Engine Full Parity Audit And Closure Implementation Plan

April 21, 2026 · View on GitHub

Last updated: 2026-04-21

Scope: engine-owned executable surfaces in src/code/qcommon/*, src/code/client/*, src/code/server/*, src/code/renderer/*, and the remaining executable-owned host/support surface (src/code/win32/*, src/common/platform/*, src/code/botlib/*) versus retail quakelive_steam.exe. The game-module layer (ui, cgame, qagame) is tracked separately as a dependency because the end-state replacement target requires the engine host and the retail/native module contract to agree at the same time.

Purpose: publish one authoritative current-worktree engine-wide parity audit that aggregates the dedicated subsystem audits, the machine-readable parity gates, the runtime evidence bundles, and the retail module-host proof into a single closure ledger.

Audit Method And Evidence

Owning retail executable:

assets/quakelive/quakelive_steam.exe

Canonical committed evidence used for this engine-wide audit:

Binary Ninja HLIL corpus:
- references/hlil/quakelive/quakelive_steam.exe/quakelive_steam.exe_hlil.txt
- references/hlil/quakelive/quakelive_steam.exe/quakelive_steam.exe_hlil_split/*
Ghidra companion corpus:
- references/reverse-engineering/ghidra/quakelive_steam/metadata.txt
- references/reverse-engineering/ghidra/quakelive_steam/imports.txt
- references/reverse-engineering/ghidra/quakelive_steam/exports.txt
- references/reverse-engineering/ghidra/quakelive_steam/functions.csv
- references/reverse-engineering/ghidra/quakelive_steam/analysis_symbols.txt
- references/reverse-engineering/ghidra/quakelive_steam/decompile_top_functions.c
Symbol and mapping support:
- references/analysis/quakelive_symbol_aliases.json
- docs/reverse-engineering/quakelive_steam_mapping_round_*.md
Dedicated subsystem audits:
- docs/reverse-engineering/qcommon-full-parity-audit-and-implementation-plan-2026-04-10.md
- docs/reverse-engineering/client-full-parity-audit-and-implementation-plan-2026-04-09.md
- docs/reverse-engineering/server-full-parity-audit-and-implementation-plan-2026-04-10.md
- docs/reverse-engineering/renderer-full-parity-audit-and-implementation-plan-2026-04-09.md
- docs/reverse-engineering/engine-host-support-full-parity-audit-and-implementation-plan-2026-04-10.md
Machine-readable engine gate artifacts:
- artifacts/qcommon_validation/logs/qcommon_full_parity_gate.json
- artifacts/client_validation/logs/client_full_parity_gate.json
- artifacts/server_validation/logs/server_full_parity_gate.json
- artifacts/renderer_validation/logs/renderer_full_parity_gate.json
- artifacts/engine_host_support_validation/logs/engine_host_support_full_parity_gate.json
Engine runtime/evidence artifacts:
- artifacts/qcommon_validation/logs/qcommon_runtime_evidence_20260410.json
- artifacts/client_validation/logs/client_runtime_evidence_20260410.json
- artifacts/server_validation/logs/server_runtime_evidence_20260410.json
- artifacts/renderer_validation/logs/renderer_runtime_evidence_latest.json
- artifacts/engine_host_support_validation/logs/engine_host_support_runtime_evidence_20260410.json
Dependent module evidence used to bound the replacement target:
- docs/reverse-engineering/game-module-parity-audit-and-implementation-plan-2026-04-10.md
- artifacts/module_validation/logs/retail_module_parity_gate.json
- artifacts/ui_validation/logs/ui_full_parity_gate.json
- artifacts/module_validation/logs/retail_module_runtime_evidence_latest.json

Method:

Treat the dedicated 2026-04-09/10 subsystem audits as the primary ownership ledgers for their respective surfaces.
Reconfirm that each engine register has both a narrative closure document and a machine-readable gate reporting overall_status: pass.
Reconfirm that each register also has either a dedicated runtime bundle or a focused evidence bundle that is already accepted by the owning audit.
Aggregate engine gap counts directly from the gate artifacts instead of inventing a new synthetic scoring model.
Keep compatibility-only host/support surfaces explicit rather than quietly blending them into the strict-retail score.
Cross-check the retail module-host seam separately so the engine-wide replacement claim does not overstep the direct retail-DLL evidence.

Committed Retail Corpus Snapshot

Retail quakelive_steam.exe metadata currently reports:

function corpus: 5473
imports: 351
exports: 2
promoted analysis symbols: 4377

Cross-check:

references/reverse-engineering/ghidra/quakelive_steam/functions.csv contains 5473 function rows, matching the metadata snapshot.

Interpretation:

the current engine-wide closure claim still rests on a large committed retail corpus, not on a narrow hint set
the closure story is now primarily about verified owner recovery and proof discipline, not missing binary discovery

Current Engine-Wide Conclusion

Current strict engine-owned parity estimate: 100%.
All five engine-owned machine-readable gate artifacts report overall_status: pass.
The aggregated engine gap surface is 31 / 31 passing gap IDs.
The strict-retail-counted engine gap surface is 29 / 29 passing gap IDs once the remaining host/support compatibility-only exclusions (EH-G03, EH-G05) are removed per the boundary contract.
No open engine-owned parity gap remains in the audited register for the current worktree.
The module dependency layer is also closed separately at 100%, with retail_module_parity_gate.json and ui_full_parity_gate.json both passing and the stable retail-DLL runtime alias still providing direct retail module-host evidence.

Important scope note:

this document does not claim a fresh runtime uplift beyond the dedicated subsystem closures; it consolidates the already-closed engine registers into one authoritative ledger

Engine Register Scorecard

Register	Audit document	Current estimate	Gap status	Machine-readable gate	Runtime or evidence bundle	Current conclusion
`qcommon`	`docs/reverse-engineering/qcommon-full-parity-audit-and-implementation-plan-2026-04-10.md`	`100%`	`5 / 5` passing	`artifacts/qcommon_validation/logs/qcommon_full_parity_gate.json`	`artifacts/qcommon_validation/logs/qcommon_runtime_evidence_20260410.json`	Bootstrap, filesystem, message transport, collision leaf ownership, and VM host/fallback behavior are closed.
`client`	`docs/reverse-engineering/client-full-parity-audit-and-implementation-plan-2026-04-09.md`	`100%`	`5 / 5` passing	`artifacts/client_validation/logs/client_full_parity_gate.json`	`artifacts/client_validation/logs/client_runtime_evidence_20260410.json`	Browser-host runtime, Steam callback lifetime, workshop bootstrap, config persistence, and final client proof are closed.
`server`	`docs/reverse-engineering/server-full-parity-audit-and-implementation-plan-2026-04-10.md`	`100%`	`6 / 6` passing	`artifacts/server_validation/logs/server_full_parity_gate.json`	`artifacts/server_validation/logs/server_runtime_evidence_20260410.json`	Steam GameServer lifecycle, `idZMQ`, stat/achievement ownership, rankings compatibility, control-plane cvars, and dedicated proof are closed.
`renderer`	`docs/reverse-engineering/renderer-full-parity-audit-and-implementation-plan-2026-04-09.md`	`100%`	`9 / 9` passing	`artifacts/renderer_validation/logs/renderer_full_parity_gate.json`	`artifacts/renderer_validation/logs/renderer_runtime_evidence_latest.json`	Export ABI, memory-image ingestion, post-process, Win32 glue, classic font lane, host text core, build lane, and strict validation are closed.
Remaining engine host/support	`docs/reverse-engineering/engine-host-support-full-parity-audit-and-implementation-plan-2026-04-10.md`	`100%`	`6 / 6` passing (`4 / 4` strict-retail-counted)	`artifacts/engine_host_support_validation/logs/engine_host_support_full_parity_gate.json`	`artifacts/engine_host_support_validation/logs/engine_host_support_runtime_evidence_20260410.json`	Win32 clipboard/raw-input, loading-window glue, botlib internal proof, and boundary formalisation are closed; compatibility-only lanes are explicit exclusions.

Historical Uplift By Register

No synthetic historical whole-engine percentage is published here because the five audited registers closed on different dates and with different scoped boundaries. The honest history is the per-register uplift shown below.

Register	Historical strict estimate at audit publication	Current strict estimate	Final closure phase
`qcommon`	`87%`	`100%`	`QC-P6`
`client`	`90%`	`100%`	`CL-P6`
`server`	`74%`	`100%`	`SV-P7`
`renderer`	`94%`	`100%`	`RG-P11`
Remaining engine host/support	`79%`	`100%`	`EH-P5` plus `EH-P1` and `EH-P6` governance closure

Interpretation:

the engine-wide closure story is real progress across every engine-owned register, not a score-only reframing of one subsystem
the lowest historical register was the server at 74%; the current register set is now uniformly closed

Detailed Findings

1. `qcommon`

Observed facts:

The dedicated qcommon audit closes QC-G01 through QC-G05 and tracks the current strict qcommon estimate at 100%.
The ownership story is backed by the committed alias ledger plus mapping rounds 34, 55, 56, 63, 78, 79, 84, 91, 99, 107, 108, and 109.
artifacts/qcommon_validation/logs/qcommon_runtime_evidence_20260410.json proves:
- qzconfig.cfg and repconfig.cfg execution
- the active search-path roots
- service-disabled launcher/resource markers
- writable-homepath DLL loading for ui, qagame, and cgame
artifacts/qcommon_validation/logs/qcommon_full_parity_gate.json records 5 / 5 passing tranches with no non-passing gap IDs.

Conclusion:

no open qcommon gap remains in the audited engine register

2. `client`

Observed facts:

The dedicated client audit closes CL-G01 through CL-G05 and tracks the current strict client estimate at 100%.
The writable client host now retains:
- browser-host runtime ownership
- qz_instance JS bridge and EnginePublish event-publication seams
- Steam client/lobby/micro callback bundles and frame pumping
- workshop-aware join/bootstrap ownership
- retail qzconfig.cfg / repconfig.cfg bootstrap plus writeClientConfig
artifacts/client_validation/logs/client_runtime_evidence_20260410.json proves:
- windowed main-menu UI bootstrap
- live qzconfig.cfg, repconfig.cfg, and writeClientConfig output
- explicit service-disabled browser-policy behavior
- a live bloodrun local-map pass with screenshots and a flushed demo
- clean lifecycle-end markers
artifacts/client_validation/logs/client_full_parity_gate.json records 5 / 5 passing tranches with no non-passing gap IDs.

Conclusion:

no open client gap remains in the audited engine register
live online services remain correctly policy-gated behind QL_BUILD_ONLINE_SERVICES; that is now an explicit repo policy surface, not silent missing client ownership

3. `server`

Observed facts:

The dedicated server audit closes SV-G01 through SV-G06 and tracks the current strict server estimate at 100%.
The writable server host now retains:
- Steam GameServer bootstrap/logon/shutdown ownership
- server-side callback/auth-session lifetime
- retained idZMQ stats-publication and remote-RCON runtime
- qagame-facing Steam stat and achievement ownership
- reconstructed control-plane cvar and policy surface
- a documented default-disabled rankings compatibility contract
artifacts/server_validation/logs/server_runtime_evidence_20260410.json proves:
- dedicated startup
- qagame load
- network-visible getstatus publication
- local rcon status and clean quit
- optional Steam and idZMQ markers recorded honestly as environment-driven booleans
artifacts/server_validation/logs/server_full_parity_gate.json records 6 / 6 passing tranches with no non-passing gap IDs.

Conclusion:

no open server gap remains in the audited engine register

4. `renderer`

Observed facts:

The dedicated renderer audit closes RG-G01 through RG-G09 and tracks the current strict renderer estimate at 100%.
The writable renderer now retains:
- the retail GetRefAPI export tail
- in-memory image/resource ingestion
- shader-backed post-process and color-correction passes
- Win32 resize/loading-window glue
- the classic font/cache/atlas lane
- the retained *fontstash host text engine
- native ui and cgame host text import switchover
- an explicit external FreeType build lane
artifacts/renderer_validation/logs/renderer_runtime_evidence_latest.json proves:
- windowed main-menu bootstrap
- retained-atlas debug rendering through r_debugFontAtlas
- live bloodrun runtime
- expected renderer init markers including InitFreeType and InitFontStash
- no fallback to compatibility-only RE_RegisterFont runtime paths
artifacts/renderer_validation/logs/renderer_full_parity_gate.json records 9 / 9 passing tranches with no non-passing gap IDs.

Conclusion:

no open renderer gap remains in the audited engine register

5. Remaining Engine Host/Support

Observed facts:

The dedicated remaining host/support audit closes EH-G01 through EH-G06 and tracks the current strict remaining-engine host/support estimate at 100%.
The writable host now retains:
- Unicode-first clipboard fetch with ANSI fallback
- active Win32 raw-input registration and WM_INPUT dispatch
- retained loading-window and bootstrap wrapper ownership
- representative botlib internal proof for AAS/reachability/goal-state helpers
artifacts/engine_host_support_validation/logs/engine_host_support_runtime_evidence_20260410.json binds those closures to focused tests and source owners instead of relying on broad narrative confidence alone.
artifacts/engine_host_support_validation/logs/engine_host_support_full_parity_gate.json records:
- 6 / 6 passing tranches overall
- explicit scope boundary metadata
- EH-G03 and EH-G05 as compatibility-only exclusions
- EH-G01, EH-G02, EH-G04, and EH-G06 as the strict-retail-counted host/support set

Conclusion:

no open strict-retail host/support gap remains outside qcommon, client, server, and renderer
the compatibility-only platform-service abstractions and the Unix/null trees remain visible, but they are now documented exclusions rather than open engine parity debt

6. Module Dependency Snapshot

Observed facts:

docs/reverse-engineering/game-module-parity-audit-and-implementation-plan-2026-04-10.md tracks the current strict retail module estimate at 100%.
artifacts/module_validation/logs/retail_module_parity_gate.json records 3 / 3 passing module tranches with overall_status: pass.
artifacts/ui_validation/logs/ui_full_parity_gate.json records 6 / 6 passing UI tranches with overall_status: pass.
artifacts/module_validation/logs/retail_module_runtime_evidence_latest.json still proves direct retail-DLL host loading for retail uix86.dll, qagamex86.dll, and cgamex86.dll under the reconstructed host. The stable alias now points at the current bounded 2026-04-21 artifact, whose remaining live-map shortfall is explicitly renderer-owned rather than a module-host ambiguity.
The only bounded live-map shortfall archived in that retail module runtime probe was explicitly renderer-owned, and the renderer register is now closed separately in the current worktree.

Conclusion:

the module dependency layer is not an active blocker for the current engine-wide replacement target
the authoritative current state is that both the engine-owned registers and the module dependency registers are closed, even though the stable runtime alias currently remains pinned to the bounded 2026-04-09 artifact

Open Gap Register

Open engine-owned strict-retail gaps:

none

Explicit non-gap exclusions that remain visible:

EH-G03: compatibility-only platform-service abstractions
EH-G05: Unix/null compatibility ports
live online-service activation remains policy-gated behind QL_BUILD_ONLINE_SERVICES

Active remaining repo work outside this engine register:

IMPLEMENTATION_PLAN.md Task A3: resolve or permanently bound the compatibility-only online-service lanes
IMPLEMENTATION_PLAN.md Task A4: modernise or explicitly contain the non-Windows portability lanes
IMPLEMENTATION_PLAN.md Task A6: finish refreshing the remaining build/runtime evidence and publication follow-ups

Implementation Plan

No open engine-specific closure phase remains. The correct execution plan after this audit is maintenance and evidence discipline, not another broad engine reconstruction tranche.

Standing execution order:

Keep the five engine gates authoritative. Whenever engine-owned code changes one of the audited contracts, refresh the owning gate artifact first.
Refresh runtime evidence only when the active host contract changes, and use the dedicated subsystem probes instead of ad-hoc broad reruns:
- qcommon: tools/qcommon/run_qcommon_runtime_probe.ps1
- client: tools/client/run_client_runtime_probe.ps1
- server: tools/server/run_server_runtime_probe.ps1
- renderer: tools/renderer/run_renderer_runtime_probe.ps1
- remaining host/support: focused harness surface plus artifacts/engine_host_support_validation/logs/engine_host_support_runtime_evidence_20260410.json
Reopen only the owning subsystem register when a gate fails or new retail evidence contradicts the current source owner; do not reopen the whole engine ledger speculatively.
Keep compatibility-only lanes explicitly out of the strict-retail score unless repo policy changes:
- src/common/platform/platform_services.c
- src/common/platform/backends/platform_backend_open_steam.c
- src/common/platform/backends/platform_backend_steamworks.c
- src/code/unix/*
- src/code/null/*
- live online services behind QL_BUILD_ONLINE_SERVICES
Keep the top-level active queue focused on non-engine remaining work: ownerdraw/stat payload completion, targeted gameplay validation, and normal gate/runtime refresh whenever audited contracts change.

Bottom Line

The reconstructed engine executable no longer has an open strict-retail parity gap in the current audited register. qcommon, client, server, renderer, and the remaining engine host/support scope are all closed with machine-readable gates and runtime or focused-evidence bundles, and the module dependency layer is also closed separately.

The practical next step is no longer broad engine reconstruction. It is disciplined maintenance: keep the subsystem gates green, refresh the dedicated probes when audited contracts change, and spend active development effort on the remaining gameplay/ownerdraw validation work outside this engine register.