Bounded exposure-default compatibility:6891 to ED5

This is a read-only comparison of local PE files. It neither opens a game process nor changes Native, production controls, a hash guard or an installed client. Mutable initial bytes are not live producer readiness.

Trace-GalaxyExposureControlsCompatibility.py binds SHA256 6891f70a021fa9f3a0e00458c284f48d1226e68ae7fc02ad48081aed47857e93 to current ED5 SHA256 ed5a8f8cd10f9a80d3bda5badd7ba1d4fcfc15996c8c7e018852cb170e264b63. Every listed complete unwind span has one full masked-body match, equal instruction count and no unexplained non-address byte changes. Actual RIP/call operands are retained in the JSON; no fixed relocation delta is assumed.

Original function	Current ED5 entry	Complete bytes	Role
351F8E0	351F880	1553	Owner constructor, EV0 and mode0 initial upper-half stores
351F450	351F3F0	1167	Shadow constructor, provisional mode later overwritten
3526320	35262C0	3444	Per-render owner100→shadow6FC EV, owner120→shadow6C8 mode
3523480	3523420	10723	Five exposure controls and actual pass selection
3525E70	3525E10	1124	Separate optional blur/downsample
3521E90	3521E30	1456	Actual embedded exposure program loader

The constructor's new351FA15 and351FA35 stores are byte-identical to old351FA75/351FA95. Each is64 bits: the zero upper DWORD establishes EV0 at100 and mode0 at120, independently of neighboring .5 atFC and1 at11C. The per-render copies remain exactly the same field offsets and integer-bit copies.

Actual paired RIP operands in the full publisher map each global to itself. Their .data initial bytes are identical:

Control	Both-image RVA	Raw little-endian bytes	Initial value
Toe	5E8B918	D7A3803F	1.0049999952316284
Shoulder	5E8B910	33331340	2.299999952316284
Offset	5E8B908	0000A040	5
Slope	5E8B914	0000803F	1
Enabled curve	5E8B997	01	true
Dispatch flag	5E8B994	01	true

The matching loader's actual old3521EC8/new3521E68 RIP operand binds descriptor54C6248→54C6218. Its actual data pointer, not a guessed neighboring-data delta, binds compressed program54C4568→54C4538; the platform remains Win64_SM50. Inflating7370 bytes in each image yields the same95472-byte program SHA256 62ef0e8e73e653df972031438fdf5a44e5c152e3c7aa0c9af8fd42e14df7cea7. Actual pass order remains0 AnalyseLuminance_AutoExposure,1 AnalyseLuminance_ManualExposure.

At new3525D6B the same copied mode is tested, then sete supplies R8D for new4F8D10. Thus mode0 still selects the named manual pass, consistent with the original source evidence. The source program's reflection and compute bytes are unchanged as part of the full program hash.

All bounded checks are closed in artifacts/galaxy-client-compatibility/exposure-controls-paired.json. This supports an explicitly labelled tool-owned ManualCS initial fallback using these proven static defaults. It does not establish current game options, a completed manual/automatic dispatch, runtime exposure history, or another camera's live histogram. No production defaults or source guards were changed by this audit.

# Bounded exposure-default compatibility:6891 to ED5 This is a read-only comparison of local PE files. It neither opens a game process nor changes Native, production controls, a hash guard or an installed client. Mutable initial bytes are not live producer readiness. `Trace-GalaxyExposureControlsCompatibility.py` binds SHA256 `6891f70a021fa9f3a0e00458c284f48d1226e68ae7fc02ad48081aed47857e93` to current ED5 SHA256 `ed5a8f8cd10f9a80d3bda5badd7ba1d4fcfc15996c8c7e018852cb170e264b63`. Every listed complete unwind span has one full masked-body match, equal instruction count and no unexplained non-address byte changes. Actual RIP/call operands are retained in the JSON; no fixed relocation delta is assumed. | Original function | Current ED5 entry | Complete bytes | Role | |---|---|---:|---| |351F8E0|351F880|1553|Owner constructor, EV0 and mode0 initial upper-half stores| |351F450|351F3F0|1167|Shadow constructor, provisional mode later overwritten| |3526320|35262C0|3444|Per-render owner100→shadow6FC EV, owner120→shadow6C8 mode| |3523480|3523420|10723|Five exposure controls and actual pass selection| |3525E70|3525E10|1124|Separate optional blur/downsample| |3521E90|3521E30|1456|Actual embedded exposure program loader| The constructor's new351FA15 and351FA35 stores are byte-identical to old351FA75/351FA95. Each is64 bits: the zero upper DWORD establishes EV0 at100 and mode0 at120, independently of neighboring .5 atFC and1 at11C. The per-render copies remain exactly the same field offsets and integer-bit copies. Actual paired RIP operands in the full publisher map each global to itself. Their `.data` initial bytes are identical: | Control | Both-image RVA | Raw little-endian bytes | Initial value | |---|---|---|---:| |Toe|5E8B918|D7A3803F|1.0049999952316284| |Shoulder|5E8B910|33331340|2.299999952316284| |Offset|5E8B908|0000A040|5| |Slope|5E8B914|0000803F|1| |Enabled curve|5E8B997|01|true| |Dispatch flag|5E8B994|01|true| The matching loader's actual old3521EC8/new3521E68 RIP operand binds descriptor54C6248→54C6218. Its **actual data pointer**, not a guessed neighboring-data delta, binds compressed program54C4568→54C4538; the platform remains `Win64_SM50`. Inflating7370 bytes in each image yields the same95472-byte program SHA256 `62ef0e8e73e653df972031438fdf5a44e5c152e3c7aa0c9af8fd42e14df7cea7`. Actual pass order remains0 `AnalyseLuminance_AutoExposure`,1 `AnalyseLuminance_ManualExposure`. At new3525D6B the same copied mode is tested, then `sete` supplies R8D for new4F8D10. Thus mode0 still selects the named manual pass, consistent with the original source evidence. The source program's reflection and compute bytes are unchanged as part of the full program hash. All bounded checks are closed in `artifacts/galaxy-client-compatibility/exposure-controls-paired.json`. This supports an explicitly labelled **tool-owned ManualCS initial fallback** using these proven static defaults. It does not establish current game options, a completed manual/automatic dispatch, runtime exposure history, or another camera's live histogram. No production defaults or source guards were changed by this audit.