Repository navigation
Description
Summary
Master (as of 6c5fd4516) fails every app build lane: Build π§± / Ubuntu π§ (OBS-Studio-Rust),
Build π§± / Flatpak π¦, Build π§± / macOS π (arm64) and (x86_64). All of them stop in the
frontend C++ compile with -Werror=conversion / -Werror,-Wsign-conversion, in files that
#2909 did not convert.
Evidence
workflow_dispatch of scheduled.yaml on fix/2921-mp4-mux-conversions (that branch is rebased
onto 6c5fd4516; its own two files compile clean before this point), run
40902391979:
Build π§± / Ubuntu π§ (OBS-Studio-Rust) job 123553698526 failure
Build π§± / Flatpak π¦ failure
Build π§± / macOS π (arm64) job 123553699030 failure
Build π§± / macOS π (x86_64) failure
Ubuntu:
frontend/utility/ExtraBrowsersModel.cpp.hpp:20:34: error: conversion from 'qsizetype' {aka 'long long int'} to 'int' may change value [-Werror=conversion]
frontend/utility/ExtraBrowsersModel.cpp.hpp:33:31: error: ... same
frontend/utility/ExtraBrowsersModel.cpp.hpp:108:32: error: ... same
frontend/utility/ExtraBrowsersModel.cpp.hpp:190:37: error: ... same
macOS arm64:
[obs-studio] Compiling qt-wrappers.core+.hpp.cpp
##[error]implicit conversion changes signedness: 'int' to 'uint32_t' (aka 'unsigned int') [-Werror,-Wsign-conversion]
43 | return obsModifiers;
Root cause
18f8a3f22 (#2909, merged 2026-10-29 14:10 UTC) added
-Wconversion;-Wsign-conversion;-Wfloat-conversion to the obs-studio frontend target in
frontend/CMakeLists.txt:203 together with fixes for only the first group of files. Its own commit
message says so:
Partial: the 48 diagnostics in the
frontend.*unity units ... The target-wide ratchet cannot
build until the other 746 diagnostics in 63 files are fixed too, so this commit is not pushed on
its own.
The second half of that sentence did not hold: the flags landed with the partial fix.
PR CI never compiles the frontend (its Ubuntu lane builds no app), so the merge queue saw a green
PR; the breakage only shows in the master-push/dispatch app lanes, which is why master went red
without a PR going red.
Options
- Restore the pins for this target until the rest is converted β re-add
-Wno-conversion;-Wno-sign-conversion;-Wno-float-conversiontoobs-studioin
frontend/CMakeLists.txt, with aci/test_ci_*.pystyle pin that fails when the flags return, and
land the 63 files as their own PRs. Restores every release lane immediately; matches what frontend: state the conversions the app target now reports (#2625)Β #2909's
message said the ordering should be. - Finish the remaining frontend conversions in one PR (~746 diagnostics / 63 files, size L) and keep
the lane dark meanwhile, which keeps master's release artifacts unbuildable and blocks every open
PR that needs a macOS or Ubuntu app build for evidence (including build(obs-outputs): state the mp4-mux and rtmp-hevc conversions macOS clang reports (#2921)Β #2922 and build(obs-outputs): state the rtmp-hevc record conversions (#2928)Β #2929).
Recommendation: option 1 now, then re-land per unity-unit group.
Not the cause
Not #2925/#2926/#2922: those touch plugins/obs-outputs/** only, and the failing files here are
frontend/shared C++ that no PR in flight modifies.
Metadata
Metadata
Assignees
Labels
Type
Projects
- StatusShow more project fieldsDone
Activity
zackees commentedon Oct 29, 2026
Claiming: #2938
Plan: turn the frontend conversion group green in place, one unity group per commit, so
Build π§± / Ubuntu π§ (OBS-Studio-Rust), Flatpak, and both macOS app lanes compile again.-Wconversion/-Wsign-conversionstay on for the target; the fix is stating the conversions at the call sites, not relaxing the flag.obs-studiofrontend target with the same flags, and record the first RED slice (the unity file that fails).frontend/utility/ExtraBrowsersModel.cpp.hpp,frontend/qt-wrappers.cpp.hpp) with explicit casts, byte/behaviour-preserving; verify with a local build of the same target.Fixes #2938from a clean master and let PR CI decide.This is a P0 blocking every app lane, so it goes first; PR follows as soon as the first group is verified.
zackees commentedon Oct 29, 2026
Status on #2938 β the cause is narrower than the issue says, and the fix splits in two.
1. Ubuntu + Flatpak came from a duplicate opt-in block, not from an unfinished conversion.
frontend/CMakeLists.txtalready optedobs-studiointo-Wsign-conversionalone("
-Wconversionand-Wfloat-conversionkeep the global pins for now"). #2909 appended asecond
target_compile_options(obs-studio ...)block at the end of the file naming all threegroups, and its own commit message said 746 diagnostics in 63 files were still unconverted. The
app configure compiles with
CMAKE_COMPILE_WARNING_AS_ERROR=ON, so the lanes stopped on the firstunconverted file. Removing that second block is the P0 fix and lands as #2944, with
ci/test_frontend_conversion_optins.pyholding the target to one opt-in block that names only thegroup it has converted.
That is option 1 in the issue, scoped to the two groups #2909 added by mistake:
-Wsign-conversionstays on for the target, the other two stay on the global pins until their files are clean.
2. macOS is a different, older red. The 05:49 UTC
scheduled.yamlon master (before #2909)already had
Build π§± / macOS π (arm64)and(x86_64)failing while Ubuntu and Flatpak weregreen; the
qt-wrappers.host-macos+.hpp.cppreturn obsModifierssite is the sign-conversionerror that run reports, and it is what #2940 states at the call site.
Why PR CI was green on #2909.
Build π§± / Qt Frontend Linux π§is selected forfrontend/changes and passed, but the app configure compiles a larger unity set than that lane, so
frontend/utility/ExtraBrowsersModel.cpp.hppwas not in any unit it built. The lanes that break(
Ubuntu π§ (OBS-Studio-Rust),Flatpak π¦, macOS) never run on a PR β ascheduled.yamldispatch on the branch is still the only evidence for them.
Next: #2944 for the pins (dispatch running on that branch for the Ubuntu/Flatpak lane evidence),
#2940 for the macOS sign-conversion sites, then the remaining frontend groups convert file by file
and the flag rejoins the single opt-in block in the PR that converts the last one. The claim stays.
zackees commentedon Oct 29, 2026
Progress from the sanitizer dispatch on master (
c2df54a7f, run 40919949570):asan π§ͺis red for a frontend site, not for the C conversions. Withmp4-mux.c.handrtmp-hevc.c.hfixed on master, the lane's asan build got past them and stopped at:This is the
1 << mixerIdx/mixers & (1 << n)bath inOBSAdvAudioCtrl. #2952 fixes that whole file's site class (1uliterals, no value change) and leaves the rest of the group to follow-ups.valgrind π§ͺandtsan π§ͺare still building in that run; the valgrind lane is the one build: state the other get_bit1 shift and link libm where the sanitizer needs it (#2923)Β #2935'slibmcommit unblocked, so its link result is the acceptance check for the conversion work.Release OBS-Studio-Rustrun (40906899118) is red for the same group:frontend/utility/ExtraBrowsersModel.cpp.hpp:20/33/108/190(qsizetype->int) on Linux, and macOS/Flatpak behind it. build(frontend): keep the unconverted app target on the global conversion pins (#2938)Β #2944 (restore the pins, queued) and build: state the frontend conversion sites the app lanes stop on (#2938)Β #2940 (the ExtraBrowsersModel +qt-host-*casts, queued) are aimed at exactly those.Nothing here needs a decision from you; the ratchet question in #2923's last comment does.
zackees commentedon Oct 29, 2026
Progress (worker tick): the app-lane conversion fixes are landing β #2940 (conversion sites stated) and #2944 (global pins kept for the unconverted target) merged at 18:28 UTC, #2947 (qt-host macOS modifier return) is in the merge queue. The macOS lanes then reached the next files, which are tracked separately as #2950 (PR #2951:
platform-osx.mm.hpp,properties-view.cpp.hpp,PluginModuleLoader_MacOS.mm.hpp). This issue stays open for the remaining unconverted app targets.zackees commentedon Oct 29, 2026
Status from the worker tick: the app-lane acceptance evidence is now being produced, on a tree that contains the fixes.
Every app-lane run on master so far predates the fixes this issue tracks: the
Release OBS-Studio-Rustfailure at 15:58 UTC ran3bff1d7c8(15:33), i.e. before #2940 (17:17) and #2944 (18:28) landed, and the last full dispatch (40919949570) ranc2df54a7f. So its redRelease gate π¦ / Ubuntu+/ Flatpakjobs β both stopping onfrontend/utility/ExtraBrowsersModel.cpp.hpp:20,33,108,190(qsizetypeβint) β are the same symptom this issue was filed for, not a new class.Dispatched
scheduled.yamlon master2fe262aa3: 40943427805. ItsBuild π§± / Ubuntu π§ (OBS-Studio-Rust),Build π§± / Flatpak π¦and bothmacOS πbuild jobs are this issue's acceptance items; I will report the verdict and, if they are green, close this with the remaining-files list as the follow-up.In flight elsewhere, all linked here:
6ca10e79f.Decision needed:(the bot applieddecision-needed+claim-blocked): the macOS conversion burndown (build: state the macOS conversion sites the mac lanes stop on (#2950)Β #2951, four commits, every in-tree site cleared) has hit six-Wsign-conversiondiagnostics inside the Qt 6.11 deps' own headers, which is a build-system choice (-isystemfor.deps/ repair that unity group / stop with build(frontend): keep the unconverted app target on the global conversion pins (#2938)Β #2944's pins) and not a call-site cast.zackees commentedon Oct 29, 2026
First verdict from the master dispatch (40943427805,
2fe262aa3):Build π§± / Ubuntu π§ (OBS-Studio-Rust)β success. This is the lane whoseExtraBrowsersModel.cpp.hpp:20,33,108,190failures opened this issue, and the tree from build: state the frontend conversion sites the app lanes stop on (#2938)Β #2940/build(frontend): keep the unconverted app target on the global conversion pins (#2938)Β #2944 builds it clean. That half of the report is closed.Build π§± / macOS π (arm64)and(x86_64)β still failure, as isAnalyze π¬ / macOS π (clang-analyze). Their logs are not published until the whole run completes (Flatpak and the Windows builds are still going), so I am not going to name the failing file from the job list alone. build: state the macOS conversion sites the mac lanes stop on (#2950)Β #2951's rounds say the in-tree macOS sites are all stated; the class it is expected to be is the Qt 6.11 deps headers under-Wsign-conversion, which is theDecision needed:now open on build: master's macOS lanes are red in frontend/utility/platform-osx.mm.hpp and shared/properties-view/properties-view.cpp.hpp (#2625, #2872)Β #2950. I will confirm from the log rather than assume, and report either way.CEF Chromium Supportfailed, by design (browser: Upgrade the bundled CEF from 7871 (Chromium 150, end of life) to a supported ChromiumΒ #1864) β not a health signal.zackees commentedon Oct 29, 2026
Verdict from the completed master dispatch (40943427805,
2fe262aa3).Green (the app lanes this issue opened):
Build π§± / Ubuntu π§ (OBS-Studio-Rust)β success.Build π§± / Windows πͺ (x64)and(arm64)β success.Build π§± / Qt Frontend Linux π§,Build π§± / Flatpak π¦β success.Build π§± / Fast Gate β‘, the Rust Tier 2 Linux matrix,Rust Tests π¦andRust Cross Lint π¦(Windows, macOS) β success.So with #2940/#2944 on master the frontend conversion group's Linux and Windows app builds are clean. That half of the report is closed.
Still red:
Analyze π¬ / macOS π (clang-analyze),Build π§± / macOS π (x86_64),Build π§± / macOS π (arm64). Logs are published now, so here is what they actually say rather than the assumed class. The arm64 and x86_64 builds stop on eight-Wsign-conversiondiagnostics in this tree, not in the Qt deps headers:frontend/utility/update-helpers.cpp.hppstr.resize(size),vsnprintf(&str[0], size, ...))frontend/utility/WhatsNewInfoThread.cpp.hppfileStream.write/read,tellg,gcount,VerifySignature)Those are exactly the sites PR #2951's fourth round states (
e54bf8092), which is not on master yet. Its own macOS dispatch (40936503160) got past them and then stopped on six-Wsign-conversionerrors whose locations are inside the.deps/obs-deps-qt6-2026-10-25-universalheaders, theDecision needed:now open on #2950. So the macOS half of this issue is two layers deep and the second layer is a maintainer decision, not a source fix.Correction to this issue's own report: the 63 frontend sites it listed were real and are now stated; nothing in this run reproduces them on Linux or Windows.
CEF Chromium Support πalso failed in this run, by design (#1864), so it is not a health signal.zackees commentedon Oct 29, 2026
Verdict from the master dispatch, now that the run finished (40943427805,
2fe262aa3):Build π§± / Ubuntu π§ (OBS-Studio-Rust)β success. The lane whoseExtraBrowsersModel.cpp.hpp:20,33,108,190failures opened this issue builds clean, as build: state the frontend conversion sites the app lanes stop on (#2938)Β #2940/build(frontend): keep the unconverted app target on the global conversion pins (#2938)Β #2944 intended. The Linux half of this issue is closed.Build π§± / macOS π (x86_64)β failure, and every diagnostic is a site PR build: state the macOS conversion sites the mac lanes stop on (#2950)Β #2951 already states:frontend/whatsnew/WhatsNewInfoThread.cpp33fileStream.write(data.data(), data.size())(size_type->streamsize), 53size_t size = fileStream.tellg()(streamoff->size_t), 57fileStream.read(&data[0], size)(size_t->streamsize), 86size_t read = file.gcount()(streamsize->size_t), 176VerifySignature(..., data.size(), ...)(qsizetype->size_t);frontend/utility/update-helpers.cpp15str.resize(size)(int->size_type), 16vsnprintf(&str[0], size, ...)(int->size_t).Analyze π¬ / macOS π (clang-analyze)β failure atfrontend/utility/PluginModuleLoader_MacOS.mm128, 160, 172 (resize(bufferSize),int->size_type), also in build: state the macOS conversion sites the mac lanes stop on (#2950)Β #2951.Build π§± / macOS π (arm64)β failure; same job shape as x86_64 (logs are identical up to the first failing TU).So there is no new site here: master's macOS red is exactly the call-site list in #2951, which is a draft because its last dispatch round cleared every site in this tree and then stopped in the Qt 6.11 deps' own headers (
qnamespace.h:1926,1930,1935,qversiontagging.h:80). That is the build-policy question now recorded asDecision needed:on #2950, which carriesclaim-blockeduntil a maintainer answers.State of this issue: Linux builds green; macOS waits on the #2950 decision, not on more conversion work here. Nothing needs a decision from you on this issue itself β I will post the next verdict after #2951's blocker is resolved and it is merged.
zackees commentedon Oct 29, 2026
Status (worker tick): the 19:48 UTC dispatch's result is in, and it is one commit stale.
Run 40943427805 ran master
2fe262aa3(19:48 UTC), i.e. before #2951 merged as217b3cf08(22:36 UTC):Build π§± / Ubuntu π§ (OBS-Studio-Rust)β successBuild π§± / Flatpak π¦β successBuild π§± / macOS π (arm64)and(x86_64)β failure, but every diagnostic it reports is a site build: state the macOS conversion sites the mac lanes stop on (#2950)Β #2951 already states:frontend/utility/WhatsNewInfoThread.cpp.hpp33/53/57/86 (size_typeβstreamsize,streamoffβsize_t,gcount())frontend/utility/update-helpers.cpp.hpp15/16 (int sizeβstd::string::resize/vsnprintf)frontend/utility/WhatsNewInfoThread.cpp.hpp176 (QByteArray::size()βsize_t)On
217b3cf08all of those arestatic_casts in place (git show 217b3cf08 -- frontend/utility/WhatsNewInfoThread.cpp.hpp frontend/utility/update-helpers.cpp.hpp). So this run measures the pre-#2951 tree, not a new class.Dispatched
scheduled.yamlon master217b3cf08: 40970380763. The acceptance for this issue is that run'sUbuntu π§ (OBS-Studio-Rust)andFlatpak π¦jobs (green above) plus the two macOS build jobs; I will report the verdict and close with the remaining-files list if macOS is green, otherwise split the next file class into its own ticket.5 remaining items
zackees commentedon Oct 30, 2026
Status (worker tick,
obs-studio3): ascheduled.yamldispatch on current master43af74187(the #2968 merge) β run 40978686832 β narrows this issue to one lane class that is not our source any more.Build π§± / Ubuntu π§ (OBS-Studio-Rust)Build π§± / Flatpak π¦Build π§± / Windows πͺ (x64)plugins/obs-qsv11/obs-qsv11.c.h(1072,25)C4310/C2220β the #2965 residual, fixed by #2969, which is queue position 1;(arm64)was cancelled behind itBuild π§± / macOS π (arm64).deps/obs-deps-qt6-2026-10-25-universal/lib/QtCore.framework/Headers/qnamespace.h:1926,1930,1935andqversiontagging.h:80, all-Werror,-Wsign-conversionBuild π§± / macOS π (x86_64)CEF Chromium Support πSo the 63 unconverted frontend files and every first-party site in this report are now past the compiler: Linux and Flatpak app builds are green on master again, and Windows follows as soon as #2969 lands.
What is left is the escalated build-policy question, not a conversion. Both macOS lanes stop only in the universal Qt 6.11 deps' own headers (
qnamespace.h,qversiontagging.h) parsed into thefrontend.signal-posixunity group. That is exactly theDecision needed:on #2950 (option 1: mark the.depsinclude directoriesSYSTEMfor the app target; option 3: stop the macOS burndown and keep the pins), and it has no answer yet, so this lane stays red while those macOS app builds are unbuildable.Nothing else from this issue's list is red on the lanes above. I did not change a flag and did not open a policy PR: the choice between
-isystemfor deps includes and keeping the pins is the maintainer's, and it is recorded on #2950 with a one-wordDecided:as the unblock.Build π§± / Sanitize π§ͺwasskippingon this dispatch (the scheduled default), so it carries no evidence either way.zackees commentedon Oct 30, 2026
Lane evidence from a
scheduled.yamldispatch on master43af7418(run 40978686832):Build π§± / Ubuntu π§ (OBS-Studio-Rust)Build π§± / Flatpak π¦Build π§± / Windows πͺ (x64)plugins/obs-qsv11/obs-qsv11.c.h(1072,25): error C2220(C4310)Build π§± / macOS π (arm64)and(x86_64)qkeycombination.h/qnamespace.h:1926,1930,1935)Analyze π¬ / macOS π (clang-analyze)So the frontend conversion scope this issue was opened for is done: Ubuntu and Flatpak are green with #2968's two
scripts.cpp.hppsites (174, 204) and nothing else in the frontend group reports.What is still red is not this issue's scope:
obs-qsv11NAL-priority mask that build(obs-qsv11): ratchet the target onto the conversion groups (#2625)Β #2965 ratcheted onto the conversion groups β theC4310in PR build(obs-qsv11): mask in unsigned so MSVC keeps the constant intact (#2965)Β #2969, which is in CI now.Decision needed:policy question (deps include directories asSYSTEMvs. stopping the burndown) escalated on build: master's macOS lanes are red in frontend/utility/platform-osx.mm.hpp and shared/properties-view/properties-view.cpp.hpp (#2625, #2872)Β #2950, and it is the only thing keeping the two macOS lanes and clang-analyze dark.Keeping this open until #2969 lands and the macOS decider answers, so the app-lane board is green end to end and not just for the frontend group.
zackees commentedon Oct 30, 2026
Status (worker tick,
obs-studio2): two things, one fix and one verdict.1. The last release-lane site in this issue's original log is now a PR. #2971 states the conversion at
frontend/utility/ExtraBrowsersModel.cpp.hpp:190(int i = deleted.size() - 1βqsizetypetoint), which is the one of the four diagnostics inRelease OBS-Studio-Rustrun 40906899118 that #2940 left behind. ItsBuild π§± / Qt Frontend Linux π§lane is the direct GREEN evidence; asan is selected too, because the file is a compile input below the ratchetedfrontenddirectory. Auto-merge is on.2. The red nightly
Sanitizersrun of 2026-10-29 13:40 (commit32c4bcc1) is not a live failure. Every diagnostic in it is fixed by a commit merged after that run started:obs-ffmpeg-vaapi.c.h:434,435(1 << log2_chroma)6c5fd4516) β now1u <<rtmp-av1.c.h:109,499,500,mp4-mux.c.h:524,rtmp-hevc.c.h:827,843v4l2-helpers.c.h:20(PRIuFAST32vsuint)(uint_fast32_t)at both call sitesI checked the current sources rather than trusting the run, and the lane was already re-run green on master after those merges:
workflow_dispatchrun 40962690971 on9ffa4e65finished success forasan π§ͺ,valgrind π§ͺandtsan π§ͺ(and 40942190769 on6ca10e79likewise). So the schedule should be green on its next pass; I canceled the redundant dispatch I had started before finding that.What is still red on the app lanes is only the
.depsQt 6.11 header question on #2950, unchanged by this tick.zackees commentedon Oct 30, 2026
Health update after #2968 and #2969 landed β the Linux/Flatpak half is now green on master, the Windows half was one regression from #2965 and is fixed.
scheduled.yamlon master43af74187(40978686832, the commit before #2969):Build π§± / Ubuntu π§ (OBS-Studio-Rust)β success. The twoplugins/frontend-tools/scripts.cpp.hppsites build(frontend-tools): state the two conversions the app lanes stop on (#2625, #2938)Β #2968 states are gone.Build π§± / Flatpak π¦β success, same.Build π§± / Windows πͺ (x64)β failure, and the only diagnostic is a regression build(obs-qsv11): ratchet the target onto the conversion groups (#2625)Β #2965 introduced when it ratchetedplugins/obs-qsv11:obs-qsv11.c.h(1072,25)warning C4310 "cast truncates constant value", promoted to error C2220 by warnings-as-errors. The mask value was right; the cast truncated the constant. Fixed by build(obs-qsv11): mask in unsigned so MSVC keeps the constant intact (#2965)Β #2969, merged as5e25ac58; the Windows lane on that branch is success (40975673680, jobBuild π§± / Windows πͺ (x64)), and(arm64)too.Build π§± / macOS π (arm64)and(x86_64)β failure, and every remaining diagnostic is the Qt 6.11 deps headers (qnamespace.h:1926,1930,1935,qversiontagging.h:80,int -> quintptr). That is build: master's macOS lanes are red in frontend/utility/platform-osx.mm.hpp and shared/properties-view/properties-view.cpp.hpp (#2625, #2872)Β #2950's openhuman-review-neededdecision (mark the app's.depsincludesSYSTEM), not this issue.Nothing from the original report is left on any lane. Once #2950's decision is made and its PR lands, this issue can be closed; until then the only red is that decision.
zackees commentedon Oct 30, 2026
Status on current master
e2782b4ac(scheduled.yamlrun 40983897736, worker tickobs-studio2):Build π§± / Ubuntu π§ (OBS-Studio-Rust),Build π§± / Flatpak π¦,Build π§± / Qt Frontend Linux π§: success β this issue's frontend conversion scope stays green.Build π§± / Windows πͺx64 and arm64: success β build(obs-qsv11): mask in unsigned so MSVC keeps the constant intact (#2965)Β #2969 cleared the build(obs-qsv11): ratchet the target onto the conversion groups (#2625)Β #2965C4310regression, so the last non-Apple app lane is restored too.Build π§± / macOS π (arm64): failure, and only in the Qt 6.11 deps headers (qkeycombination.h1926/1930,qnamespace.h:1935,qversiontagging.h:80). That is the escalatedDecision needed:on build: master's macOS lanes are red in frontend/utility/platform-osx.mm.hpp and shared/properties-view/properties-view.cpp.hpp (#2625, #2872)Β #2950, not this issue; the x86_64 and clang-analyze jobs were cancelled behind it, so they did not report.Release lane: #2971 (
ExtraBrowsersModel.cpp.hpp:190, the one site from this issue'sRelease OBS-Studio-Rustlog that #2940 left behind) is merge-queue position 1.So nothing in this issue's original list is red any more; once #2971 lands and #2950's include-directory decision is answered, this tracker has no children left and can close.
zackees commentedon Oct 30, 2026
Lane evidence on current master
e2782b4ac(worker tick,obs-studio4), and the one thing this run could not answer.scheduled.yamldispatch run 40983897736 one2782b4ac(the #2970 frontend-api ratchet):Build π§± / Ubuntu π§ (OBS-Studio-Rust)Build π§± / Flatpak π¦Build π§± / Windows πͺ (x64)Build π§± / Windows πͺ (arm64)Build π§± / Qt Frontend Linux π§Build π§± / macOS π (arm64)The Windows half of this issue is closed on master. #2969 (
5e25ac58) is in this tree and both Windows app builds are green, so theobs-qsv11.c.hC4310/C2220residual the last comment was waiting on is gone.#2970's frontend-api ratchet is verified, not just PR-green.
e2782b4acis the commit that putsobs-frontend-apion-Wconversion;-Wsign-conversion;-Wfloat-conversion, and PR CI never compiles that library with the app configure β the Ubuntu, Flatpak and both Windows app builds above do, and all four are green. So this ratchet does not repeat the #2938 shape (a target turned on with unconverted sources behind a lane PR CI does not build).What is still red is only the escalated decision.
Build π§± / macOS π (arm64)job 123830620422 stops on:every one inside
.deps/obs-deps-qt6-2026-10-25-universalparsed into thefrontend.signal-posixunity group. That is #2950's openDecision needed:(mark the app's.depsincludesSYSTEM), unchanged by this tick; there is no first-party conversion to write.One honest gap in the table: I started a duplicate
scheduled.yamldispatch on the same ref before finding the one above, and the concurrency group cancelled both runs, somacOS π (x86_64),Analyze π¬ / macOS π (clang-analyze)andFast Gate β‘endedcancelledin this run and carry no evidence either way. The arm64 build reaching the deps headers after compiling every first-party unity group is the same signal those jobs gave on43af74187.PR #2971 (
ExtraBrowsersModel.cpp.hpp:190) is green and queued.zackees commentedon Oct 30, 2026
The macOS half now has its fix on master: #2979 (
b0f2e8e3a, 04:21 UTC) keeps the frontend's-Wsign-conversionopt-in off the Darwin TUs, which is the #2950 decision (option B) that was the last thing keepingBuild π§± / macOS π (arm64),(x86_64)andAnalyze π¬ / macOS πred.It was green on the branch across five lanes (run 40996937748) and is now re-running on master:
scheduled.yamldispatch 41002744398 onb0f2e8e3a. That run's Ubuntu, Flatpak, Windows x64/arm64, macOS arm64/x86_64 and clang-analyze jobs are this tracker's acceptance list; Linux, Flatpak and Windows were already green ate2782b4ac. I will close here with the verdict once the run finishes, and the only follow-up this leaves is #2989 (restoring the Unix-side ratchet on Darwin later), not a red lane.zackees commentedon Oct 30, 2026
Every app lane this issue was filed for is green on master, so the tracker closes.
scheduled.yamldispatch 41002744398 on masterb0f2e8e3a(#2979):Build π§± / Ubuntu π§ (OBS-Studio-Rust)Build π§± / Flatpak π¦Build π§± / Windows πͺ (x64)and(arm64)Build π§± / macOS π (arm64)and(x86_64)Analyze π¬ / macOS π (clang-analyze)The red set that opened this issue β the 63 unconverted frontend files,
ExtraBrowsersModel,WhatsNewInfoThread,update-helpers,PluginModuleLoader_MacOS, thefrontend-tools/scripts.cpp.hpppair (#2968) and theobs-qsv11C4310regression (#2969) β is gone from every lane. The last piece was macOS: #2979 keeps the frontend's-Wsign-conversionopt-in off the Darwin TUs, which is #2950's option B, because those lanes stopped inside Qt 6.11's own headers rather than in obs sources.Remaining follow-ups, none of them a red lane:
-Wsign-conversionratchet on the Darwin frontend TUs once the Qt deps prefixes can be madeSYSTEMincludes under Xcode.CEF Chromium Support πstill fails in these runs by design; it is the CEF 7871 pin, unrelated to the conversion work.Nothing else on the app-lane board is red. Closing.
build(frontend): keep the unconverted app target on the global converβ¦
build: state the frontend conversion sites the app lanes stop on (obsβ¦
fix(frontend): state the audio mixer bit masks the asan lane's sign-cβ¦
build(frontend-tools): state the two conversions the app lanes stop on (
build(frontend): state the row-index conversion the release lane stopβ¦