's
diagnostics opt-out case, which does not apply to this addon)
- Perf: PASS by the no-scenarios exception; the addon ships no
tests/perf.lua, so perf was skipped, not measured
- Complexity: PASS, lizard 1.24.0 ran; 17849 NLOC, 2222 functions,
avg CCN 2.0
- CCN <= 15: PASS, 0 warnings, max CCN 15
- Sighted: PASS, blindFiles 0
Files:
- PanelMaster.toc: ## Version 1.3.0
- core/Namespace.lua: NS.version = "1.3.0"
- core/Database.lua: the [Init] summary example comment reads v1.3.0
- docs/smoke-tests.md: /pm version, /pm help header and the TOC-vs-constant
smoke case read 1.3.0
- README.md: new 1.3.0 row at the top of Version History (bulleted
highlights, de-AI pass run, plus the perf-skip note as a final "- "
line); old rows untouched
- docs/automated-tests/20261009-191730/: the release bundle, with its
ANALYSIS.md
- docs/automated-tests/RESULTS.md: the runner's row, plus watch-list
dispositions. tests/test_slash.lua newly crossed into the 1000-1500
band (1166); tests/test_libka0s.lua grew to 1184 and is on its second
consecutive Accepted release run.
Left alone: the CLAUDE.md LibKa0s provenance (v1.71.0, the vendored tag),
the auto-derived CurseForge badge, the luacheck 1.2.0 mention in
DEPENDENCIES.md, and the existing history rows. docs/test-cases.md did not
change. Post-edit harness re-run: 1063/0/1 of 1064, lint 0/0.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J2w6FgfFYw93meWZN9Ahn6
commit 0fcb643d9bf049d5774210dd3e43460bb8232819
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 9 17:49:43 2026 +0530
SD-FIN-02: sync docs to code (collection sync-docs pass 2026-10-09)
Docs and comment citations reconciled against today's tree: stale counts,
drifted file:line citations, orphan anchors and toolchain gaps, each verified
independently before applying. Comment edits are comment-only. Dead-export
candidates reported, not removed. Green gate run after the edits.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01J2w6FgfFYw93meWZN9Ahn6
commit 7497562e8a8a4e2b89025382707939bbfc6f99e1
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 21:04:50 2026 +0530
PM-STD: Confirm the v2.77.0 standards reference; sync docs to PM-01..12
Release-side item for the 2026-10-07 remediation (no finding of its own).
Standards reference (WowAddonStandards v2.77.0, feature branch, WS-11):
- The three places are already in the v2.77.0 form and carry no version:
PanelMaster.toc `## X-Standard:` (:12), the README Standard badge (:6,
bare image, `_` spaces), and CLAUDE.md `## Standards compliance (read
first)`. No change.
- Swept every live file (outside libs/, tests/_kit/ and the frozen
audits/reviews/automated-tests/revendor stores) for `<file>-§N`: every
citation resolves to a `### N.` heading in v2.77.0's section files (the
one non-resolving hit is CLAUDE.md's literal `filename-§N` placeholder).
Nothing cites the changelog at the top of STANDARDS.md (WS-01), no
cross-repo citation carries a bare line number (documentation-§6
SHOULD, WS-07), and nothing restates the versioning-git branch rule
(WS-08).
Docs sync against the tree after PM-01..PM-12:
- smoke-tests.md: PM-02's PANEL-8a and PM-04's COMBAT-3b broke the doc's
own rule that a new check takes the next free number in its theme;
they are now PANEL-33 and COMBAT-7, moved to the end of their themes
(unsigned, branch-only IDs, referenced nowhere else). Index ranges
updated. Pending sign-off gains rows for PM-01 (FRAME-29), PM-02
(PANEL-33), PM-03 (FRAME-28, PANEL-16, PANEL-24), PM-04 (COMBAT-7) and
PM-05 (LOC-2, LOC-3); the PM-06 NUM row now names the item and finding
PM-R-03. None is marked passed.
- module-map.md: core/Util.lua (IsFinite, the non-finite Clamp fallback,
hex-encoding Slugify, FoldName), modules/Registry.lua (one repair rule
per template field, R.UNREPAIRED_FIELDS, folded name uniqueness),
modules/Unlock.lua (DropPending and the stand-down), settings/Slash.lua
(its four confirm popups, Sl.ActiveProfileName) and
settings/PanelEditor.lua (Delete/Reset ask; E.AcceptDelete/AcceptReset).
- scope.md + ARCHITECTURE.md Known Limitations: the Panels list's
ASCII-only sort (left out of PM-05, LOC-4) is now a listed limitation;
six -> seven.
- In-repo line cites moved by PM-items' edits re-pointed: NS.COMMANDS
Slash.lua:329 -> :371 (ARCHITECTURE, common-tasks), R:Recover's
GetScreenSize Registry.lua:912 -> :929, the media dropdown
PanelEditor.lua:285 -> :311, COERCE.media PanelSchema.lua:96 -> :98
(compat-layer), STAND_UP_EVENTS LifecycleSetup.lua:73-75 -> :75-77
(performance), currentRecord's caller PanelEditor.lua :653 -> :682
(settings-panel).
- Verified unchanged: schema.md, data-flow.md, settings-panel.md rows and
testing.md already carry PM-01..PM-12's content; Documentation map
lists every live docs/ page; slash-dispatch 21 verbs and compat 8 shims
re-counted; README Tests badge 1063/1063 equals docs/test-cases.md
Total 1063, and a fresh `--list` is byte-identical.
No RegistryRepair.lua exists (PM-07 kept Registry.lua at 990 lines), so
module-map has no new row. No version bump and no Version History row.
Deviations: renumbering PANEL-8a/COMBAT-3b and the Known Limitations
addition go beyond the item's list but are the docs sync it asks for.
Gates: 1063 passed / 0 failed / 1 skipped; luacheck 0/0; lizard -C 15
no warnings; largest authored .lua 1184 lines.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 7497562e8a8a4e2b89025382707939bbfc6f99e1
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 21:04:50 2026 +0530
PM-STD: Confirm the v2.77.0 standards reference; sync docs to PM-01..12
Release-side item for the 2026-10-07 remediation (no finding of its own).
Standards reference (WowAddonStandards v2.77.0, feature branch, WS-11):
- The three places are already in the v2.77.0 form and carry no version:
PanelMaster.toc `## X-Standard:` (:12), the README Standard badge (:6,
bare image, `_` spaces), and CLAUDE.md `## Standards compliance (read
first)`. No change.
- Swept every live file (outside libs/, tests/_kit/ and the frozen
audits/reviews/automated-tests/revendor stores) for `<file>-§N`: every
citation resolves to a `### N.` heading in v2.77.0's section files (the
one non-resolving hit is CLAUDE.md's literal `filename-§N` placeholder).
Nothing cites the changelog at the top of STANDARDS.md (WS-01), no
cross-repo citation carries a bare line number (documentation-§6
SHOULD, WS-07), and nothing restates the versioning-git branch rule
(WS-08).
Docs sync against the tree after PM-01..PM-12:
- smoke-tests.md: PM-02's PANEL-8a and PM-04's COMBAT-3b broke the doc's
own rule that a new check takes the next free number in its theme;
they are now PANEL-33 and COMBAT-7, moved to the end of their themes
(unsigned, branch-only IDs, referenced nowhere else). Index ranges
updated. Pending sign-off gains rows for PM-01 (FRAME-29), PM-02
(PANEL-33), PM-03 (FRAME-28, PANEL-16, PANEL-24), PM-04 (COMBAT-7) and
PM-05 (LOC-2, LOC-3); the PM-06 NUM row now names the item and finding
PM-R-03. None is marked passed.
- module-map.md: core/Util.lua (IsFinite, the non-finite Clamp fallback,
hex-encoding Slugify, FoldName), modules/Registry.lua (one repair rule
per template field, R.UNREPAIRED_FIELDS, folded name uniqueness),
modules/Unlock.lua (DropPending and the stand-down), settings/Slash.lua
(its four confirm popups, Sl.ActiveProfileName) and
settings/PanelEditor.lua (Delete/Reset ask; E.AcceptDelete/AcceptReset).
- scope.md + ARCHITECTURE.md Known Limitations: the Panels list's
ASCII-only sort (left out of PM-05, LOC-4) is now a listed limitation;
six -> seven.
- In-repo line cites moved by PM-items' edits re-pointed: NS.COMMANDS
Slash.lua:329 -> :371 (ARCHITECTURE, common-tasks), R:Recover's
GetScreenSize Registry.lua:912 -> :929, the media dropdown
PanelEditor.lua:285 -> :311, COERCE.media PanelSchema.lua:96 -> :98
(compat-layer), STAND_UP_EVENTS LifecycleSetup.lua:73-75 -> :75-77
(performance), currentRecord's caller PanelEditor.lua :653 -> :682
(settings-panel).
- Verified unchanged: schema.md, data-flow.md, settings-panel.md rows and
testing.md already carry PM-01..PM-12's content; Documentation map
lists every live docs/ page; slash-dispatch 21 verbs and compat 8 shims
re-counted; README Tests badge 1063/1063 equals docs/test-cases.md
Total 1063, and a fresh `--list` is byte-identical.
No RegistryRepair.lua exists (PM-07 kept Registry.lua at 990 lines), so
module-map has no new row. No version bump and no Version History row.
Deviations: renumbering PANEL-8a/COMBAT-3b and the Known Limitations
addition go beyond the item's list but are the docs sync it asks for.
Gates: 1063 passed / 0 failed / 1 skipped; luacheck 0/0; lizard -C 15
no warnings; largest authored .lua 1184 lines.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 1642bc2672e422d9ad8b89bcd653b5f56572ed3a
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:58:49 2026 +0530
PM-12: Order the CLAUDE.md stub; drop README placeholders and 1.2.0 tail
CLAUDE.md (PM-A-06, documentation-§2 "in this order"): move the
"Green gate before every commit" paragraph up to sit directly after the
docs pointer list, ahead of the LibKa0s paragraph and the provenance
line. Pure move, text unchanged; the provenance line is byte-identical,
so test_vendor_sync still finds it.
README.md (PM-A-08, documentation-§1): CurseForge strips <...>, so the
two `/pm profile <name>` examples now read `/pm profile Raid` with
"switches to a profile called Raid" prose; no other <...> placeholder
remains in player prose. The 1.2.0 Version History row loses its
contributor-facing "Released on lint, tests and complexity only..."
tail (deleted rather than prefixed, per the verified remediation and
LH-A-07), leaving only "- " prefixed highlights. The edited paragraph
is reflowed to the README's line width.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 91aa9c0b601ef765a0d1e8cf5df2f63bfc5d9da9
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:56:29 2026 +0530
PM-11: Sweep docs and comments against the tree; trim ARCHITECTURE
Docs-only and comment-only; no behavior change.
PM-A-09 (comments re-read against the code they describe):
- core/LifecycleSetup.lua: the launcher survivor bullet no longer says the
left click changes with the disabled state (Launcher minor 4 opens the
panel in either state); NS.StandUp's note no longer says OnEnable
registers the three events (OnEnable calls NS.StandUp, which registers
STAND_UP_EVENTS).
- core/LauncherSetup.lua: Register runs at OnInitialize, not OnEnable.
- settings/Slash.lua: drop the README command table claim (the README has
none); the lock pair's co-writer is the launcher menu's Locked entry,
not the minimap left click; DisabledLine's note drops the launcher
refusal claim and qualifies the bare section as slash-commands-§7.
- docs/ARCHITECTURE.md events-frames-taint-§8 row: trigger sweep re-run;
the API list now names UnitAffectingCombat and the C_AddOns readers,
UnitClass removed. Decided column left at 2026-08-05.
- docs/scope.md: both UnitClass claims restated (class color resolves via
LibKa0s-Core ResolveColor, player-scoped).
- docs/testing.md complexity row defers to the CLAUDE.md provenance line
instead of naming kit revision 36.
PM-A-10: dropped the Perf-decline preamble (it lives in the register row
and docs/performance.md); the census keeps its heading and the empty-
census statement plus the measuring command, with the hard figures, the
peel narrative and the 1000-1500 band prose cut to one paragraph linking
automated-tests/RESULTS.md. 424 -> 387 lines.
PM-R-08: core/Database.lua cites R.Sanitize in modules/Registry.lua by
name, with no line numbers.
PM-A-07: docs/automated-tests/README.md complexity row quotes
`bash tests/_kit/run-automated-tests.sh --suite complexity` and the
sighted-shadow (-L 1500) note, matching docs/testing.md.
Deviations: the cited line numbers had shifted after PM-04/PM-09 (the
Slash.lua sites are now near :358, :408 and :745); each was located by
content. The census has no table to keep (nothing is over the cap, so
the kit gate reads the empty-census statement), so the heading and that
statement were kept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit fe313d9c90eb22f02259c5f0f8597714808d24c2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:49:49 2026 +0530
PM-09: Delete the dead bare-global AddOns fallbacks in EnvSetup and Compat
GetAddOnMetadata, GetNumAddOns and GetAddOnInfo went in the 11.0 AddOns
purge and exist on no Interface 120100 client, so the rungs that read them
were dead code (finding PM-A-04).
- core/EnvSetup.lua NS.Meta: drop the bare GetAddOnMetadata rung; Env first,
then C_AddOns, else nil. Header comment updated to match.
- core/Compat.lua Compat.AddOnFolders: drop the global GetNumAddOns /
GetAddOnInfo fallback; with either C_AddOns member missing it answers nil
("cannot tell"), which callers already handle.
- .luacheckrc: drop the three globals from read_globals; luacheck stays 0/0,
proving nothing reads them.
- tests/test_compat.lua: two new cases, red first, planting the bare globals
as functions with C_AddOns (and, for NS.Meta, Env) absent and asserting
nil. The NS.Meta case reloads core/EnvSetup.lua alone under a LibStub that
answers nothing, since Env is captured at load.
- tests/test_envsetup.lua: remove the redundant mocks.GetAddOnMetadata
save/restore (wow_mock never defined the bare globals; they live only as
C_AddOns members, so wow_mock.lua needs no change).
- docs/compat-layer.md: AddOnFolders section rewritten without the fallback;
the shim table and section line cites refreshed to current Compat.lua
lines (MediaList and MouseIsOver were already 6 lines stale).
- docs/test-cases.md regenerated; README Tests badge 1061 -> 1063.
tests/test_lintconfig.lua does not pin these globals, so it is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit d7f9ca6f7746db91f1cf4eedfa2d6cc3caa7fe00
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:44:14 2026 +0530
PM-08: Render the library-absent FormatKV plainly, not the library's colors
The degraded Slash arm in settings/Slash.lua reproduced LibKa0s's colored
`|cFFFFFF00%s|r = |cFFFFFFFF%s|r` key = value format, and
tests/test_surface_parity.lua pinned that copy byte for byte against
lib.FormatKV. slash-commands-§1 lets a stub carry only DISABLED_LINE_FORMAT.
Sl.FormatKV stays assigned on the degraded arm (the PANEL verbs call it), but
now returns a plain `tostring(path) .. " = " .. tostring(valueStr)` with no |c
escapes. The comment says why it must exist and that it is deliberately not
the library's format. The parity case's byte pin is replaced by assertions
that the degraded line has the path, the value and " = ", and no "|c"; the
now-unused `mocks` local is dropped.
Finding: PM-A-03 (audit PM-050).
Deviations: no docs/ARCHITECTURE.md or docs/module-map.md text claimed the
stub reproduces FormatKV byte for byte (the slash-dispatch.md "byte for byte"
line is about DISABLED_LINE_FORMAT and is correct), so no doc edit. Case name
and count unchanged, so docs/test-cases.md and the README badge are unchanged
(regenerated list verified identical). The stale WowAddonStandards testing.md
citation is out of scope per the item.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit e08ded67bda42f186979acb10861f06cb971e73e
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:40:34 2026 +0530
PM-07: Repair every template field in Sanitize; list enabled==nil as on
PM-R-04: R.Sanitize / R.SanitizeField share one REPAIR table that claims
the two "cannot diverge", but four template fields had no rule. Now:
- accentColor and accentBorderColor get the color rule beside bgColor,
borderColor and artColor;
- artDesaturate joins BOOL_FIELDS;
- artBlend joins ENUM_FIELDS (enumMatch already knew its list through
C.PANEL_FIELD_ENUM, so no separate REPAIR.artBlend was needed).
New exported R.UNREPAIRED_FIELDS = { "name", "frameName", "id" } names the
keys intentionally stored as given. A property case in
tests/test_registry.lua probes every C.PANEL_TEMPLATE key with a sentinel
table and proves it has a rule unless it is on that list (and that a
listed key really has none). A second case covers fill-from-template and
junk repair for the four fields.
The editor's color swatch reads (makeColorPair's initial read and its
refresher in settings/PanelEditor.lua, and the composed bind's get in
settings/PanelEditorTabs.lua) pass C.PANEL_TEMPLATE[field] as
Util.Color's fallback, so a junk accentBorderColor shows black, not white.
PM-R-07: settings/Slash.lua (BuildPanelLines) and settings/PanelEditor.lua
(refreshPicker) test `rec.enabled ~= false`, matching REPAIR.enabled and
the renderer. New test_slash case pins a nil-enabled record listing in the
enabled color.
docs/schema.md notes the repair for the four fields and the property
case; docs/test-cases.md regenerated; README Tests badge 1058 -> 1061.
Deviations: the item's line cites had drifted (Slash.lua:156 is now
:189, PanelEditor.lua:601 is :627, PanelEditor.lua:334 is the color
reads at :360/:392); the intended sites were changed. Registry.lua is
990 lines, under the 1000 threshold, so no RegistryRepair.lua split.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 82da1e60401a9311311a450492640e6156d84d3a
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:35:26 2026 +0530
PM-06: Refuse NaN and infinity in panel numbers; recover repairs NaN
tonumber reads "nan", "inf", "-inf" and "1e999" as numbers, so they
slipped past every "is it a number" guard: COERCE.number stored them,
Util.Clamp returned NaN unchanged (both comparisons are false),
freeNumber and R:SetPosition passed them through, and /pm recover
reported "moved 1 panel" on every run for a NaN x because NaN ~= NaN.
- core/Util.lua: new Util.IsFinite(n) (type number, n == n, not +/-inf).
Util.Clamp sends a non-finite n to the fallback when that is finite,
else to lo.
- settings/PanelSchema.lua: COERCE.number refuses a non-finite result
with the existing "expected a number".
- modules/Registry.lua: freeNumber returns the template value for a
non-finite input; R:SetPosition drops a non-finite x or y as it drops
a non-number. RecoveredOffsets already clamps through Util.Clamp with
fallback 0, so a non-finite x/y recovers to 0 and IsOffScreen/Recover
compare against that finite value: one recover repairs it, the next
reports nothing moved. Registry.lua is 983 lines (< 1000).
- Tests (red first, notes inline): IsFinite cases, Clamp with NaN/inf,
panel rows refusing nan/inf/-inf/1e999/-1e999 for width and x with the
stored value unchanged, SetPosition dropping NaN/inf, Recover repairing
a planted NaN x / inf y in one run (second run 0), Sanitize falling
back on a non-finite free number.
- docs/schema.md: number coercion is finite-only. docs/smoke-tests.md:
new NUM-1 to NUM-3 theme (CLI refusal, reload, recover on a planted
SavedVariables NaN) plus index and pending sign-off rows.
- docs/test-cases.md regenerated; README Tests badge 1052 -> 1058.
Finding: PM-R-03.
Deviations: Clamp(math.huge, lo, hi) now falls back (fallback, then lo)
instead of returning hi, as the item spec says. Canvas.BuildSpec's
tonumber(rec.x) (cited in the finding) is left as is; it is not one of
the item's five sites and the write and recover guards cover it. The
smoke check is a new NUM-1..3 theme because the doc had no NUM rows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 1380d569206c5cd90904f64e9dd11daf056da683
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:28:58 2026 +0530
PM-05: Hex-encode non-ASCII bytes in frame slugs; fold UTF-8 case
Non-Latin and accented panel names collapsed to one frame name: %w is
ASCII-only, so every Cyrillic/Han/Hangul name slugged to "Panel" and
accented pairs (Aerger/Oerger with umlauts) both slugged to "rger"; the
second panel was refused for a frame-name clash (PM-R-01, PanelMaster#26).
- core/Util.lua: Util.Slugify writes every byte >= 0x80 as two upper-case
hex digits joined to the surrounding ASCII (e.g. "C384rger",
"D09ED0B1D0B7D0BED180"); ASCII names are byte-identical to before and an
all-punctuation name still falls back to "Panel". Only new frame names
(R:Create, and the backfill for a record that has none) use it; stored
rec.frameName values are never rewritten, so no anchor moves.
New Util.FoldName: string.lower plus a UTF-8 simple case fold held in a
load-time map (Latin-1 Supplement minus U+00D7, Latin Extended-A pairs,
Greek U+0391-U+03A9, Cyrillic U+0400-U+042F). Slugify comment rewritten.
- modules/Registry.lua: R:FindByName compares FoldName on both sides, so
the duplicate-name guard no longer depends on the slug colliding.
975 lines.
- tests: test_util.lua (Slugify distinctness/encoding/ASCII unchanged,
FoldName), test_registry.lua (two Cyrillic + two Han panels coexist,
Uebersicht/uebersicht refused as a duplicate and resolved, stored frame
name survives ReloadProfile), each with a red-under note; 5 were red
before the fix. test_docs.lua header comment updated.
- docs: README anchoring paragraph (hex example), data-flow.md frame-name
section, smoke-tests.md non-English section and LOC-2/LOC-3 expected
results; regenerated test-cases.md; README tests badge 1052/1052.
Deviations: FoldName also folds U+0178 (Y diaeresis) to U+00FF, and the
Latin Extended-A fold follows the block's real upper/lower parity
(even-upper 0100-0137 and 014A-0177, odd-upper 0139-0148 and 0179-017E)
rather than a naive even/odd pairing, which would merge U+0138 with
U+0139. docs/schema.md describes no slug rule, so it is unchanged. A
pre-v2 record with no stored frameName and a non-ASCII name is backfilled
under the new slug, as the item specifies. The Panels list sort
(settings/PanelEditor.lua:221, LOC-4) still folds ASCII only; out of scope.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit c9b24f92960a229c9d7e5193e6829b74cb4fb7f3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:22:32 2026 +0530
PM-04: Drop combat-held unlocks on stand-down; apply unlocks while down
A combat-held unlock (pendingUnlock / pendingPanels) survived NS.StandDown.
PLAYER_REGEN_ENABLED is unregistered while the addon is down, so after a
re-enable the next unrelated combat exit unlocked every panel (PM-R-06).
- modules/Unlock.lua: new U:DropPending(reason) empties both held queues
(the held-request half of dropAllUnlocks, factored out as dropHeld, so
NS.State.unlockedPanels is untouched), returns the count and writes one
[Unlock] line "dropped %d held unlock(s): %s" when n > 0. Both combat
gates (SetUnlocked and SetPanelUnlocked) stand aside while the addon is
down (NS.IsAddonEnabled() false or NS.Lifecycle:IsDown()), so an unlock
then applies at once instead of queueing. ForgetPending is unchanged.
- core/LifecycleSetup.lua: NS.StandDown calls
NS.Unlock:DropPending("stood down") (guarded) before unregistering
PLAYER_REGEN_ENABLED; the header comment no longer claims there is
nothing to defer while down.
- settings/Slash.lua: the unlock-verb read-back comment notes the
stood-down case.
- tests/test_disabled.lua: Disabled 11 (held unlocks dropped by
/pm disable, no unlock at a later combat exit) and Disabled 11b (while
stood down an unlock applies at once, nothing queued), each with a
red-under note; both were red before the fix.
- docs: data-flow.md, ARCHITECTURE.md (stand-down section), debug.md
(new drop line), smoke-tests.md COMBAT-3b, regenerated test-cases.md,
README tests badge 1045/1045.
Deviations: the "apply at once while down" rule is applied to the
per-panel gate (SetPanelUnlocked) as well as SetUnlocked, since a
per-panel hold queued while down would otherwise survive to the same
later combat exit. Slash.lua:389 in the item is now the comment at
:412 (the unlock verbs' read-back note); that is the one edited.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit ef02f20d3e8b814b5eb23801e9bc6cbd10ba380d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:17:09 2026 +0530
PM-03: Confirm the editor's per-panel Delete and Reset, naming the panel
The Panels editor's Delete and Reset buttons sit side by side and both
acted on one click, and the Registry comment claimed R:Reset "is
confirmed on its own control", which was false (PM-R-05).
- settings/Slash.lua: two new StaticPopupDialogs beside
KA0S_PANELMASTER_DELETEALL: KA0S_PANELMASTER_DELETE ('Delete the panel
"%s"? This cannot be undone.') and KA0S_PANELMASTER_RESET ('Reset the
panel "%s" to how a new panel starts? Its size, position, textures and
colors are lost; its name is kept.'), Yes/No, timeout 0, whileDead,
hideOnEscape, showAlert, preferredIndex 3. OnAccept(self, data) hands
the panel id to the editor.
- settings/PanelEditor.lua: pageAction.delete/reset now show the popup
with StaticPopup_Show(name, rec.name, nil, rec.id). The old bodies moved
into E.AcceptDelete / E.AcceptReset, which resolve the id afresh with
Registry:Get (a panel deleted while the popup was up is a no-op);
selectedID is still cleared before Delete and Reset's error print is
kept. Headless fallback (no StaticPopup_Show) acts at once, as
doDeleteAll does.
- Tooltips on Reset and Delete say they ask first; the Registry.lua
ResetPositions comment now names the KA0S_PANELMASTER_RESET popup.
- tests: wow_mock's StaticPopup_Show records the 4th (data) argument as
__popupData. Four new test_panel cases (written first, red: no popup
shown / dialog missing): Delete asks and deletes nothing until
OnAccept; Reset asks and a changed width survives until OnAccept;
OnAccept with a since-deleted id is a no-op for both; the headless
fallback acts at once. The two existing editor-delete rebuild cases
and the panels-page "acts follow the picker" case now accept the popup.
- Docs: settings-panel.md Reset/Delete rows quote the confirms;
smoke-tests.md FRAME-28, PANEL-16 and PANEL-24 (already in Pending
sign-off) cover the confirms; docs/test-cases.md regenerated and the
README Tests badge moved to 1043/1043.
Finding: PM-R-05.
Deviations: E.AcceptDelete clears selectedID only when it is the panel
being deleted (it used to clear unconditionally), since the popup can be
accepted after the picker has moved to another panel. The CLI
`/pm delete` is unchanged (it never confirmed; the finding is about the
editor buttons).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit adc75e8e100839a3110d137748a4d2f9909010aa
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:11:58 2026 +0530
PM-02: address review — sweep session rows from OnProfileReset so both resets match
Review of 3060ec4: the session-only sweep lived only in Sl:DoResetAll, so
Reset all settings relocked the screen and closed the debug console while
Profiles -> Reset Profile (AceDBOptions, db:ResetProfile() directly) did
neither. options-ui-§12 says the two MUST be the same act, and the docs and
the Reset all tooltip ("the same thing Profiles -> Reset Profile does")
promised an equivalence the code no longer kept.
- core/Database.lua: the sweep moves into the OnProfileReset handler, the
one place both resets arrive. After the reload it drives every row
flagged sessionOnly (or at state.debugConsole) through ApplyDefault,
inside its own bracket that reports profileReset, so the reset keeps its
one [Set] reset profile line whether it is nested in DoResetAll's bracket
or alone (the Profiles page).
- settings/Slash.lua: DoResetAll drops its own sweep. It now reads "the
reset was logged" from the handler having consumed the snapshot, so
anything raising after the handler logged (reload, sweep) marks the
bracket without a second line, and a reset that raised before it still
logs "reset all ... (stopped by an error)".
- tests/test_debuglog.lua: new case, written first and red ("Profiles ->
Reset Profile left the screen unlocked"): unlock, show the console, call
db:ResetProfile() directly; asserts relocked, Lock frame reads locked,
console hidden, exactly one [Set] line, bracket closed.
- Docs and comments: settings-panel.md (Lock frame / Debug console rows and
the sweep paragraph now name both resets, so the tooltip is true again),
schema.md, smoke-tests.md PANEL-8a (adds the Reset Profile pass),
OptionsSetup.lua, Schema.lua, test_slash.lua red-under notes;
docs/test-cases.md regenerated, README badge 1039/1039.
Gates: 1039 passed / 0 failed / 1 skipped; luacheck 0/0; lizard no
function above CCN 15; largest touched file settings/Schema.lua 849 lines.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 3060ec4b0f301a76fdcb483c745e16ad9dc6ae1c
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 20:04:24 2026 +0530
PM-02: Sweep the session-only rows on Reset all settings
Sl:DoResetAll only bracketed db:ResetProfile(), so the session-only rows
(Lock frame at state.locked, Debug console at state.debugConsole)
survived a global reset: the screen stayed unlocked, the console stayed
open, and the next panel was born draggable. options-ui-§12 requires
those rows restored row by row, and the OptionsSetup.lua comment that
defended the omission is what the rule forbids.
- settings/Slash.lua: after a reset that landed, and still inside the
existing reset/all bulk bracket, DoResetAll walks S.ProfileRows() and
drives every row flagged sessionOnly (or at state.debugConsole)
through NS.SchemaRuntime.ApplyDefault, so each goes through its own
set: SetUnlocked(false) relocks and drops held unlocks, DebugLog:Hide
closes the console. Found by flag, so a future session-only row is
swept automatically. No DB write; the bracket tallies the writes, so
the reset still logs its one [Set] reset profile line. A sweep that
raises marks the bracket and re-raises without a second line.
- tests/test_slash.lua: the blast-radius case (PM-A-02), written first
and red ("the reset left the screen unlocked"): three panels, a second
profile, unlocked, console shown, /pm resetall then the popup's
OnAccept; asserts zero panels, profile list and active profile
unchanged, unlocked false, state.locked true, console hidden, exactly
one PanelsChanged, and a panel made afterwards is not mouse-enabled.
- Comments rewritten: OptionsSetup.lua (the "deliberate rather than an
oversight" block now describes the sweep), DoResetAll's doc comment,
and two Schema.lua comments that said the reset leaves the rows alone.
- Docs: settings-panel.md (Lock frame / Debug console rows and a sweep
paragraph), schema.md, smoke-tests.md PANEL-8a; docs/test-cases.md
regenerated and README Tests badge moved to 1038/1038.
Findings: PM-A-01, PM-A-02.
Deviations: the item's test asserts S:Get('settings.locked'); the
composed row is wired onto path state.locked (settings/Schema.lua), so
the test reads S:Get('state.locked'). Profiles -> Reset Profile still
calls db:ResetProfile() directly and does not sweep the session rows;
the item scopes the sweep to Sl:DoResetAll.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 58892c177a579e7f2755315eeb9d118e1a4e190a
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 19:58:59 2026 +0530
PM-01: address review — restore doDeleteAll's header comment placement
Sl.ActiveProfileName had been inserted between doDeleteAll's header
comment and doDeleteAll itself. Move it (with its own comment) above the
"confirm-gated wipe" block so that block heads doDeleteAll again.
Comment/ordering only; no behaviour change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 5c42c0d59142a15452884ae651c4a837e2bac5a1
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 19:56:45 2026 +0530
PM-01: Name the active profile in the delete-all confirm
KA0S_PANELMASTER_DELETEALL said "on this character", but panels live in
db.profile and AceDB's "Default" profile is shared, so accepting it wiped
every character's layout on that profile. The text is now a format string
naming the profile ("Delete ALL Ka0s Panel Master panels in the profile
\"%s\"? Every character using this profile loses them. ...") and both
entry points pass the profile as text_arg1 through the new
Sl.ActiveProfileName() (NS.db:GetCurrentProfile(), guarded, else
"current"): doDeleteAll in settings/Slash.lua and the Panels page
Defaults closure in settings/Panel.lua. Headless fallbacks unchanged.
Tests (red first): tests/wow_mock.lua's StaticPopup_Show now records its
text args in __popupArgs (parallel to __popupsShown); new cases in
test_slash.lua (/pm panel deleteall) and test_panel.lua (Panels page
Defaults) assert the formatted text names "Default" and no longer says
"this character". docs/test-cases.md regenerated (1037), README Tests
badge 1035 -> 1037, smoke FRAME-29 expects the profile name.
Finding: PM-R-02.
Deviations: the item says `/pm delete all`; the actual verb is
`/pm panel deleteall`. The Panels page Defaults button raises the same
popup, so it passes the profile name too (otherwise it would show a raw
"%s"). No other "this character" panel wording exists in settings/,
modules/, core/, docs/ or README.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 9fa5c52cb32cc8738dff9627f9eaa9236724f8ed
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 19:51:11 2026 +0530
RV-PM: Re-vendor LibKa0s v1.71.0 (kit 38); back-fill v1.69.0-v1.70.0
Replace libs/LibKa0s/ and tests/_kit/ whole from the LOCAL LibKa0s tag
v1.71.0 (tag 3bf1b97, commit cb274a4) and roll the CLAUDE.md provenance
line v1.70.0 -> v1.71.0 in the same commit. Minors moved: Env 2, Slash
20 / SlashParse 2, WidgetsLineChart 3, WidgetsAutocomplete 2,
OptionsIdList 4; kit 37 -> 38 adds testkit/secrets.lua.
Regenerate docs/test-cases.md under kit 38: Totals now count only cases
that run, the diagnostics opt-out skip sits on "| Skipped | 1 |", and
Total 1035 equals the README badge 1035/1035 (PM-R-09).
Write docs/revendor/2026-10-07-v1.71.0/ (01_DELTA, 02_CANDIDATES with
every surface "not adopted in this run" per owner ruling 5, 05_SUMMARY)
and the consolidated span bundle docs/revendor/2026-10-07-v1.69.0-v1.70.0/
for the unrecorded re-vendors 4e15691 and f61f2b7, base v1.68.1
(PM-A-11). Step 3h's vendored-minus-recorded listing is now empty.
No red from the fresh payload. Recorded, not fixed here:
core/EnvSetup.lua's library-absent fallback keeps its bare
GetAddOnMetadata rung (PM-09 owns it). docs/testing.md:242 is left for
PM-11.
Deviation: the item names the kit's installer "Kit.installIssecretvalue";
the kit ships it as Kit.installSecretValue, and the bundle uses the
real name.
Findings: PM-A-11, PM-R-09.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit 9e41cd97bf5199c49b9993945e4d74572d673739
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 16:51:59 2026 +0530
RA-00: record 2026-10-07 standards audit and review bundles
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RoCGDEVoaSdzqSHHsASPMi
commit f61f2b7b33b32444fa48b94edddff96334b80d06
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 11:38:41 2026 +0530
chore: re-vendor LibKa0s v1.70.0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4XiFWzQrd9ee19aBpkVmt
commit f61f2b7b33b32444fa48b94edddff96334b80d06
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Oct 7 11:38:41 2026 +0530
chore: re-vendor LibKa0s v1.70.0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4XiFWzQrd9ee19aBpkVmt
commit 4e15691e16bd6f09a31dec1167983938c0b48379
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Oct 6 14:52:47 2026 +0530
chore: re-vendor LibKa0s v1.69.0 (kit 37; adds the line chart widget)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4XiFWzQrd9ee19aBpkVmt
commit cbc3743f5703e2ba0add7c3a47aebbab8d1f3e76
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Oct 4 15:13:49 2026 +0530
DC-REV-01: re-vendor LibKa0s v1.68.1 (kit 36; rename-only)
Both payloads copied whole from the local annotated tag v1.68.1 (9000cbd).
No LibStub minor moved; tests/_kit moves to revision 36, which names the
dev-copilot plugin's commands where revision 35 named /wow-addon:*. The
CLAUDE.md provenance line rolls v1.68.0 -> v1.68.1 and docs/testing.md's
"kit revision 35" moves to 36 in the same commit.
No blockers (no major moved, no __Attach site). Zero adoption candidates,
no interview, no issue filed; the Widgets/Perf/Pool/Item declines stay
settled. Bundle: docs/revendor/2026-10-04-v1.68.1/.
Gate: tests 1035/0/1 of 1036 (vendor-sync green), luacheck 0/0 in 71
files, lizard sighted max CCN 15. No version bump.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MsGF7a1YxB69EsRnBC8rcE
commit cbc3743f5703e2ba0add7c3a47aebbab8d1f3e76
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Oct 4 15:13:49 2026 +0530
DC-REV-01: re-vendor LibKa0s v1.68.1 (kit 36; rename-only)
Both payloads copied whole from the local annotated tag v1.68.1 (9000cbd).
No LibStub minor moved; tests/_kit moves to revision 36, which names the
dev-copilot plugin's commands where revision 35 named /wow-addon:*. The
CLAUDE.md provenance line rolls v1.68.0 -> v1.68.1 and docs/testing.md's
"kit revision 35" moves to 36 in the same commit.
No blockers (no major moved, no __Attach site). Zero adoption candidates,
no interview, no issue filed; the Widgets/Perf/Pool/Item declines stay
settled. Bundle: docs/revendor/2026-10-04-v1.68.1/.
Gate: tests 1035/0/1 of 1036 (vendor-sync green), luacheck 0/0 in 71
files, lizard sighted max CCN 15. No version bump.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MsGF7a1YxB69EsRnBC8rcE
commit 8d96fe34fbd35dd77fef8327d389be34bd631936
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Oct 4 14:32:58 2026 +0530
DC-REN-02: automated-tests README names /dev-copilot:bump-version as the release-gate evaluator
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MsGF7a1YxB69EsRnBC8rcE
commit 8d96fe34fbd35dd77fef8327d389be34bd631936
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Oct 4 14:32:58 2026 +0530
DC-REN-02: automated-tests README names /dev-copilot:bump-version as the release-gate evaluator
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MsGF7a1YxB69EsRnBC8rcE
commit 83a9390c5a8e4a112335c126cc0eea7d6db60083
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sun Oct 4 14:21:13 2026 +0530
DC-REN-01: point slash-command references at the dev-copilot plugin
The wow-addon plugin merged into dev-copilot (v2.0.0). Live docs and
test comments now name /dev-copilot:* (wow-* for the addon-only
commands). No assertion changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MsGF7a1YxB69EsRnBC8rcE
commit 0d80c0de722a6e3b18d3905fe5de30a66b868844
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Sat Oct 3 12:48:06 2026 +0530
SD-FIN-01: sync docs (diagnostics in the ARCHITECTURE verb list)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GBEbct8mwzFqZSG1oTkv8f
commit 8f7c4c7e1b3fc9f2533cb11a6683f27485dee87d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 23:37:30 2026 +0530
TP-PM-01: re-vendor LibKa0s v1.68.0 (WidgetsDragHandle 4)
Both payloads copied whole from the local tag v1.68.0; only
libs/LibKa0s/WidgetsDragHandle.lua differs (DRAG_MINOR 3 -> 4, the
optional tooltipPlace hook). Kit revision 35 unchanged. The CLAUDE.md
provenance line rolls to v1.68.0 in the same commit.
No blockers: this addon does not look up LibKa0s-Widgets-1.0. The
tooltipPlace candidate is declined as not applicable (no drag strip;
unlock mode drags the panel frame with no tooltip), recorded in
docs/revendor/2026-10-02-v1.68.0/ with no issue filed: not a gap, and
the Widgets decline stays settled in #43.
Gate: tests 1035/0/1 of 1036 (vendor-sync green), luacheck 0/0,
lizard sighted max CCN 15. No version bump.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit 8f7c4c7e1b3fc9f2533cb11a6683f27485dee87d
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 23:37:30 2026 +0530
TP-PM-01: re-vendor LibKa0s v1.68.0 (WidgetsDragHandle 4)
Both payloads copied whole from the local tag v1.68.0; only
libs/LibKa0s/WidgetsDragHandle.lua differs (DRAG_MINOR 3 -> 4, the
optional tooltipPlace hook). Kit revision 35 unchanged. The CLAUDE.md
provenance line rolls to v1.68.0 in the same commit.
No blockers: this addon does not look up LibKa0s-Widgets-1.0. The
tooltipPlace candidate is declined as not applicable (no drag strip;
unlock mode drags the panel frame with no tooltip), recorded in
docs/revendor/2026-10-02-v1.68.0/ with no issue filed: not a gap, and
the Widgets decline stays settled in #43.
Gate: tests 1035/0/1 of 1036 (vendor-sync green), luacheck 0/0,
lizard sighted max CCN 15. No version bump.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit 9a8eddc1549d8edf9ba18535e54f0f99b5ff1e72
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 02:00:28 2026 +0530
CA-PM-01R: re-point two Panel.lua citations shifted by a04e2d4
.luacheckrc:195 now cites settings/Panel.lua:344 for the
NS.Slash:ConfirmResetAll() probe (was :343, now P:RestoreDefaults),
and docs/settings-panel.md:572 cites :383-388 for the comment above
ctx.panel.defaultsOnClick (was :380-385).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit 9a8eddc1549d8edf9ba18535e54f0f99b5ff1e72
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 02:00:28 2026 +0530
CA-PM-01R: re-point two Panel.lua citations shifted by a04e2d4
.luacheckrc:195 now cites settings/Panel.lua:344 for the
NS.Slash:ConfirmResetAll() probe (was :343, now P:RestoreDefaults),
and docs/settings-panel.md:572 cites :383-388 for the comment above
ctx.panel.defaultsOnClick (was :380-385).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit a04e2d4a93f8c9dfe5e00cc2229768300f6a3643
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 01:58:03 2026 +0530
CA-PM-01: read O.PADDING_X for the Profiles container (PanelMaster#56)
settings/Panel.lua drops its local PADDING_X = 16 and anchors the Profiles
page's AceConfigDialog container at the LibKa0s-Options-1.0 instance's
O.PADDING_X, so the container moves with the header and divider the library
draws at the same inset. No visible change: the library value is 16.
- settings/OptionsSetup.lua: the degraded stub gains PADDING_X = 0 beside
its other zero layout placeholders (D3.9).
- tests/test_surface_parity.lua: PADDING_X leaves the exemption list and
joins the zero-on-the-stub loop.
- tests/test_profiles.lua: the characterization case now also ties both
insets to ns.Helpers.PADDING_X.
- Citations re-pointed for the moved lines (OptionsSetup.lua 225->228,
256->259; Panel.lua 375->376).
- docs/smoke-tests.md: PROFILE-18, index range, Pending sign-off row.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit 113f6c4463c3a303a886f58795dba4138d717670
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 01:55:56 2026 +0530
CA-PM-01: pin the Profiles container inset (characterization)
The AceConfigDialog container on the Profiles page is anchored by hand at
TOPLEFT +16/-8 and BOTTOMRIGHT -16/+8 of the page body. Nothing pinned that
geometry. The new case builds a fresh full-library environment whose
SimpleGroup frames record SetPoint (no production seam) and asserts both
anchors, ahead of reading O.PADDING_X there (PanelMaster#56).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit 664a4417ae9c8662310b2a64ad5d9a2e0f8f317c
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 01:48:26 2026 +0530
CA-PM-NMR: re-point docs/performance.md scheduleTimer citation to OptionsSetup.lua:256
The addonName descriptor lines added in CA-PM-NM moved the scheduleTimer seam
from line 253 to 256; this citation was missed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit 74429c790fcdea0e340f66eebaf737af0d8ef5fb
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 01:45:25 2026 +0530
CA-PM-NM: Options descriptor passes addonName (LibKa0s#42)
settings/OptionsSetup.lua keeps its first vararg as `addonName` and hands
it to the LibKa0s-Options-1.0 descriptor beside parentTitle and
mainPanelName, so OptionsIdList's help-mark art resolves to this addon's
own vendored info glyph. Latent here: no list carries help yet. The
degraded (library-absent) arm returns before the descriptor and is
unchanged.
Tests first in tests/test_libka0s.lua: a comment-stripped source guard
that the vararg is kept and the lib:New descriptor passes
`addonName = addonName,` (red before the change), and an on-disk assert
for libs/LibKa0s/media/icons/info.tga. docs/test-cases.md regenerated,
README badge 1034/1034, smoke PANEL-32 added with its Pending sign-off
row, and the `212/filter` line citation moved to OptionsSetup.lua:225
(docs/testing.md; tests/test_lintconfig.lua's :203 was already stale).
Gate: 1034 passed / 0 failed / 1 skipped; luacheck 0/0 in 71 files;
vendor diff empty; sighted complexity green, max CCN 15.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit 1c826624f2f0543f34b1dd4348377228c41c346b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 01:39:39 2026 +0530
CA-PM-RV: re-vendor LibKa0s v1.67.0 (Core 10, Options 28, OptionsIdList 3)
Re-vendor from the local tag v1.67.0 (LibKa0s HEAD == v1.67.0^{commit},
clean tree; whole folders compared):
- libs/LibKa0s/: Core 9 -> 10 (MakeResizable canResize / onResizeStop /
gripParent), Options 27 -> 28 (docblock), OptionsIdList 2 -> 3 (the
addonName loaded-addon guard). Nothing added or removed upstream; payload
stays 32 files.
- tests/_kit/: unchanged (kit revision 35); diff empty.
- CLAUDE.md provenance line rolled to v1.67.0 in this commit.
No contract blocker: this addon makes no MakeResizable call and builds no
O.IdList. No consumer test broke. Candidates: addonName -> CA-PM-NM;
MakeResizable opts -> none. Record: docs/revendor/2026-10-02-v1.67.0/.
Gate: tests 1032 passed / 0 failed / 1 skipped / 1033 total (unchanged;
docs/test-cases.md regenerates byte-identical, README badge unchanged);
luacheck 0/0 in 71 files; sighted complexity pass, max CCN 15, 2161
functions; diff -r --strip-trailing-cr of both payloads against the tag is
empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit aec946d80f5a9d8120b51bc77ba0e7dc404f79cf
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 00:38:00 2026 +0530
Delete Util.CountChanged: no shipped caller since GI-PM-03
GI-PM-03 (#54) moved the bulk copy and reset onto the schema's S:CountChangedSince, which left
Util.CountChanged called only by its own test. Removes the function and its test; the inventory
and the README badge drop to 1032.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit aec946d80f5a9d8120b51bc77ba0e7dc404f79cf
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 00:38:00 2026 +0530
Delete Util.CountChanged: no shipped caller since GI-PM-03
GI-PM-03 (#54) moved the bulk copy and reset onto the schema's S:CountChangedSince, which left
Util.CountChanged called only by its own test. Removes the function and its test; the inventory
and the README badge drop to 1032.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit f9c21e2fb91040b129e98371032cfb5bf91fc1ff
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 00:13:28 2026 +0530
GI-FIN-02: record the 2026-10-02 smoke passes; doc sync
FRAME-12 and PANEL-27..31 passed in the owner's 2026-10-02 client run and leave Pending sign-off.
Doc sync re-points line citations that moved with GI-PM-02/03 (Constants, Registry, OptionsSetup)
and COERCE.media's new home in settings/PanelSchema.lua.
Gate: 1033 pass / 0 fail / 1 skip; luacheck 0/0 in 71 files; vendor diff empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit f9c21e2fb91040b129e98371032cfb5bf91fc1ff
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Fri Oct 2 00:13:28 2026 +0530
GI-FIN-02: record the 2026-10-02 smoke passes; doc sync
FRAME-12 and PANEL-27..31 passed in the owner's 2026-10-02 client run and leave Pending sign-off.
Doc sync re-points line citations that moved with GI-PM-02/03 (Constants, Registry, OptionsSetup)
and COERCE.media's new home in settings/PanelSchema.lua.
Gate: 1033 pass / 0 fail / 1 skip; luacheck 0/0 in 71 files; vendor diff empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X7u34U99DrEPHJTMx6E6u8
commit b58cba5245bbadbfdbdcc4df8b438cb06076a3c2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 14:27:54 2026 +0530
GI-PM-03: per-panel fields are instance-addressed schema rows; retire the architecture-§5 row (#54)
Option (a), host-side only (decision D8). Every per-panel field write now
goes through the one schema seam with the panel id as the instance id.
- settings/Schema.lua: S.ResolveRoot(parts, id) maps a `panel.*` path to the
Registry record (first = 2, resolved id = rec.id; "no such panel" / "panel
rows need a panel" otherwise) and leaves every other path on the profile.
NS.Schema:Set/:Get forward the id; new :SetMany. The descriptor gains a
`format` hook (panel rows render through R.FormatValue and name the panel,
built only while logging is on), `announce` and `announceBatch` (the
Registry's). S:Register resolves panel rows against C.PANEL_TEMPLATE.
SnapshotPersisted reads the profile rows. S.BulkLine is gone (no caller).
- settings/PanelSchema.lua (new, TOC after Schema.lua): one hidden,
skipRender `panel.<field>` row per C.PANEL_FIELD_ORDER field but `name`
(50 rows), normalize = the field parser (the COERCE table, moved here from
the Registry) then R.SanitizeField. S.ProfileRows / S.FindProfileRow, which
the Slash and Options descriptors and the diagnostics dump now read, so no
profile surface sees a panel row.
- modules/Registry.lua: the repair rules are one REPAIR table behind both
R.Sanitize and the new per-field R.SanitizeField. R:Set routes through
SchemaRuntime.Set; R:SetPosition(key, x, y, point, relPoint), R:Reset,
R:CopyFrom, R:FitToArtwork, R:Recover and R:ResetPositions through SetMany
with the panel id. Record-to-record writes repair rather than parse (so a
copied texture whose media addon is not loaded still copies, as before).
R.AnnounceWrite/AnnounceBatch repair the record and send PanelChanged once
per act; Recover/ResetPositions hold it back and send one PanelsChanged.
Reset still drops undeclared keys and counts them into its one line.
- modules/Unlock.lua: the drag-stop no longer writes point/relPoint raw; it
passes all four fields to SetPosition as one act.
- Debug lines: a panel field write is `[Set] panel.width = 300 on 'Alpha'`
(the library's shape); the Registry's own per-write, "moved to" and "fit"
[Panel] lines are gone. Bulk lines are unchanged.
Tests first: tests/test_panel_schema.lua's instance-row cases were red
against the old code (9 of 11: resolver x2, rows, seam write, [Set] shape,
seam routing, drag-stop act, profile surfaces, degraded load; the reset
legacy-key and boot-check cases pin behaviour that already held). The
characterization cases from the previous commit and every existing R:Set,
Reset, Copy, Recover, ResetPositions and debug-line test pass untouched.
Profile-surface shape tests in test_schema, test_slash and test_libka0s now
iterate S.ProfileRows() (what /pm list and the General page actually read).
Docs: ARCHITECTURE.md retires the architecture-§5 (fields on a panel) row
into Retired rows and describes the instance rows; schema.md gains
"Instance rows (panel.*)"; data-flow.md the field write path; module-map.md
PanelSchema.lua and its load position; debug.md the [Set] panel line;
common-tasks.md; smoke-tests.md PANEL-28 to PANEL-31 (owed); test-cases.md
regenerated; README badge 1033/1033; luacheck file count 71.
Gate: tests 1033 passed / 0 failed / 1 skipped / 1034 total (was 1022/0/1);
luacheck 0/0 in 71 files; sighted complexity pass, max CCN 15, 0 warnings.
Registry.lua 987 -> 972 lines, Schema.lua 794 -> 846, no file enters the
1000-1500 band.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ng9STmzmM1JYwT2WWSakpL
commit 962ce158fef648d708d1e5b48829d853334f4119
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 14:12:18 2026 +0530
GI-PM-03: characterization tests for every panel verb before the seam move (#54)
New suite tests/test_panel_schema.lua pins what each whole-record and bulk
verb stores and broadcasts today: R:Set, R:Set of `name` (Rename routing),
R:Reset, R:CopyFrom, R:FitToArtwork, R:Recover, R:ResetPositions,
R:SetPosition and Unlock's drag-stop. Each case counts the panel messages
sent (PANEL with its id, or PANELS), so moving the verbs onto
SchemaRuntime.Set/SetMany is provably a change of route, not of behaviour.
Green against the current code. Suite totals: 1013 passed / 1 skipped
(1014) -> 1022 passed / 1 skipped (1023). README badge, docs/test-cases.md,
the module-map test table and the luacheck file count (70) move with it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ng9STmzmM1JYwT2WWSakpL
commit 3d2cc4031ebe61af1c70a534245543f95d108dc3
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 13:59:19 2026 +0530
GI-PM-02: Frame level slider on the Position and size tab, named bounds (#15)
The per-panel `level` was settable only from the CLI. The Position and size
tab gains a "Frame level" slider (0-100, step 1) on its own half-width row
directly under Frame strata, so the two layer controls read together. It
writes through NS.Registry:Set, the seam `/pm panel <name> level <n>` uses;
no new write path.
The bounds are named once, C.MIN_PANEL_LEVEL = 0 and C.MAX_PANEL_LEVEL =
100 (options-ui-§8), beside C.PANEL_LEVEL_STRIDE, and the Registry's clamp
row and Canvas's BuildSpec read them instead of inline 0/100.
Tests first (red against the old code, 1009/4/1):
- test_panels_page: the Position and size tab builds Frame level.
- test_constants: the slider-bounds scan expects the level slider on the
named bounds (its pattern now captures the two bound arguments, so a
slider that also passes a step and tooltip is still read); a new case
pins the two constants, that the top stack (100 * stride + 7) stays under
the client's 10000 frame-level ceiling, and that Registry and Canvas
clamp by the names.
- test_registry: R:Set level 101 clamps to C.MAX_PANEL_LEVEL, -1 to
C.MIN_PANEL_LEVEL.
Docs: scope.md drops the "No per-panel strata level UI" known limitation
(the equal-level creation-order caveat moves to the Frame level row);
ARCHITECTURE.md Known Limitations Seven -> Six; settings-panel.md gains the
Frame level row; data-flow.md names the bounds; README mentions the
control; smoke-tests.md FRAME-12 sets levels on the slider with the CLI as
the alternative, new PANEL-27 (slider reads a CLI-set 7, live re-order,
clamped display), index and Pending sign-off rows updated. test-cases.md
regenerated, README badge 1013/1013.
Gate: tests 1013 passed / 0 failed / 1 skipped / 1014 total; luacheck 0/0
in 69 files; sighted complexity pass, max CCN 15, 0 warnings, 0 blind.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ng9STmzmM1JYwT2WWSakpL
commit c75d5e3aa9df8d6661c8d0656aeba50bf6c87e50
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 13:55:31 2026 +0530
GI-PM-01: delete the dead S.Installed() export (#42)
S.Installed() had no caller in core/, modules/, settings/, defaults/ or
locales/; the feature silences itself through S.Rows(S.Themes()). Deleted,
with its misleading "keeps the whole feature silent" comment, as triaged on
2026-08-07.
Tests first (red against the old code: the first detection case's new
assertNil(S.Installed) guard failed, 1010/1/1):
- every S.Installed() assertion rewritten on #S.Themes(), or dropped where
the next line already asserts #S.Themes() == 0 (:704);
- the dedicated case retitled "SunnArt: a SunnArt whose every theme is
uninstalled offers nothing", asserting #S.Themes() == 0 beside its S.Rows
assertion;
- assertNil(S.Installed, ...) guards a reintroduction.
Docs: docs/rendering.md drops the folder-gate paragraph's last sentence;
docs/test-cases.md regenerated (one retitle, count unchanged, README badge
stays 1011/1011). docs/compat-layer.md's modules/SunnArt.lua:236 citation
sits above the deleted block and still points at the AddOnFolders call.
Gate: tests 1011 passed / 0 failed / 1 skipped / 1012 total; luacheck 0/0
in 69 files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ng9STmzmM1JYwT2WWSakpL
commit 406aa2c65983594cdb2bbd6e6fd40d23ac54d084
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 13:53:34 2026 +0530
GI-PM-RV: re-vendor LibKa0s v1.66.0 (kit 35), wire test_lizard_sighted, record sighted complexity
Re-vendor from the local tag v1.66.0 (git archive, whole folders, never a
branch tip):
- libs/LibKa0s/: Widgets 11 -> 12 + WidgetsReorder 1, DebugLog 18 -> 19,
Slash 18 -> 19 + SlashParse 1, OptionsWidgets 33 -> 34, OptionsTabs 7 -> 8,
Perf 13 -> 14 + PerfSampler 1 + PerfCommands 1. Payload 28 -> 32 files.
Nothing removed upstream, nothing deleted. The TOC and both harness load
lists derive from LibKa0s.xml, so the new files need no TOC edit.
- tests/_kit/: revision 34 -> 35 (lizard_sighted.lua, test_lizard_sighted.lua).
- CLAUDE.md provenance line rolled to v1.66.0 in this commit.
- tests/run.lua wires { name = "test_lizard_sighted", dir = "tests/_kit/" }.
No contract blocker: the host parse adapter ignores Slash 19's optional
third argument and the host L carries RESET_ALL alone. No consumer test
broke on the library change.
Raw lizard command replaced by the runner's complexity suite in
DEPENDENCIES.md and docs/testing.md (CLAUDE.md quotes none).
Sighted complexity (run-automated-tests.sh --suite complexity --no-bundle):
- before: fail, lizard blind in 1 file, tests/test_slash.lua (155 of 156
functions listed): a function literal in a for-in header at :48.
- fix: hoisted into a local (the kit 35 prescription); no case changes.
- after: pass, maxCcn 15, warnings 0, blindFiles 0, 2086 functions.
Records: docs/revendor/2026-10-01-v1.66.0/ (candidates listed, not
interviewed; GI-LK-13 picks them up) and the span bundle
docs/revendor/2026-10-01-v1.64.0-v1.65.0/ for the two tags vendored
without one.
Gate: tests 1011 passed / 0 failed / 1 skipped / 1012 total (1003/0/1/1004
before; +8 kit cases); docs/test-cases.md regenerated, README badge
1011/1011; luacheck 0/0 in 69 files; diff -r of both payloads against the
tag is empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ng9STmzmM1JYwT2WWSakpL
commit 2f53fdd4138f5de840feb34cd1a5c4d230c87e66
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 02:16:03 2026 +0530
DG-PM-02: sync-docs: luacheck covers 69 files, the profile verb's and the disabled gate's refusals are library [Cmd] lines
- docs/testing.md: the scoped lint figure is 69 files (luacheck . reports
0/0 in 69), not 66, in both places it is quoted.
- docs/slash-dispatch.md: "the library logs nothing itself" was stale since
LibKa0s v1.65.0; the profile verb's refusals and the disabled gate's
refusal are each one library `[Cmd]` line through the `debug` sink.
Gate: lua tests/run.lua 1003 passed / 0 failed / 1 skipped; luacheck 0/0
in 69 files; vendor diffs empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 2f53fdd4138f5de840feb34cd1a5c4d230c87e66
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 02:16:03 2026 +0530
DG-PM-02: sync-docs: luacheck covers 69 files, the profile verb's and the disabled gate's refusals are library [Cmd] lines
- docs/testing.md: the scoped lint figure is 69 files (luacheck . reports
0/0 in 69), not 66, in both places it is quoted.
- docs/slash-dispatch.md: "the library logs nothing itself" was stale since
LibKa0s v1.65.0; the profile verb's refusals and the disabled gate's
refusal are each one library `[Cmd]` line through the `debug` sink.
Gate: lua tests/run.lua 1003 passed / 0 failed / 1 skipped; luacheck 0/0
in 69 files; vendor diffs empty.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 911e6ab33655aab535bf2b278b9faecbd4c218a2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 01:08:41 2026 +0530
DG-PM-01: adopt the library's debug lines (Slash and Lifecycle sinks, console change gate, at-enable queue), drop the duplicate host lines
LibKa0s v1.65.0 has every module log its own refusals and edges through
the host's gated sink (standard v2.73.0, debug-logging §4/§8/§9):
- settings/Slash.lua passes `debug` to the Slash descriptor: the
dispatcher's refusals (disabled gate, unknown verb, get/set/reset usage,
not found, parse/write, no default, the profile verb's) are one
`[Cmd] refused <verb>: <guard>` line each. The host logged none of
them, so nothing to delete; the /pm panel refusals stay `[Panel]`
(host-owned verbs).
- core/LifecycleSetup.lua passes `debug` to the Lifecycle descriptor and
NS.StandDown / NS.StandUp drop their own `[Lifecycle]` lines: each edge
is the latch's one `stood down: added <key> (holds: ...)` /
`stood up: released <key> (holds: none)` line.
- Options already took `debug`; the combat lock's new `Cfg` refusal
lines and the park's flush line land through it unchanged.
- core/LauncherSetup.lua passes `debugAtEnable`, and modules/SunnArt.lua's
`Sunn adapter` line goes through the new NS.DebugAtEnable (the console's
D.DebugAtEnable): both are state lines written at OnEnable with logging
off, and now land after the [Init] summary the first time logging is on.
- NS.DebugOnce keeps its (site, key, ...) call shape but its memory is the
console's D.DebugOnce: the hand-rolled seen-set and the setEnabled
re-arm are deleted, and a console Clear now re-arms it too. site and key
are joined only past the flag test, so the off path still builds nothing.
Tests: new tests/test_library_debug.lua (11 cases) pins each library line
landing in this addon's log exactly once, the silent no-edge call, the
Clear re-arm and the at-enable flush order; each case mutation-checked red
(Slash sink removed, host Lifecycle line restored, debugAtEnable removed,
Sunn line back on NS.Debug, a host seen-set restored).
Docs: docs/debug.md Coverage names the library's tags (Debug, Init, Diag,
Cmd, Lifecycle, Cfg, Launcher, and the schema seam's Set) and rewrites the
Lifecycle, Cmd, Cfg, Launcher and Artwork rows, the "Not logged here"
paragraph and the NS.DebugOnce paragraph (the Clear residual is closed);
docs/module-map.md rows; docs/smoke-tests.md DIAG-28 (a Slash refusal in
the console), DIAG-29 (a Lifecycle edge), DIAG-30 (the at-enable lines),
index and pending sign-off. docs/test-cases.md regenerated (1004 cases,
1 declared skip), README badge 1003/1003. Citations re-pointed:
.luacheckrc (DebugLogSetup.lua:164), tests/test_surface_parity.lua
(DebugLogSetup.lua:115).
Gate: lua tests/run.lua 1003 passed / 0 failed / 1 skipped; luacheck 0/0
in 69 files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit a0d81a99fc32057a24698f4e1ef9b7ce28d7c7e6
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Thu Oct 1 00:57:15 2026 +0530
DG-PM-01: re-vendor LibKa0s v1.65.0 (Slash 18, DebugLog 18 + DebugLogGates 1, Options 27, Launcher 5, Lifecycle 3, kit 34)
Re-vendor from the local tag v1.65.0 (git archive, whole folders):
- libs/LibKa0s/: Slash 17 -> 18, DebugLog 17 -> 18 with the new
secondary file DebugLogGates.lua (loaded by LibKa0s.xml), Options
key 26.1.32.1.1.6.1.7.4.1 -> 27.2.33.2.2.7.1.7.4.2, Launcher 4 -> 5,
Lifecycle 2 -> 3. Line endings normalized through .gitattributes.
- tests/_kit/: revision 34, unchanged. Nothing addon-owned lives there.
- CLAUDE.md provenance line rolled to v1.65.0.
The DebugLog degradation stub gains DebugOnce, DebugChanged, DebugForget
and DebugAtEnable (each answering false, or nothing): the parity case reads
the live instance by name, so it went red on the copy alone.
Standards reference v2.73.0: the TOC X-Standard line, README and CLAUDE.md
carry no version stamp, so nothing to change.
Gate: lua tests/run.lua 992 passed / 0 failed / 1 skipped; luacheck 0/0.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 65ebbf396803a678181ef11442aa46ebd527b0a4
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 21:55:48 2026 +0530
DL-PM-05: re-measure the layout-cap census in ARCHITECTURE.md
The debug-coverage work grew tests/test_libka0s.lua to 1160 lines and modules/Registry.lua to 987,
so the census's dated measurement is refreshed to today's numbers. Nothing is over the cap.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 65ebbf396803a678181ef11442aa46ebd527b0a4
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 21:55:48 2026 +0530
DL-PM-05: re-measure the layout-cap census in ARCHITECTURE.md
The debug-coverage work grew tests/test_libka0s.lua to 1160 lines and modules/Registry.lua to 987,
so the census's dated measurement is refreshed to today's numbers. Nothing is over the cap.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit c6773b1b0e1922f0ae9729603bb8ba4eb4a16886
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 21:19:47 2026 +0530
DL-PM-04: README uses bullets, never numbered lists (standard v2.72.0)
CurseForge's description page does not render numbered lists. The four
setup steps and the Reporting a bug steps become "- " bullets; the setup
intro says "in this order" so the words carry the sequence the numbers
did. Reporting a bug now matches the v2.72.0 fixed body with /pm.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 29dc70d10780fc389563754a1d7deb8d4648a3e2
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 19:36:32 2026 +0530
DL-PM-03: re-vendor final LibKa0s v1.64.0 (DebugLog 17, DebugLogDiagnostics 2, kit 34), diagnostics turns logging on, Diagnostics link smoke checks
Re-vendor from the local tag v1.64.0 (git archive, whole folders):
- libs/LibKa0s/: DebugLog 15 -> 17 (the orange Diagnostics link in the
console title bar, minor 16; the diagnosticsEnablesLogging descriptor
field, minor 17), DebugLogDiagnostics 1 -> 2 (a run turns debug logging
on for the session before it writes). Byte-identical to the tag export,
already CRLF.
- tests/_kit/: revision 34. test_diagnostics_contract's "lands with logging
off and leaves it off" becomes "turns it on for the session", plus the
opt-out case (a declared skip here: this addon keeps the default) and the
already-on case. Nothing addon-owned lived in tests/_kit.
- CLAUDE.md provenance line already reads v1.64.0 (same tag, re-cut).
Standards reference v2.71.0: the three places carry no version stamp, so
nothing to change.
Tests: tests/test_diagnostics.lua's read-only case restores NS.State.debug
after its RunDiagnostics, which now turns the flag on, and says why.
Docs:
- docs/debug.md: running the report (either form, or the link) turns
logging on for the session through SetEnabled(true), never off, no second
enable line; /reload turns it off. The descriptor keeps the default. The
Diagnostics link beside the Debug toggle. Coverage table rows reworded.
- docs/smoke-tests.md: DIAG-15 corrected (the report lands with logging
off; the flag after it is DIAG-27) and taken off the passed list; DIAG-21
minimum includes the link; new DIAG-26 (the orange Diagnostics link:
plain text, small gap after the toggle, click writes the report) and
DIAG-27 (diagnostics, link or /pm diagnostics or /pm debug diagnostics,
turns logging on for the session; /reload turns it off). Index and
Pending sign-off updated.
- docs/test-cases.md regenerated (993 cases, 1 declared skip); README
badge 992/992.
Gate: lua tests/run.lua 992 passed / 0 failed / 1 skipped, luacheck 0/0 in
68 files, lizard 0 over CCN 15 on the touched test file; vendor payloads
identical to the v1.64.0 export.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 751ac73d781206641df7c70cebbbee6276a446c0
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 16:57:10 2026 +0530
DL-PM-02R: slash and recover refusals logged, DebugOnce re-armed per enable, safeRun keyed per closure, citations re-pointed
Review round 1 fixes for DL-PM-02:
- /pm panel's two slash-local refusals (no such panel, unknown field) and
R:Recover's screen guard now write `[Panel] <verb> refused: <reason>`
through the Registry's refuse helper, published as R.Refuse. /pm recover
prints the reason instead of claiming every panel is on screen.
- NS.DebugOnce's seen-set is cleared on every logging enable edge (the
descriptor's setEnabled), so a recurring error lands again in each
logging session. A console Clear cannot re-arm it: the library exposes
no clear hook (library residual).
- safeRun keys NS.DebugOnce on the closure's tag, so two closures raising
one message are two lines.
- .luacheckrc probe citations re-pointed to the lines they name, and the
live-doc citations this change shifted.
Four new tests (987 -> 991), each mutation-checked red.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 22abc573b1ea2a3c683e71717bc1f2a7948d7f18
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 16:37:21 2026 +0530
DL-PM-02: debug coverage audit and fill (debug-logging v2.70.0 §8, §9)
Diagnosis lines the log was missing: the loading-screen and combat edges the
renderer acts on, the combat-held unlocks with their flush and drops, the
Registry's refusals with the guard's reason, the optional dependencies on the
[Init] line, and the errors and media fallbacks a repeating path swallows.
The render summary now says how many panels the show ladder left shown.
New NS.DebugOnce seam (gated, once per distinct key, stub on the degraded
arm) keeps the 10Hz mouseover tick, the media fallback and the page closures
to one line per distinct error. Tests pin each line and the quiet steady
state; docs/debug.md gains a tag-by-tag Coverage section.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 9f7ca360d39b4a83987c6f46bfd251e9dad5876e
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 16:07:52 2026 +0530
DL-PM-01: re-vendor LibKa0s v1.64.0 (kit revision 33), standards ref v2.70.0, resize smoke checks
Re-vendor from the local tag v1.64.0 (git archive, whole folders):
- libs/LibKa0s/: Core minor 9 (MakeResizable), DebugLog 15.1, Widgets
11.3, PerfPanel 6. Byte-identical to the tag export; archive bytes were
already CRLF, run-automated-tests.sh stays LF and +x.
- tests/_kit/: revision 33, adds mock_resize.lua. Nothing addon-owned
lived in tests/_kit (every file matched the kit), so it was replaced whole.
- CLAUDE.md provenance line rolled to v1.64.0 in this commit.
Fallout: none. The Core parity case here compares core/CoreSetup.lua's NS
publications, not the library's Core major, so it needs no MakeResizable
stub; no suite asserted on IsResizable/IsUserPlaced/GetResizeBounds.
Nothing adopted from the delta.
Standards reference v2.70.0: the three places (PanelMaster.toc X-Standard,
the README badge, CLAUDE.md Standards compliance) carry no version stamp,
so nothing to change; v2.70.0 renames or retires nothing, and no live doc
cites the anti-pattern range.
Docs:
- docs/smoke-tests.md DIAG gains DIAG-20 to DIAG-25: console grip,
minimum, size kept across close/Esc/reopen and back to 700x344 after
/reload, another addon's console independent; the copy window's grip,
minimum, size kept and default after /reload. No perf panel checks: Perf
is declined here (ratified deviation). Index row, optional-addon list
and Pending sign-off table updated.
- docs/debug.md: the console is 700x344 by default and resizable for the
session, no longer a fixed size.
Gate: lua tests/run.lua 975/0/0, luacheck 0/0 in 68 files, lizard 0 over
CCN 15; vendor payloads identical to the v1.64.0 export.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit f44cffd4a4fc2a4474cbf04a5f348794007cbfb4
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 08:52:25 2026 +0530
SP-FIN-01: re-vendor the test kit at revision 32 (LibKa0s v1.63.0, re-cut)
LibKa0s folded kit revision 32 into the unpublished v1.63.0: tests/_kit/README.md now counts eleven
consumers and thirteen copies, and Kit.VERSION is 32. Prose and the revision integer only; no code,
case or mock changes. tests/_kit/ is byte-identical to the tag again (diff -r empty), so the
vendor-sync case stays green.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit f44cffd4a4fc2a4474cbf04a5f348794007cbfb4
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 08:52:25 2026 +0530
SP-FIN-01: re-vendor the test kit at revision 32 (LibKa0s v1.63.0, re-cut)
LibKa0s folded kit revision 32 into the unpublished v1.63.0: tests/_kit/README.md now counts eleven
consumers and thirteen copies, and Kit.VERSION is 32. Prose and the revision integer only; no code,
case or mock changes. tests/_kit/ is byte-identical to the tag again (diff -r empty), so the
vendor-sync case stays green.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 26472302f0f6a1e05efde66c4edb60984ce290f0
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Wed Sep 30 01:38:04 2026 +0530
SP-FIN-01: sync docs (settings-panel names the General-page refresh on a profile switch)
The profile callbacks in core/Database.lua now end in NS.Panel:Refresh() so
an open General page re-reads in place when /pm profile switches from chat;
docs/settings-panel.md described the handler as delegating to
Registry:ReloadProfile alone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit ec052e5e0779f34b5a1e652be4c91f4298dac175
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 23:05:43 2026 +0530
SP-PM-03R: take PANEL-12 off Pending sign-off
PANEL-12 (old § 9 step 5-x, Panels opened after another addon) has a
recorded owner pass: the 2026-09-26 navrail adoption's report lists "the
PanelMaster Panels band fix (5-w, 5-x)" as passing. Step 5-x predates that
pass (3a7ac19) and PANEL-12 keeps its expectation, so under the clarified
Pending policy it is not owed. The intro now counts nine recorded passes
and cites the report for 5-x; 73d9b8b had filled only 5-w and step 16.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 79d9c9b3cf0488e9455ac33e45c3b18eb0d10c86
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 22:53:39 2026 +0530
SP-PM-03R: correct the General tab layouts and FRAME-12's level step
PANEL-2 and PANEL-3 now expect 7 Master controls rows, with Minimap
button alone on a fourth line above the reset pair, and the Editing
lines [Show names | Snap to grid], [Grid size | Unlock outline
thickness], which is how the flow engine pairs them. FRAME-12 sets the
levels with /pm panel <name> level, since the Panels page has no Level
control. The three checks get their own "corrected" rows in Pending
sign-off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 90784833c9ba6bdf8ce7805d36b5658d9ebe5217
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 22:36:03 2026 +0530
SP-PM-03R: take the six diagnostics-plan passes off Pending sign-off
DIAG-14 to DIAG-17, DIAG-19 and DEGRADED-6 have a recorded owner pass
(2026-09-26, PM-S1 to PM-S5, PM-S7, PM-X1 in the diagnostics plan's
report) and unchanged expectations, so they are not pending. DIAG-13 and
DIAG-18 move to a "partly run" row, and the intro names all eight
carried-over passes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit a2a84d2f4f93230762ade63ac5ae3a23667c38c9
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 22:25:43 2026 +0530
SP-PM-03R: close the third review's smoke defects and apply the Pending policy
The five review defects:
- STATE-6 switches between two disabled profiles on the Profiles page. The
`/pm profile` route prints a line, and PROFILE-15's enabled profile brings
the panels up.
- LOC-3 compares against `/pm panel wide` resolving `Wide`. `/pm panel chat bg`
reads only `chat`. Its second half names both refusals, because the
frame-name check refuses a second `übersicht` even when the name guard does
not fold it.
- LOC-1 hovers Panel name on the General tab, not in the band.
- STATE-3 no longer lists `/pm profile`. PROFILE-15 owns that assertion.
- Before you start and the ART lead-in say that ART-13 needs no Sunn folder.
Re-verifying every chat and console string against the code found six
more mismatches:
- INSTALL-7 expected a `[Preview]` line. The sweep runs at load, when
logging is off.
- INSTALL-8 read the `[Init]` line in chat. It goes to the console only.
- DEGRADED-4 expected the library's green `ON` ack. The stub prints
`debug logging is on`.
- LOOK-6, DEGRADED-5 and DEGRADED-14 read font dropdowns that PanelMaster
does not have, and a degraded install has no settings panel. The media
checks are now LSM IsValid dumps.
- STATE-3's disabled line sits under the help header.
- LAUNCH-11 and FRAME-18 quote the exact echo and tooltip.
Pending sign-off now follows the clarified policy. The old suite recorded a
result for PANEL-11 and PANEL-25 only, so every other check is listed, with
separate rows for the old NOT YET RUN items, the new checks and every
corrected expectation.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 096ac332ce151cb332fa91cbb1123b7dcbab0c1b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 21:31:02 2026 +0530
SP-PM-03R: fix the second review's smoke-suite defects
The review found checks whose steps cannot be run as written. ART-12 used
`/pm panel set <name> ...`, which reads `set` as the panel name. FRAME-9 used
`/pm panel Wide reset`, which is not a field or an action. Both now use forms the
client accepts: `/pm panel <name> artFill SQUISH` and the General tab's Reset button.
COMBAT-4 and PROFILE-10 could not be reached in the client. The Panels and Profiles
pages sit under the combat cover and `/pm profile` refuses in combat, so no route
queues a per-panel unlock or switches profile mid-combat. COMBAT-4 is dropped
because tests/test_panels_page.lua and tests/test_media.lua already assert it.
PROFILE-10 had no headless cover, so this adds a test_profiles.lua case: a switch
drops a queued per-panel unlock and keeps a queued global one. The case fails when
Registry's ForgetPending call is removed. The later PROFILE and COMBAT checks are
renumbered to keep the index contiguous. No ID has shipped yet.
Also: DEGRADED-13 now names the real divergence point (after the closing parenthesis,
since DEGRADED-3 continues with a semicolon). Before you start says `/pm panel` steps
need a one-word name and lists LOOK-13's LSM-texture addon. LAUNCH-3 and PROFILE-14/15
are reflowed. The test inventory and README badge are updated to 975.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit eab7097a3f410cb2b0a8bdf9cf531b07e735340b
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 21:09:06 2026 +0530
SP-PM-03R: fix the review's six smoke-suite defects
The independent review of SP-PM-03 found checks that no longer tested
what the old steps did. DEGRADED-13 now compares the four lines built
from NS.LIBKA0S_MISSING (DEGRADED-3, 4, 7, 11), not DEGRADED-6, which
has no cause clause. LAUNCH-3 drops the hide-entry assertion that
LAUNCH-10 owns. PROFILE-15 switches away before the single-quote form,
so it expects a switch and not "Already on". DEGRADED-14 restores the
round-trip proof from old § 14 step 9b: the Ka0s media names return
and a panel naming one draws it again. Before you start now lists
LOOK-23's other-class character and the other-Ka0s-addon comparisons
in DIAG-1, PANEL-3, PANEL-4 and PANEL-14.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit ade48adc48d78c981f95bc1ceaf5cd7d2fcc9112
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 20:55:56 2026 +0530
SP-PM-03: rewrite docs/smoke-tests.md as one evergreen, theme-grouped suite
The old suite had grown to 1233 lines across 22 numbered sections, with
adoption sign-off batches, change-history narration and the same behavior
checked in several places (profile switching in seven, reset-all in eight).
It is now 864 lines of 225 checks under 14 themes with stable IDs
(INSTALL, SLASH, FRAME, LOOK, ACCENT, ART, PANEL, PROFILE, STATE, COMBAT,
DIAG, LAUNCH, DEGRADED, LOC), each ending on a Result line.
Every in-client behavior the old doc checked is kept once. Stale claims are
corrected rather than carried: the degraded install's unlock/lock now print
the library-absent line, the CLI clamps an out-of-range outline thickness
instead of refusing it, the console draws icon marks rather than words, and
the General Defaults tooltip no longer claims the panels are untouched.
PROFILE-12 to PROFILE-18 and DEGRADED-12 cover the new /pm profile verb.
The never-run owner checks (M4-01, M4-05, M4-18, M4-19, M4-22, the session-3
General-tab acts and the non-English pass) are listed under Pending sign-off
with their new IDs. The Non-English client section keeps what
tests/test_docs.lua pins, and cites ConsumableMaster LOC-1 and KickCD LOC-1.
The prose pointer in tests/test_libka0s.lua now names DIAG-2.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 0603fa22909de7883479ed890b64caf315af5fde
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 12:54:05 2026 +0530
SP-PM-02R: re-point the TOC's Slash load-order citation to 630-634
The TOC comment explaining why Slash.lua loads after Schema.lua cites the
descriptor lines that capture NS.SchemaRuntime's members as values. The
profile COMMANDS row, the liveVerbs copy loop, the stub's CliProfile and
ProfileSwitch and the new descriptor fields added in SP-PM-02 pushed those
lines down 28, so 601-605 pointed into the degraded-arm run code. They now
sit at settings/Slash.lua:630-634.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit 9bceeabccc99a9473ff368f858919d38a5bb1089
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 12:39:42 2026 +0530
SP-PM-02: /pm profile via CliProfile
Adds the profile verb the 2026-09-29 rollout gives every Ka0s addon (owner
decisions D1-D3): the row is this addon's, next to resetall, and what it does
is LibKa0s-Slash-1.0 minor 17's CliProfile. Bare /pm profile lists the stored
profiles with the current one marked; /pm profile <name> switches to an
existing profile only (quotes stripped, case and spaces kept), refuses an
unknown name with the list instead of letting AceDB create it, and refuses a
switch in combat. The switch is AceDB's SetProfile, so the existing
OnProfileChanged handler does the latch, the panel rebuild and the one
[Profile] line, exactly as a Profiles-page switch does.
- Descriptor: profiles = function() return NS.db end, and liveVerbs is now a
copy of lib.LIVE_VERBS plus "profile", so the verb answers while disabled
(a player can switch to a profile where the addon is on). Sl.ALWAYS_LIVE
reads the same array; the thirteen-verb pin becomes thirteen plus profile.
- Degraded stub carries CliProfile and ProfileSwitch on route (b), each
printing the library-absent line, because the live NS.Slash republishes
both and the parity case compares the two tables.
- The profile handler now refreshes an open General page (options-ui-§11):
/pm profile can land with it on screen, a Profiles-page switch never could.
- The description routes through NS.L (localization-§2), so enUS.lua lists a
second key.
- Tests first: the row and verb order, list, switch plus handler, already
current, unknown and did-you-mean with nothing created, quotes, combat,
disabled switch (Disabled 7d) and the degraded stub. The AceDB fake gains
GetProfiles/SetProfile. 964 -> 974; docs/test-cases.md and the README badge
regenerated.
- Docs: ARCHITECTURE (21 verbs, live set, localization row), slash-dispatch,
profiles, module-map, localization, testing, README usage and FAQ.
Gate: 974/974, luacheck 0/0 in 68 files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
commit ada4eeeb9ad2a9bea35dc0f35f6b0aea8c8c2ac4
Author: Tushar Saxena <tushar.saxena@gmail.com>
Date: Tue Sep 29 12:27:01 2026 +0530
SP-PM-02: re-vendor LibKa0s v1.63.0
Copies both payloads whole from the local v1.63.0 tag (git archive, never a
working tree) so the profile verb has its library half: Slash minor 16 -> 17
adds the profiles descriptor field, CliProfile, ProfileSwitch, ProfileNames and
nine PROFILE_* strings. Every other file and the kit (revision 31) are
unchanged. The CLAUDE.md provenance line moves in the same commit because
tests/test_vendor_sync.lua compares the payloads against the tag it names.
No host change was owed: the Slash parity case compares two NS.Slash tables
this addon builds, so the two new instance members do not reach it until the
host republishes them. The bundle in docs/revendor/2026-09-29-v1.63.0/ records
the delta and that only the Slash minor 17 profile surface is adopted, by the
next commit of this item.
Gate: 964/964, luacheck 0/0 in 68 files.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R83GGMfYm2GoXLBwdCsbUH
This mod has no additional files