[logicallysynced]

Pulls FFXIVClientStructs forward to the 7.57 snapshot (f0741b760 → d6397a76b, 333 commits spanning 7.56 hotfix2 and 7.57, game build 2026.10.06) and Lumina to 7.7.1. Lumina.Excel stays at 7.5.0 — that's already the newest release. Supersedes dependabot #135.

7.57 deprecated three members we read. In every case FCS left an [Obsolete] alias at the identical offset, so the reads were still returning correct bytes — but the aliases will be deleted upstream eventually, and they were the only CS warnings in the build.

• ActionBarSlotNumberArray.ActionType → Type at offset 0. Pure rename; the mapper only ever wanted the offset.
• BlackMageGauge.EnochianFlags → [BitField] EnochianActive / ParadoxActive / AstralSoulStacks over a private BitFieldD byte at 0x0D.
• SummonerGauge.Attunement → [BitField] AttunementType / AttunementCount over a private BitFieldE byte at 0x0E.

The two gauge replacements are bit accessors over a private backing field, so those offsets now come from FieldOffsetReader string-name lookups — the same mechanism already carrying _materia, _status and _attributes. JobResourceResolver keeps doing its own masking and the surfaced values are unchanged, so this is behaviour-neutral. Both new lookups get integrity tests per the DEPENDENCY.md contract, plus one pinning BitFieldD=0x0D, BitFieldE=0x0E and Type=0 so an upstream layout move can't quietly relocate a live memory read instead of failing a test.

Reading those new [BitField] declarations turned out to be worth more than the migration itself. They're an authoritative statement of the bit layout, and our hand-rolled masks match exactly: BLM Enochian at bit 0, Paradox at bit 1, AstralSoul at bits 2-4; AST Card1/2/3 at bits 0-3 / 4-7 / 8-11 with CurrentArcana at 12-15. Those two gauges are now cross-validated against upstream rather than taken on faith, which is a small dent in the open "verify the gauges in-game" pile.

On the checks that compile and tests can't do for a range this size: no [StaticAddress] signature and no declared size changed for any of the 21 scanner-key types, and every private field we resolve by string name survived intact. Of the structs we depend on, PlayerState changed but none of our fields moved, Utf8String only altered a member-function signature we never call, and AstrologianGauge.Cards is still a short at 0x08 — FCS added bitfield accessors beside it and deprecated a CurrentCards helper we don't use.

Build is warning-free again and 228 tests pass, built through build.ps1 so it's compiled against real 7.57 structs rather than a cached DLL.

Not covered here: no live in-game run. The gauge bit layouts are now verified against upstream's own declarations, which is stronger than before, but reading them off a running client is still the real check.

Three unrelated dependabot PRs are also open and untouched by this — #133/#134 (NLog 6.2.1) and #136 (ILRepack 2.0.48). Say the word and I'll fold those in too.

🤖 Generated with Claude Code

[logicallysynced]

Scope grew after the dependency bump landed, per your go-ahead on both counts.

Beastmaster now has a real gauge. 7.57 added BeastmasterGauge to FCS and wired it into JobGaugeManager as a union member like every other gauge — the thing that was missing when BST support went in at 9.2.2. All nine fields are mapped (0x08–0x10): TP, pet TP, pet TP at last use, active battlehorn, instinctual combo state, affinity, chain count, plus the Kinship and Instinct bytes. Those last two are single bytes FCS splits with [BitField], so the resolver surfaces the raw byte alongside the decoded halves — Kinship into Battlehorn (bits 0-3) and Classification (4-7), Instinct into Natural (0-1) and Mastered (2-3). Keeping the raw byte means you can re-mask if the game turns out to use spare bits. BeastmasterAffinity is mirrored with a test asserting the values and member count still track FCS, so an upstream addition can't silently mislabel a pet.

The three remaining dependabot PRs are folded in: NLog and NLog.Schema 6.2.1 (the schema regenerates substantially) and ILRepack 2.0.48. Supersedes #133, #134, #136.

One gap worth calling out, because it was a case of my comments claiming more than my tests delivered. I'd added tests pinning FCS's [BitField] declarations and named them "MatchResolverMasks" — but nothing connected them to the masks the resolver actually ran. Changing (kinship >> 4) & 0xF to >> 3 would have left FCS untouched and every test green. The bit positions now live once in GaugeBitField, read by both the resolver and the tests, so the loop closes in both directions: positions asserted against FCS, and the extraction asserted to do what those positions describe. I verified it by renumbering a field to bit 3 and confirming both layers fail. That refactor also absorbed the hand-written shifts for BLM Enochian, SMN Attunement and the AST cards — previously correct but unverifiable.

Also: the harness now prints a BST eye-check line with the Kinship and Instinct bytes in hex and binary next to the decoded values, so a mismatch against the in-game gauge points at the mask rather than the offset.

242 tests, zero compiler warnings, CI green, built through build.ps1 against real 7.57 structs. Version 9.2.5.

Still unverified in-game — nothing here has been read off a live Beastmaster. The harness line is the check.