Document JobGauge padding, private cached fields, and two previously-missing fields

[RoseOfficial]

Summary

Exhaustive vtable analysis of all 21 JobGauge subclasses. For each gauge, every overridden vtable function was inspected to determine which bytes are read or written. Bytes that are confirmed to never be accessed by any override are documented as padding. Additionally, several previously-undocumented fields have been named and typed.

---

Methodology

For each gauge:

1. The full vtable listing was pulled to identify all overrides.
2. Init, Tick, and ReceivePacketValues were checked to establish which bytes are packet-sourced vs. client-computed vs. untouched.
3. Every other job-specific override was checked for gauge field accesses.
4. ProcessDeferredReplaceAction was confirmed as a safe skip — it does not read gauge fields in any gauge.
5. JobGaugeManager::ChangeGauge was analyzed to confirm exact zeroing ranges per gauge, providing additional corroboration of struct data sizes.

---

Changes

New public fields

NinjaGauge.NinjutsuActivated (0x0C, byte)
Previously noted as an unknown with a commented-out stub. Confirmed by analysis: Init zeroes this byte; ReceivePacketValues never writes it; it is set to 1 locally when a mudra sequence is in progress (confirmed via ProcessDeferredReplaceAction). This is a client-local flag, not server-sourced.

AstrologianGauge.ActiveSpell (0x0E, AstrologianSpell)
A previously undocumented server-sourced byte at 0x0E. ReceivePacketValues copies a QWORD covering 0x08–0x0F from the packet; vf6 (the action upgrade resolver) reads this byte and maps values 1/2/3 to action rows for Fall Malefic (29246), Aspected Benefic (29247), and Gravity II (29248). A new AstrologianSpell : byte enum has been added to JobGaugeEnums.cs for the three values.

New private cached fields

These fields are written only by Init (using sub_14080B560 for action sheet row lookups or direct computation) and are never exposed to consumers. They are documented as private to make struct layout explicit and to prevent future confusion about "unknown" gaps.

BlackMageGauge._betweenTheLinesRow (0x20, nint)
Action sheet row pointer for Between the Lines (Action 7419), cached by Init. Used by vf10 and vf11 for action upgrade and cost resolution.

ReaperGauge._regressRangeSquared (0x20, float)
Precomputed Action.Range² for Regress (Action 24403), cached by Init. Allows external code to perform squared-distance comparisons without a square root.

AstrologianGauge._retrogradeRangeSquared (0x20, float)
Same pattern as Reaper — precomputed Action.Range² for Retrograde (Action 41507), cached by Init.

RedMageGauge._enchantedRiposteRow through _enchantedRipostePvPRow (0x10–0x48, 8× nint)
Eight action sheet row pointers cached by Init for PrimaryCostValue lookups on the eight enchanted actions (7527, 7528, 7529, 7530, 16528, 29685, 29688, 41488). These explain the large gap between ManaStacks at 0x0A and the end of the struct.

Padding documentation

// offset: padding comments added to every gauge where gap bytes are confirmed. Each comment reflects the specific reason the bytes are padding:

Gauge	Ranges	Confirmation
ScholarGauge	0x0D–0x0F	Within WORD copy range; never read
AstrologianGauge	0x0B–0x0D, 0x0F, 0x10–0x1F, 0x24–0x2F	QWORD copy range ends at 0x0F; ChangeGauge zeroing; no vtable reads
SageGauge	0x0D–0x0F	Within WORD copy range; never read
BlackMageGauge	0x0E–0x0F, 0x10–0x1F, 0x28–0x2F	Outside DWORD+WORD copy ranges; ChangeGauge zeroing; no vtable reads
RedMageGauge	0x0B–0x0F	Outside DWORD copy range; never read by any override
PictomancerGauge	0x09, 0x0D–0x0F	Within DWORD/WORD copy ranges; never read
BardGauge	0x0A–0x0B	Within bulk QWORD copy; never read
DancerGauge	0x0F	Not packet-sourced; never read
DragoonGauge	0x0D–0x0F	Within WORD copy range; never read
NinjaGauge	0x09, 0x0B, 0x0D–0x0F	Within DWORD copy range; never read
SamuraiGauge	0x08–0x09, 0x0E–0x0F	Within DWORD copy range at 0x08; never read at 0x0E
ReaperGauge	0x0E–0x1F, 0x24–0x2F	ChangeGauge zeroing; no vtable reads
DarkKnightGauge	0x0E–0x0F, 0x12–0x17	Within bulk QWORD copy; ChangeGauge zeroing; no vtable reads
PaladinGauge	0x09, 0x0D–0x0F	Within DWORD/WORD copy ranges; never read
WarriorGauge	0x09–0x0F	Not packet-sourced; never read
GunbreakerGauge	0x09, 0x0D–0x0F	Within DWORD/WORD copy ranges; never read

[wolfcomp]

So a few points:

1. You didn't understand what I posted in Fix FishingEventHandler AtkEventInterface secondary vtable address #1755 when it comes how addresses are located.
2. Introducing padding documentation just for the sake of completeness is wastefull time better spent elsewhere.
3. Clearly you don't understand how to document things properly with the project as [FieldOffset(0x<offset>)] private nint _<name>; isn't something that is done in the project unless the field is of:

• Padding in sequential structs
• Enumerator structs

4. If you went through the other functions you would try to document them with virutal functions but clearly not if you are using AI to do this.
5. The summary reaks of AI more specifically Claudes LLM

With all that in mind come back to this when you stop using AI for most of your job.

[RoseOfficial]

The two public fields (NinjutsuActivated and ActiveSpell) are real findings regardless of your other objections. Happy to resubmit those cleanly if you're open to it.