Resolve levelshot, macOS provider, dialogue media, and loadscreen issues - #104
Merged
Conversation
This was referenced Aug 13, 2026
Distinguish base dialogue media from numbered language patches, include the process ID in generated-loadscreen staging paths, and keep future macOS support notes pinned to the live visual-parity tracker. Align the OpenAL build guide with the strict pkg-config lookup.\n\nReferences #73, #87, #94, #98, and #102.
Replace the new shared process-ID ABI with existing OS CSPRNG staging tokens and keep the source levelshot when secure staging is unavailable. This preserves engine/GameLib header parity while closing the remaining cross-process collision window.
Update the broad macOS bridge assertion for the strict pkg-config OpenAL policy and require the secure loadscreen staging path to fail closed before publication.
Prepare and verify the pinned MoltenVK payload before recording both ARM64 and Intel thin build trees. This makes the fresh-runner provenance gate match the required package layout.
Prepare and verify the pinned MoltenVK payload on every thin architecture before provenance recording and archival. Harden the candidate workflow contract to enforce unconditional install, preparation, verification, and recording order. References #95
Normalize and re-sign every installed macOS loadable module before thin provenance capture, and validate both slices when packaging universal2 payloads. Cover renderer-vk alongside the SP and MP modules in manual release checks. Prepare the Windows x64 OpenAL provider in commit validation so runtime staging matches push verification. References #95
Verify both recorded universal MoltenVK binaries and every signature slice, then compare signature-stripped temporary copies so host-dependent ad-hoc identifiers cannot break universal2 assembly or conceal different provider bytes. References #95
Merge intact signed thin slices before normalizing loadable-module IDs, then ad-hoc sign and verify every fused universal2 binary. This avoids leaving Mach-O LINKEDIT data unprocessable after signature stripping. References #95
themuffinator
force-pushed
the
codex/resolve-open-issues
branch
from
August 13, 2026 15:23
7355f41 to
ef2196d
Compare
Keep leading-hyphen universal2 variant suffixes attached to their option in both commit validation and the non-publishing release-candidate lane. Add scoped contracts for the ad-hoc and Developer ID paths. References #95
Install Python-Markdown in each universal2 package-producing job before release documentation is rendered. Guard the candidate and commit-validation ordering so isolated hosted Python environments cannot skip the dependency. References #95
Resolve the packaged app path against the GitHub workspace before changing to the Finder-style working directory. Add a scoped contract that prevents the relative-path launch regression. References #95
Record the macOS SDL launcher status without short-circuiting the Finder-style smoke, then validate normal exit shape, required runtime markers, resource resolution, module selection, and absence of fatal diagnostics. Preserve hard timeout and signal failures. References #95
Persist the packaged app console alongside its engine log, validate pre-log filesystem resolution and exact universal SP module selection from that console, and scan both outputs for fatal diagnostics. References #95
Owner
Author
|
Final hosted evidence for
Push Verification passed on Windows, Linux x64, and Linux ARM64. Its four macOS failures are the exact inherited hosted-renderer baseline already present on The workflow-dispatch candidate is not represented in the PR check rollup, so its exact-head evidence is recorded here explicitly. |
Keep ignored Python cache files out of cross-runner source provenance and all staging refresh inventories so identical clean sources assemble reproducibly. References #95
Apply the existing Linux cross-DSO vptr exclusion to in-process BSE objects whose renderer-interface RTTI is owned by runtime-loaded modules. ASan and all other requested UBSan checks remain enabled. References #95
Renderer modules must not delete idFile objects allocated by the engine. Route the MD5R export lifetime through idFileSystem::CloseFile and guard both renderer source sets without weakening sanitizer coverage. References #95
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This draft collects the engine/content changes produced by a full audit of the 14 open tracker reports.
Frameworks/libMoltenVK.dyliblayout used by macOS packages, then normalize and re-sign installed SP, MP, and Vulkan renderer modules before thin-build provenance is recorded.zpak_<language>.pk4dialogue-media base archive is mounted; numbered patch archives no longer mask a missing base archive, while dedicated servers and assetless CI remain unaffected.Issue disposition
#11: fully implemented; 24 missing sets added, bringing the source-derived inventory to 63/63.#94,#95,#102: root-caused and fixed here.#87: diagnostic improvement only; the issue remains open pending the reporter's current Linux log. Missing retail language media is the strongest hypothesis, with output routing still possible.#99: documentation is brought back in line with shipped code; implementation lives in the coordinatedopenQ4-gamedraft.The tracker audit separately closed
#73,#78,#82, and#96because current-main fixes and reporter evidence already resolved them. No whole-issue duplicates were found; partial overlaps were consolidated in issue comments.Validation
3107f8f0,03c20342,8cfa4ccc,19dd2192,b7d421cb,eeba86b2,0ab0e97b,ef2196d1,6db32cc0,be2e95de,6f743c72,5153ba45, ande84b89acfixed them, restored the engine/GameLib header-parity gate, aligned the OpenAL guide with the strict pkg-config implementation, prepared the pinned MoltenVK payload before every recorded thin build, normalized Meson-rewritten loadable-module IDs, validated both universal2 slices, aligned Windows commit validation with its OpenAL runtime staging contract, and made universal2 assembly compare verified signature-stripped MoltenVK bytes so host-dependent ad-hoc identifiers cannot conceal or mimic provider drift.9196401182has SHA-256203b8644eedb9101f9c700ad8a84f4fd4b74571e1cd5217ca39477bf7726a6b1.02cbe9b4andbe7ce029preserve ASan and every requested UBSan check except the existing cross-modulevptrboundary, keep BSE linkable, and close engine-owned MD5R export files through the engine instead of deleting them inside renderer modules.Closes #11
Fixes #94
Fixes #95
Fixes #102
References #87
References #99