Skip to content

Upgrade to upstream WebKit 3722912ff800 - #383

Merged
sosukesuzuki merged 1015 commits into
mainfrom
bun/upgrade-to-3722912ff800
Aug 3, 2026
Merged

sosukesuzuki merged 1015 commits into
mainfrom
bun/upgrade-to-3722912ff800

Conversation

@robobun

@robobun robobun commented Aug 2, 2026

Copy link
Copy Markdown
Collaborator

Merges upstream WebKit/WebKit 3722912ff800 (2026-08-02) into the Bun fork.

Previous sync point: 01aaa3e0be0c (2026-07-25). 474 upstream commits, 110 touching JSC/WTF/bmalloc.

Merge notes

  • c8878bda5421 records 01aaa3e0be0c as an ancestor via -s ours: oven-sh/WebKit Upgrade to upstream WebKit 01aaa3e0be0c #352 landed that upgrade as a squash commit, so without this git computed the merge-base back at 2603e9eb41f0 (Jul 17) and produced ~380 spurious conflicts. Same pattern as c01c66740939.
  • 17 real conflicts resolved in JSC/WTF; fork's USE(BUN_JSC_ADDITIONS) hunks preserved.

Follow-up fixups for upstream's 32-bit removal (29ceb3c03de3, 857bd4334690)

Upstream deleted USE_JSVALUE64/USE_JSVALUE32_64, is64Bit(), ARMv7, and the payload/tag split. Fork-only code that referenced those:

  • ffi/*: dropped #if USE(JSVALUE64) guards (now always-false), dropped dead !is64Bit() bailout, payloadFor -> lowWordFor.
  • tools/JSDollarVM.cpp: dropped && USE(JSVALUE64) from the FFI helper guards.
  • inspector/agents/InspectorDebuggerAgent.cpp: handle BunTranspiledModule/Synthetic in the new scriptTypeForScript switch.
  • CMakeLists.txt: dropped StringRecursionChecker.h from the private headers list (removed upstream in f2f2c2ddf637).

Bun-side changes required

AbstractModuleRecord now threads ScriptFetchParameters::Type through export/star entries (upstream 90b2ecf79ae3). Four call sites in src/jsc/bindings/BunAnalyzeTranspiledModule.cpp need the extra argument; handled in the bun PR.

Verified: jsc debug build on Linux x64, full bun run build:local links and -p '42' runs.

achristensen07 and others added 30 commits July 29, 2026 09:11
https://bugs.webkit.org/show_bug.cgi?id=320479
rdar://183446539

Reviewed by Youenn Fablet.

This re-merges 318000@main but with an additional check of
__IPHONE_OS_VERSION_MIN_REQUIRED >= 260400 because the ios-safer-cpp
bot is still building with the iOS 26.0 SDK, which doesn't have the
needed Network.framework symbols.

Test: Tools/TestWebKitAPI/Tests/WebKit/WKWebView/WebTransport.mm

* Source/WTF/wtf/PlatformHave.h:
* Source/WebKit/NetworkProcess/cocoa/NetworkSoftLink.h:
* Source/WebKit/NetworkProcess/cocoa/NetworkSoftLink.mm:
* Source/WebKit/NetworkProcess/webtransport/NetworkTransportSession.cpp:
* Source/WebKit/NetworkProcess/webtransport/NetworkTransportSession.h:
* Source/WebKit/NetworkProcess/webtransport/NetworkTransportStream.cpp:
* Source/WebKit/NetworkProcess/webtransport/NetworkTransportStream.h:
* Source/WebKit/NetworkProcess/webtransport/cocoa/NetworkTransportSessionCocoa.mm:
(WebKit::createParameters):
(WebKit::NetworkTransportSession::create):
(WebKit::NetworkTransportSession::initialize):
(WebKit::NetworkTransportSession::setupDatagramConnection):
(WebKit::NetworkTransportSession::createStream):
(WebKit::NetworkTransportSession::terminate):
(WebKit::NetworkTransportSession::isSessionClosed const):
(WebKit::NetworkTransportSession::exportKeyingMaterial):
* Source/WebKit/NetworkProcess/webtransport/cocoa/NetworkTransportStreamCocoa.mm:
(WebKit::NetworkTransportStream::initializeReadyConnection):
(WebKit::NetworkTransportStream::cancelReceive):
(WebKit::NetworkTransportStream::cancelSend):
* Source/WebKit/Shared/WebPreferencesDefaultValues.cpp:
(WebKit::defaultWebTransportEnabled):
* Tools/TestWebKitAPI/Helpers/cocoa/NetworkConnection.h:
* Tools/TestWebKitAPI/Helpers/cocoa/NetworkConnection.mm:
(TestWebKitAPI::Connection::abortReads):
(TestWebKitAPI::Connection::abortWrites):
(TestWebKitAPI::Connection::setRemoteReceiveErrorHandler):
(TestWebKitAPI::Connection::setRemoteSendErrorHandler):
(TestWebKitAPI::ConnectionGroup::drainWebTransportSession):
* Tools/TestWebKitAPI/Helpers/cocoa/WebTransportServer.h:
* Tools/TestWebKitAPI/Helpers/cocoa/WebTransportServer.mm:
(TestWebKitAPI::WebTransportServer::isAvailable): Deleted.
* Tools/TestWebKitAPI/Tests/WebKit/WKWebView/WebTransport.mm:
(TestWebKitAPI::TEST(WebTransport, ClientBidirectional)):
(TestWebKitAPI::TEST(WebTransport, ClientBidirectionalBYOB)):
(TestWebKitAPI::TEST(WebTransport, Datagram)):
(TestWebKitAPI::TEST(WebTransport, Unidirectional)):
(TestWebKitAPI::TEST(WebTransport, ServerBidirectional)):
(TestWebKitAPI::TEST(WebTransport, NetworkProcessCrash)):
(TestWebKitAPI::TEST(WebTransport, Worker)):
(TestWebKitAPI::TEST(WebTransport, WorkerAfterNetworkProcessCrash)):
(TestWebKitAPI::TEST(WebTransport, ServiceWorker)):
(TestWebKitAPI::CreateStreamsBeforeReady)):
(TestWebKitAPI::DISABLED_CSP)):
(TestWebKitAPI::ServerCertificateHashes)):
(TestWebKitAPI::ServerConnectionTermination)):
(TestWebKitAPI::BackForwardCache)):
(TestWebKitAPI::ServerDrain)):
(TestWebKitAPI::DISABLED_ClientStreamAborts)):
(TestWebKitAPI::DISABLED_ServerStreamAborts)):

Canonical link: https://commits.webkit.org/318161@main
…buted-string-across-shadow-boundaries tests are constant TEXT failures.

https://bugs.webkit.org/show_bug.cgi?id=320539
rdar://182859944

Reviewed by Tim Horton and Abrar Rahman Protyasha.

317100@main ("Refactor Style::ComputedStyleProperties generation") made adjacent NSAttributedString ranges that carry
identical attributes coalesce into a single run across shadow boundaries. This is a progression: the runs had byte-identical
attributes (font, kern, stroke, color), so merging them into one range is the canonical form -- no attribute value changed. The
cross-platform baselines already expected the merged output; only the stale mac-wk1 overrides still held the pre-merge split form.

* LayoutTests/platform/mac-wk1/editing/mac/attributed-string/attributed-string-across-shadow-boundaries-1-expected.txt:
* LayoutTests/platform/mac-wk1/editing/mac/attributed-string/attributed-string-across-shadow-boundaries-2-expected.txt:
* LayoutTests/platform/mac-wk1/editing/mac/attributed-string/attributed-string-across-shadow-boundaries-3-expected.txt:
* LayoutTests/platform/mac-wk1/editing/mac/attributed-string/attributed-string-across-shadow-boundaries-4-expected.txt:
* LayoutTests/platform/mac-wk1/editing/mac/attributed-string/attributed-string-across-shadow-boundaries-5-expected.txt:

Canonical link: https://commits.webkit.org/318162@main
https://bugs.webkit.org/show_bug.cgi?id=320533
rdar://183490993

Reviewed by Yijia Huang.

Greedy * or + non-capturing parens in the tail position does not need inter-iteration backtracking state.

1. For * case (term.quantityMinCount == 0)

    Pattern like /(?:AA|A)*/ can never fail, because * succeeds even with zero iterations.
    And because it is in the tail position, all backtracking are coming from the
    greedy matching attempt of (?:AA|A). This means that we will never restore the
    previous iteration's state to explore a different matching. When we failed in the current
    iteration, we can just say "the matching is complete" because we are at the terminal position.

2. For + case (term.quantityMinCount == 1)

    Pattern like /(?:AA|A)+/ will need to match at least once. If we failed in this 1st iteration,
    since this is the initial iteration, we have no context to restore, and we can simply fail.
    If we matched once and in the second iteration, if we failed to match, then we also do not need
    to restore to the 1st iteration since we already succeeded the 1st iteration (and + suffices),
    and since it is a terminal position, we can just say "the matching is complete"

As a result, term.quantityMinCount <= 1 cases never require per-iteration ParenContext when
we have these greedy patterns at the terminal position. Thus we can mark them `isTerminal` to
skip ParenContext generation as an optimization.

On the other hand, term.quantityMinCount >= 2 cannot work in this way. Let's have a RegExp matching
`/(?:AA|A){2,}/.exec("AA")`. The right answer is "AA" since "A" and "A" matches with 2 iterations.
But this is achieved by restoring the 1st iteration and taking a different alternative "A" instead of "AA".
Thus, inter-iteration backtracking state is necessary for this case.

Tests: JSTests/stress/yarr-terminal-parentheses-captures.js
       JSTests/stress/yarr-terminal-parentheses-frame-layout.js
       JSTests/stress/yarr-terminal-parentheses-min-count.js

* JSTests/stress/yarr-terminal-parentheses-captures.js: Added.
(shouldBe):
(testExec):
* JSTests/stress/yarr-terminal-parentheses-frame-layout.js: Added.
(shouldBe):
(test):
* JSTests/stress/yarr-terminal-parentheses-min-count.js: Added.
(shouldBe):
(test):
* Source/JavaScriptCore/yarr/Yarr.h:
* Source/JavaScriptCore/yarr/YarrInterpreter.cpp:
(JSC::Yarr::Interpreter::matchParenthesesTerminalBegin):
(JSC::Yarr::Interpreter::matchParenthesesTerminalEnd):
(JSC::Yarr::Interpreter::backtrackParenthesesTerminalBegin):
* Source/JavaScriptCore/yarr/YarrJIT.cpp:
* Source/JavaScriptCore/yarr/YarrPattern.cpp:
(JSC::Yarr::YarrPatternConstructor::checkForTerminalParentheses):
* Source/JavaScriptCore/yarr/YarrPattern.h:
(JSC::Yarr::BackTrackInfoParenthesesTerminal::entryPositionIndex):

Canonical link: https://commits.webkit.org/318163@main
…one device system-wide on every capture start (unconditional VPIO instantiation)

rdar://183517329
https://bugs.webkit.org/show_bug.cgi?id=320452

Reviewed by Eric Carlson.

To remove microphone interruption when echo cancellation is off, we should disable VPIO prewarming in GPUProcess when not needed.
To do so, we compute in WebPageProxy whether we will use echo cancellation from the gum request constraints.
We then pass that information via IPC to GPUConnectionToWebProcess::updateCaptureAccess which will prewarm the VPIO unit only if echo cancellation is on.
In case of GPU process crash, we lot the info whether echo cancellation is on or off.
We do not prewarm the VPIO unit, as this is quite rare and minimising the work on GPU process crash is probably best.

Manually tested.

* Source/WebCore/platform/mediastream/MediaConstraints.cpp:
(WebCore::MediaConstraints::willUseEchoCancellation const):
* Source/WebCore/platform/mediastream/MediaConstraints.h:
* Source/WebKit/GPUProcess/GPUConnectionToWebProcess.cpp:
(WebKit::GPUConnectionToWebProcess::updateCaptureAccess):
* Source/WebKit/GPUProcess/GPUConnectionToWebProcess.h:
* Source/WebKit/GPUProcess/GPUProcess.cpp:
(WebKit::GPUProcess::createGPUConnectionToWebProcess):
(WebKit::GPUProcess::updateCaptureAccess):
* Source/WebKit/GPUProcess/GPUProcess.h:
* Source/WebKit/GPUProcess/GPUProcess.messages.in:
* Source/WebKit/UIProcess/GPU/GPUProcessProxy.cpp:
(WebKit::GPUProcessProxy::updateCaptureAccess):
* Source/WebKit/UIProcess/GPU/GPUProcessProxy.h:
* Source/WebKit/UIProcess/UserMediaPermissionRequestProxy.cpp:
(WebKit::UserMediaPermissionRequestProxy::willUseEchoCancellation const):
* Source/WebKit/UIProcess/UserMediaPermissionRequestProxy.h:
* Source/WebKit/UIProcess/WebPageProxy.cpp:
(WebKit::WebPageProxy::willStartCapture):
(WebKit::WebPageProxy::gpuProcessExited):

Canonical link: https://commits.webkit.org/318164@main
…ne' sharing groups

https://bugs.webkit.org/show_bug.cgi?id=320558

Reviewed by Antti Koivisto.

A flex line is at least as tall as each of its baseline-sharing groups, and a group is as tall as its largest
ascent plus its largest descent. align-self: baseline and align-self: last baseline align to opposite edges of the
line, so they are two separate groups.

crossSizeForFlexLines kept the two ascents apart but shared one descent between them, and each group's total read
that shared value. So a last-baseline item's descent was added to the first-baseline group's ascent, and the other
way round, making the line taller than either group really is. It also made the result depend on the order the
items appear in, since a descent only reached the other group once it had been seen.

Give the last-baseline group its own descent. Both groups now sum the ascent and descent of their own members.

Chrome and Firefox both size these lines correctly.

* LayoutTests/imported/w3c/web-platform-tests/css/css-flexbox/flexbox-baseline-and-last-baseline-groups-line-size-expected.txt: Added.
* LayoutTests/imported/w3c/web-platform-tests/css/css-flexbox/flexbox-baseline-and-last-baseline-groups-line-size.html: Added.
* Source/WebCore/layout/formattingContexts/flex/FlexFormattingContext.cpp:
(WebCore::FlexFormattingContext::crossSizeForFlexLines):

Canonical link: https://commits.webkit.org/318165@main
https://bugs.webkit.org/show_bug.cgi?id=317022

Reviewed by Keith Miller.

recursivelyBuildSwitch picks a jump table when (last - first + 1) / n is
below a density limit. When case values span more than 2^63, that
subtraction overflows signed int64; the wrapped result can look dense and
tableSize becomes enormous, OOMing in BitVector / data-section allocation.

Compute the span with CheckedInt64 and skip the table path on overflow.
Add a testb3 case with the sparse i64 values from the bug report.

* Source/JavaScriptCore/b3/B3LowerMacros.cpp:
(JSC::B3::LowerMacros::recursivelyBuildSwitch): Checked span for density.
* Source/JavaScriptCore/b3/testb3.h:
* Source/JavaScriptCore/b3/testb3_1.cpp:
* Source/JavaScriptCore/b3/testb3_5.cpp:
(testSwitchSparseI64RangeOverflow): Added.

Canonical link: https://commits.webkit.org/318166@main
https://bugs.webkit.org/show_bug.cgi?id=312794

Reviewed by Yusuke Suzuki.

Memory64 bulk ops (fill/copy/init) are implemented in OMG, but this stress
test disabled OMG with --useOMGJIT=0. Drop that flag so default tiering and
wasmTestLoopCount can reach OMG (the path that regressed JetStream3).

* JSTests/wasm/stress/memory64-bulk-memory.js:

Canonical link: https://commits.webkit.org/318167@main
https://bugs.webkit.org/show_bug.cgi?id=277686

Reviewed by Yusuke Suzuki.

BBQ used to emit a full binary search over every br.table index even when
long stretches shared the same target. Collapse consecutive identical
targets into runs and dispatch with unsigned range checks. Prefer the
jump table only when the number of runs is large enough, so dense tables
with few runs stay on the shallow path.

BinarySwitch takes those runs via a dedicated constructor over
span of (caseValue, runLength), rather than overloading the flat
span<int64_t> case list.

Stress forces BBQ and skips wasm-no-jit / wasm-no-wasm-jit modes,
which would set IPInt or BBQ off against the required options.

* Source/JavaScriptCore/jit/BinarySwitch.h:
* Source/JavaScriptCore/jit/BinarySwitch.cpp:
(JSC::BinarySwitch::BinarySwitch):
(JSC::BinarySwitch::advance):
(JSC::BinarySwitch::buildCheckRuns):
* Source/JavaScriptCore/wasm/WasmBBQJIT.cpp:
(JSC::Wasm::BBQJIT::addSwitch):
* JSTests/wasm/stress/br-table-consecutive-target-runs.js: Added.
* JSTests/microbenchmarks/wasm-bbq-br-table-runs.js: Added.

Canonical link: https://commits.webkit.org/318168@main
https://bugs.webkit.org/show_bug.cgi?id=142657
<rdar://problem/20149674>

Reviewed by Yusuke Suzuki.

Rather than trying to keep a manual list, leverage the pretty format of IDL to autogenerate `WI.NativeConstructorFunctionParameters` and `WI.NativePrototypeFunctionParameters`.

* Source/WebCore/bindings/scripts/CodeGenerator.pm:
(ProcessInterfaces):
(WriteInspectorNativeFunctionParameters): Added.
Modify the IDL parser to output `*.inspector-native-function-parameters.json` files containing a pretty string of the arguments for any class-level and instance-level methods.

* Source/WebCore/bindings/scripts/combine-inspector-native-function-parameters.pl: Added.
Combine all `*.inspector-native-function-parameters.json` into a single `InspectorNativeFunctionParameters.json`.

* Source/WebInspectorUI/UserInterface/Models/NativeFunctionParameters.js: Removed.
* Source/WebInspectorUI/UserInterface/Models/NativeFunctionParameters-overrides.json: Added.
Pure JS APIs do not have anything like pretty formatted IDL to draw from like other web APIs.
Instead, have these be handwritten in a JSON file until a solution can be found for them.
This also allows for Web Inspector to override anything that is automatic if needed.

* Source/WebInspectorUI/Scripts/update-NativeFunctionParameters.py: Added.
Merge the `InspectorNativeFunctionParameters.json` with the overrides file and generate `NativeFunctionParameters.js`.
If WebCore's generated JSON cannot be read or parsed, warn and generate an overrides-only file rather than failing the build, since these parameter names are only a minor visual enhancement.

* Source/WebInspectorUI/Scripts/copy-user-interface-resources.pl:
Read `InspectorNativeFunctionParameters.json` from WebCore's private headers and copy `NativeFunctionParameters.js` alongside all other resources.

* Source/WebInspectorUI/UserInterface/Views/ObjectTreePropertyTreeElement.js:
(WI.ObjectTreePropertyTreeElement.prototype._functionParameterString):
Drive-by: also handle `Atomics`, `Intl`, and `WebAssembly` namespace objects.

* Source/CMakeLists.txt:
* Source/WebCore/CMakeLists.txt:
* Source/WebCore/DerivedSources.make:
* Source/WebCore/DerivedSources-output.xcfilelist:
* Source/WebCore/WebCoreMacros.cmake:
* Source/WebCore/WebCore.xcodeproj/project.pbxproj:
* Source/WebInspectorUI/CMakeLists.txt:
* Source/WebInspectorUI/WebInspectorUI.xcodeproj/project.pbxproj:
Generate and consume the parameter data in each build system.
Export `InspectorNativeFunctionParameters.json` as a WebCore private header and consume it from `WEBCORE_PRIVATE_HEADERS_DIR`, matching the existing `InspectorBackendCommands.js` pattern.
This avoids a cross-project target dependency, which cannot be satisfied when WebInspectorUI is built standalone against installed WebKit.
Depend on `WebCore_CopyPrivateHeaders` in CMake and add `WebCore.framework` to Xcode's `Product Dependencies` build phase so combined builds still order WebCore before WebInspectorUI.

* Source/WebCore/bindings/scripts/test/JS/TestCallTracer.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestCallbackInterface.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestConditionalIncludes.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestDOMJIT.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestDomainSecurity.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestEnabledBySetting.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestEventTarget.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestGlobalObject.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestIndexedSetterWithIdentifier.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestInterface.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestLegacyOverrideBuiltIns.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestNamedAndIndexedSetterWithIdentifier.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestNamedDeleterWithIdentifier.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestNamedGetterWithIdentifier.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestNamedSetterWithIdentifier.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestNamedSetterWithIndexedGetter.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestNamedSetterWithIndexedGetterAndSetter.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestNamespaceObject.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestObj.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestScheduledAction.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestSerializedScriptValueInterface.inspector-native-function-parameters.json: Added.
* Source/WebCore/bindings/scripts/test/JS/TestTypedefs.inspector-native-function-parameters.json: Added.

Canonical link: https://commits.webkit.org/318169@main
…struction

https://bugs.webkit.org/show_bug.cgi?id=310640
rdar://problem/173252605

Reviewed by Abrar Rahman Protyasha

Remove construction of ISO-18013 request info

* Source/WTF/wtf/PlatformEnable.h:
* Source/WebCore/Headers.cmake:
* Source/WebCore/Modules/identity/DigitalCredentialsMobileDocumentRequestDataWithRequestInfo.h: Removed.
* Source/WebCore/Modules/identity/DigitalCredentialsRequestData.h:
* Source/WebCore/Modules/identity/DigitalCredentialsRequestDataBuilder.cpp:
* Source/WebCore/Modules/identity/DigitalCredentialsRequestDataBuilder.h:
* Source/WebCore/WebCore.xcodeproj/project.pbxproj:
* Source/WebKit/Shared/WebCoreArgumentCodersAuth.serialization.in:
* Source/WebKit/UIProcess/DigitalCredentials/WKDigitalCredentialsPicker.mm:
* Source/WebKit/UIProcess/WebPageProxy.h:
* Source/WebKit/UIProcess/WebPageProxy.messages.in:
* Source/WebKit/WebProcess/DigitalCredentials/DigitalCredentialsCoordinator.messages.in:

Canonical link: https://commits.webkit.org/318170@main
…s on Linux

https://bugs.webkit.org/show_bug.cgi?id=320559

Unreviewed test gardening.

The Linux ports (GTK and WPE) build and test on the CI with asserts enabled
for Release.

This tests are quite slow on a standard release build, but on release+asserts they
often timeout completely. On the CI each one of this tests is run several times
with different configurations, and at least 15-20 of those runs usually timeout.
Each one of those timeouts costs around ~10 minutes (because first it hits
the soft timeout and then the hard, so around 5+5) with the CPU spinning like
crazy and a worker waiting that can't advance to the next test.

* JSTests/wasm/stress/type-index-abstract-heap-types-concrete-vs-abstract.js:
* JSTests/wasm/stress/type-index-abstract-heap-types-globals-and-tables.js:
* JSTests/wasm/stress/type-index-abstract-heap-types-nulls-and-casts.js:
* JSTests/wasm/stress/type-index-abstract-heap-types-subtype-validation.js:

Canonical link: https://commits.webkit.org/318171@main
…am before WebTransport.ready should fail if the server closes the connection

https://bugs.webkit.org/show_bug.cgi?id=320510
rdar://183471508

Reviewed by Chris Dumez.

Like getStats requests, createStream requests need to wait until the session has been initialized
either successfully or unsuccessfully before completing.

* LayoutTests/imported/w3c/web-platform-tests/webtransport/close.https.any-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/webtransport/close.https.any.serviceworker-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/webtransport/close.https.any.sharedworker-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/webtransport/close.https.any.worker-expected.txt:
* Source/WebKit/NetworkProcess/webtransport/NetworkTransportSession.cpp:
(WebKit::NetworkTransportSession::~NetworkTransportSession):
(WebKit::NetworkTransportSession::completeRequestsAfterInitialization):
(WebKit::NetworkTransportSession::completeStatsRequestsAfterInitialization): Deleted.
* Source/WebKit/NetworkProcess/webtransport/NetworkTransportSession.h:
* Source/WebKit/NetworkProcess/webtransport/cocoa/NetworkTransportSessionCocoa.mm:
(WebKit::NetworkTransportSession::initialize):
(WebKit::NetworkTransportSession::createStream):

Canonical link: https://commits.webkit.org/318172@main
https://bugs.webkit.org/show_bug.cgi?id=320235

Reviewed by Darin Adler.

Now that all <length> and <length-percentage> values can use evaluation time zoom,
we no longer have anything that requires style building time zoom. That means its
a good time to remove the EvaluationTimeZoomEnabled setting, as we can get the complete
benefit from its removal now.

Along with removing the setting, all the zoom state on CSSToLengthConversionData can
be remove.

One place that zoom was still being used was in the theme code when using the version
of emToPx that takes a Style::ComputedStyle. Its very unclear from call sites that it
is distinct, so a new function, emToPxZoomed was added for those callers, and now the
normal emToPx is always unzoomed. There is quite a bit of cleanup in the theme code
around zoom that can be done, but that is not handled in this change.

* Source/WTF/Scripts/Preferences/UnifiedWebPreferences.yaml:
* Source/WebCore/css/CSSProperties.json:
* Source/WebCore/css/CSSToLengthConversionData.cpp:
* Source/WebCore/css/CSSToLengthConversionData.h:
* Source/WebCore/css/parser/SizesAttributeParser.cpp:
* Source/WebCore/css/query/MediaQueryEvaluator.cpp:
* Source/WebCore/css/values/primitives/CSSPrimitiveNumericRange.h:
* Source/WebCore/dom/Document.cpp:
* Source/WebCore/html/canvas/CanvasRenderingContext2DBase.cpp:
* Source/WebCore/html/shadow/TextControlInnerElements.cpp:
* Source/WebCore/platform/graphics/FontDescription.cpp:
* Source/WebCore/platform/graphics/FontDescription.h:
* Source/WebCore/rendering/RenderBlock.cpp:
* Source/WebCore/rendering/RenderTheme.cpp:
* Source/WebCore/rendering/cocoa/RenderThemeCocoa.mm:
* Source/WebCore/rendering/ios/RenderThemeIOS.mm:
* Source/WebCore/rendering/mac/RenderThemeMac.mm:
* Source/WebCore/rendering/mathml/RenderMathMLBlock.cpp:
* Source/WebCore/style/StyleBuilderCustom.h:
* Source/WebCore/style/StyleBuilderState.cpp:
* Source/WebCore/style/StyleExtractorCustom.h:
* Source/WebCore/style/StyleResolveForDocument.cpp:
* Source/WebCore/style/StyleResolveForFont.cpp:
* Source/WebCore/style/computed/StyleComputedStyleBase+GettersInlines.h:
* Source/WebCore/style/computed/StyleComputedStyleBase+SettersInlines.h:
* Source/WebCore/style/computed/StyleComputedStyleBase.h:
* Source/WebCore/style/computed/StyleComputedStyleProperties+InitialCustomInlines.h:
* Source/WebCore/style/computed/data/StyleInheritedRareData.cpp:
* Source/WebCore/style/computed/data/StyleInheritedRareData.h:
* Source/WebCore/style/values/backgrounds/StyleLineWidth.cpp:
* Source/WebCore/style/values/backgrounds/StyleLineWidth.h:
* Source/WebCore/style/values/inline/StyleLineHeight.cpp:
* Source/WebCore/style/values/non-standard/StyleWebKitTextStrokeWidth.cpp:
* Source/WebCore/style/values/primitives/StyleLengthResolution.cpp:
* Source/WebCore/style/values/primitives/StyleLengthResolution.h:
* Source/WebCore/style/values/primitives/StylePrimitiveNumericTypes+Conversions.cpp:
* Source/WebCore/style/values/primitives/StylePrimitiveNumericTypes+Conversions.h:
* Source/WebCore/style/values/sizing/StyleContainIntrinsicSize.cpp:
* Source/WebCore/style/values/text/StyleLetterSpacing.cpp:
* Source/WebCore/style/values/text/StyleWordSpacing.cpp:
* Source/WebCore/style/values/transforms/StylePerspective.cpp:
* Source/WebCore/style/values/transforms/StyleTransformFunction.cpp:
* Source/WebCore/svg/SVGLengthContext.cpp:
* Source/WebCore/svg/SVGLengthContext.h:

Canonical link: https://commits.webkit.org/318173@main
…ration

https://bugs.webkit.org/show_bug.cgi?id=320554
rdar://problem/183517726

Reviewed by Aditya Keerthi.

This method is not implemented (or called) anywhere, so let's remove the
declaration.

* Source/WebKit/WebProcess/WebPage/WebPage.h:

Canonical link: https://commits.webkit.org/318174@main
https://bugs.webkit.org/show_bug.cgi?id=320511
rdar://problem/183473145

Reviewed by Alan Baradlay.

A negative grid line counts backward from the end edge of the explicit grid: for
a grid with N explicit tracks, -1 is the last line and -(N+1) is the first line,
with more-negative values counting past the start edge into the leading implicit
grid. Previously explicitLineToIndex() passed negative lines through unchanged
(e.g. -1 stayed -1), which is not quite how they are supposed to get
resolved.

To fix this we can just thread in the number of explicit tracks through
the code that computes the UnplacedGridItem's GridPosition so that it
takes the number of explicit tracks into account. For example, if you
have three expliti columns with lines L1 L2 L3 L4, then -1 will map to
L4 and -5 will map to the first implicit track that is created right
before the first explicit one.

* Source/WebCore/layout/formattingContexts/grid/GridFormattingContext.cpp:
(WebCore::Layout::GridFormattingContext::constructUnplacedGridItems const):
* Source/WebCore/layout/formattingContexts/grid/UnplacedGridItem.cpp:
(WebCore::Layout::explicitLineToIndex):
This is what does the mapping as described above. This means that this
may still return a negative number for those lines that are part of the
implicit grid that is part of the beginning. That is ok though since
those will be translated and be relative to the implicit grid when we
actually create the GridPosition itself.

(WebCore::Layout::UnplacedGridItem::GridPosition::create):
This is where the line numbers will get transformed to be relative to
the implicit grid. So here -5 from the example above would be fed in as
-1 but then transformed into 0 since it is the first line in the
implicit grid.

(WebCore::Layout::UnplacedGridItem::resolveDefinitePosition):
Plumb the number of explicit tracks to explicitLineToIndex so that the
values take them into consideration like described above in the
explanation of the function.

Canonical link: https://commits.webkit.org/318175@main
…rigin.y

https://bugs.webkit.org/show_bug.cgi?id=320548
rdar://183507280

Reviewed by Mike Wyrzykowski and Aditya Keerthi.

Fix the typo and rebaseline the impacted tests.

Test: Tools/TestWebKitAPI/Tests/WebKit/WKWebView/ios/CALayerTreeAsText.mm

* LayoutTests/compositing/background-color/background-color-clamping-expected.txt:
* LayoutTests/interaction-region/guard-crash-expected.txt:
* LayoutTests/interaction-region/interaction-layers-culling-layer-type-change-expected.txt:
* LayoutTests/interaction-region/layer-tree-expected.txt:
* LayoutTests/interaction-region/layer-tree-shape-reset-expected.txt:
* LayoutTests/platform/ios/compositing/background-color/background-color-clamping-expected.txt:
* LayoutTests/platform/visionos/transforms/separated-expected.txt:
* Source/WebKit/UIProcess/API/Cocoa/WKWebViewTesting.mm:
(dumpCALayer):
(-[WKWebView _propertiesOfLayerWithID:]):

* Tools/TestWebKitAPI/TestWebKitAPI.xcodeproj/project.pbxproj:
* Tools/TestWebKitAPI/Tests/WebKit/WKWebView/ios/CALayerTreeAsText.mm: Added.
(TEST(WebKit, CALayerTreeAsTextDumpsYCoordinates)):
Add an API test similar to the one from 317839@main.

Canonical link: https://commits.webkit.org/318176@main
…ringe in vertical writing modes.

https://bugs.webkit.org/show_bug.cgi?id=320538
rdar://183496114

Reviewed by Alan Baradlay.

For vertical writing modes, TextBoxPainter paints the highlight background
fill inside a rotated graphics context whose translation is derived from an
inline-axis coordinate that is intentionally not device-pixel aligned (to
preserve glyph advances). A fill snapped to integers in the box's local space
therefore lands on fractional device pixels and gets anti-aliased, leaving a
one-pixel fringe that an ordinary unrotated element background does not have.
Fix this by snapping the fill rect in device space for vertical writing modes.

imported/w3c/web-platform-tests/css/css-highlight-api/painting/custom-highlight-painting-vertical-writing-mode-001.html

* LayoutTests/TestExpectations:
* LayoutTests/platform/ios/TestExpectations:
* Source/WebCore/rendering/TextBoxPainter.cpp:
(WebCore::TextBoxPainter::paintBackgroundFillForRange):

Canonical link: https://commits.webkit.org/318177@main
https://bugs.webkit.org/show_bug.cgi?id=320491
rdar://183457853

Reviewed by Yusuke Suzuki.

This patch fixes the path run-regexp-tests tries to chdir to.

* Tools/Scripts/run-regexp-tests:

Canonical link: https://commits.webkit.org/318179@main
…FlexItem's unreachable fallback

https://bugs.webkit.org/show_bug.cgi?id=320572

Reviewed by Antti Koivisto.

flowAwareMarginBeforeForFlexItem returns a margin of the flex item, but the fallback after its switch returns one of
the flex container's. The four flow-aware accessors next to it are container-level, so returning flexBox().marginTop()
is right for them and was carried over to this one.

* Source/WebCore/layout/formattingContexts/flex/FlexFormattingUtils.cpp:
(WebCore::FlexFormattingUtils::flowAwareMarginBeforeForFlexItem const):

Canonical link: https://commits.webkit.org/318180@main
… container is sized to min-content

https://bugs.webkit.org/show_bug.cgi?id=320564

Reviewed by Antti Koivisto.

computeIntrinsicLogicalWidths measures a flex item's margins once and adds the same value to both the min-content
and the max-content width. The two are measured over different sets of flex lines, though: max-content puts every
item on one line, while min-content for a wrapping container puts a break between every item, so each one is alone
on its line.

margin-trim takes the inline margins of the items at the start and end of a line. On the single line max-content
measures there are two such items; on the min-content lines every item is one, at both ends. RenderFlexibleBox
trims the margins of the container's first and last flex items before layout, which is the right answer for
max-content and feeds the same number into min-content, where it leaves every interior item contributing margins
that will not be there.

So a container that sizes to min-content reports a width wider than the content it then lays out. Three 50px items
with 100px inline margins and margin-trim: inline-start inline-end come out 250px wide when the same container's
layout puts each item alone on its line at 50px with both margins gone, leaving 200px of the container empty.

Take the margins off the min-content contribution of every item when the container wraps and the corresponding
side is trimmed. Column flow keeps its existing behaviour: its min-content is a max over items in the cross axis,
not a per-line break.

* LayoutTests/imported/w3c/web-platform-tests/css/css-box/margin-trim/flex-min-content-trims-every-line-expected.txt: Added.
* LayoutTests/imported/w3c/web-platform-tests/css/css-box/margin-trim/flex-min-content-trims-every-line.html: Added.
* Source/WebCore/rendering/RenderFlexibleBox.cpp:
(WebCore::RenderFlexibleBox::computeIntrinsicLogicalWidths const):

Canonical link: https://commits.webkit.org/318181@main
…yncData types

https://bugs.webkit.org/show_bug.cgi?id=320517
rdar://183482026

Reviewed by Brady Eidson.

Sometimes it's desirable to log said data close to their IPC
send/receive methods, where they are usually bundled as ProcessSyncData
types. As such, this patch teaches generate-process-sync-data.py to
create TextStream operator<< overloads for said types.

We also regenerate the TestSyncData output, which had drifted from the
generator after 312295@main and 316357@main; each of those commits
altered the script without regenerating these baselines.

* Source/WebCore/Scripts/generate-process-sync-data.py:
(generate_synched_data_header):
(generate_synched_data_header.TextStream):
(generate_synched_data_impl):
* Source/WebCore/Scripts/tests/TestSyncClient.cpp:
(WebCore::TestSyncClient::broadcastAudioSessionTypeToOtherProcesses):
(WebCore::TestSyncClient::broadcastMainFrameURLChangeToOtherProcesses):
(WebCore::TestSyncClient::broadcastIsAutofocusProcessedToOtherProcesses):
(WebCore::TestSyncClient::broadcastUserDidInteractWithPageToOtherProcesses):
(WebCore::TestSyncClient::broadcastAnotherOneToOtherProcesses):
(WebCore::TestSyncClient::broadcastMultipleHeadersToOtherProcesses):
* Source/WebCore/Scripts/tests/TestSyncData.cpp:
(WebCore::TestSyncData::update):
(WebCore::operator<<):
* Source/WebCore/Scripts/tests/TestSyncData.h:
* Source/WebCore/Scripts/tests/TestSyncData.serialization.in:

Canonical link: https://commits.webkit.org/318182@main
…redProcessEnabled

https://bugs.webkit.org/show_bug.cgi?id=320495
rdar://183460049

Reviewed by Sihui Liu.

SiteIsolationSharedProcessEnabled may soon be enabled by default. While
this is a useful performance optimization to complement Site Isolation,
it _may_ mask real issues unique to frames in separate processes, since
there will be more process re-use now.

We should introduce test coverage around this feature, so let's retrofit
the EmbeddedPDFHUDSiteIsolation suite for that purpose.

* Tools/TestWebKitAPI/Tests/WebKit/WKWebView/UnifiedPDFTests.mm:
(TestWebKitAPI::EmbeddedPDFHUDSiteIsolation::siteIsolationSharedProcessEnabled const):
(TestWebKitAPI::EmbeddedPDFHUDSiteIsolation::testNameGenerator):
(TestWebKitAPI::TEST_P):

Canonical link: https://commits.webkit.org/318183@main
https://bugs.webkit.org/show_bug.cgi?id=310698

Reviewed by Yusuke Suzuki.

* JSTests/stress/iterator-prototype.js:
    This test checks whether properties on `Iterator.prototype` are expected.
* Source/WTF/Scripts/Preferences/UnifiedWebPreferences.yaml:

Canonical link: https://commits.webkit.org/318184@main
https://bugs.webkit.org/show_bug.cgi?id=320390
rdar://183347616

Reviewed by Youenn Fablet.

This makes WebTransport receive streams and bidirectional streams work with byob buffers that
don't line up exactly with what bytes are received from the network.

Tests: imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.html
       imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.serviceworker.html
       imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.sharedworker.html
       imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.worker.html

* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any-expected.txt: Added.
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.html: Added.
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.js: Added.
(async echo_bidirectional_stream):
(ascending_bytes):
(promise_test.async t):
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.serviceworker-expected.txt: Added.
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.serviceworker.html: Added.
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.sharedworker-expected.txt: Added.
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.sharedworker.html: Added.
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.worker-expected.txt: Added.
* LayoutTests/imported/w3c/web-platform-tests/webtransport/receive-stream-pull.https.any.worker.html: Added.
* Source/WebCore/Modules/webtransport/DatagramByteSource.cpp:
(WebCore::DatagramByteSource::receiveDatagram):
(WebCore::DatagramByteSource::closeStreamIfPossible):
* Source/WebCore/Modules/webtransport/WebTransport.cpp:
(WebCore::WebTransport::WebTransport):
(WebCore::WebTransport::receiveIncomingUnidirectionalStream):
(WebCore::createBidirectionalStream):
(WebCore::WebTransport::receiveBidirectionalStream):
(WebCore::WebTransport::streamReceiveBytes):
(WebCore::WebTransport::cleanup):
(WebCore::WebTransport::didDrain):
* Source/WebCore/Modules/webtransport/WebTransportBidirectionalStreamSource.h:
(WebCore::WebTransportBidirectionalStreamSource::stopReceivingIncomingStreams):
* Source/WebCore/Modules/webtransport/WebTransportReceiveStream.cpp:
(WebCore::WebTransportReceiveStream::create):
* Source/WebCore/Modules/webtransport/WebTransportReceiveStreamByteSource.cpp:
(WebCore::WebTransportReceiveStreamByteSource::pull):
(WebCore::WebTransportReceiveStreamByteSource::deliverBytes):
(WebCore::WebTransportReceiveStreamByteSource::closeStream):
(WebCore::WebTransportReceiveStreamByteSource::errorStream):
(WebCore::WebTransportReceiveStreamByteSource::receiveBytes):
(WebCore::WebTransportReceiveStreamByteSource::receiveError):
(WebCore::WebTransportReceiveStreamByteSource::cancel):
* Source/WebCore/Modules/webtransport/WebTransportReceiveStreamByteSource.h:
* Source/WebCore/Modules/webtransport/WebTransportReceiveStreamSource.h:
(WebCore::WebTransportReceiveStreamSource::stopReceivingIncomingStreams):

Canonical link: https://commits.webkit.org/318185@main
…have a click handler (matching HTML anchors)

rdar://181924326
https://bugs.webkit.org/show_bug.cgi?id=319107

Reviewed by Tyler Wilcock.

Introduces or guard clause within SVGAElement. This if statement only assigns an AccessibilityRole::Link if the element has an href link or has a click handler. Bare anchor svgs will no longer emit an AcessibilityRole::Link. Also, changes were made to update SVG objects on the AXIsolated Tree as well.

Test: accessibility/svg-anchor-with-click-handler-is-link.html

* LayoutTests/accessibility/svg-anchor-with-click-handler-is-link-expected.txt: Added.
* LayoutTests/accessibility/svg-anchor-with-click-handler-is-link.html: Added.
* Source/WebCore/accessibility/AccessibilitySVGObject.cpp:
(WebCore::AccessibilitySVGObject::determineAccessibilityRole):
* Source/WebCore/accessibility/AXObjectCache.cpp:
(WebCore::AXObjectCache::handleClickHandlerChanged):
(WebCore::AXObjectCache::handleAttributeChange):

Canonical link: https://commits.webkit.org/318186@main
…-of-process iframes

https://bugs.webkit.org/show_bug.cgi?id=320563
rdar://183529776

Reviewed by BJ Burg.

Automation commands that operate on a specific frame route their IPC to the
process hosting that frame via `sendWithAsyncReplyToProcessContainingFrame...`,
but pass `webPageIDInMainFrameProcess()` as the destination page identifier.

That identifier is only valid in the main frame's process. With site isolation,
a cross-origin (out-of-process) frame lives in a different `WebProcess`, where the
`RemotePage`'s `WebPage` is registered under a freshly generated `PageIdentifier`.

The `WebProcess`-side handler resolves the page with `WebProcess::webPage(pageID)`,
which returns null for the main frame id, so the command fails with a
`WindowNotFound ("no such window")` error.

Add `WebPageProxy::webPageIDInProcessForFrame()`, which return the page identifier valid
in the process that actually hosts the frame. For the main frame this resolves to the
main frame process and returns m_webPageID, so the non-isolated path is unchanged.
Use it at every frame-scoped automation call site.

* Source/WebKit/UIProcess/Automation/WebAutomationSession.cpp:
(WebKit::WebAutomationSession::switchToBrowsingContext):
(WebKit::WebAutomationSession::evaluateJavaScriptFunction):
(WebKit::WebAutomationSession::resolveChildFrameHandle):
(WebKit::WebAutomationSession::resolveParentFrameHandle):
(WebKit::WebAutomationSession::computeElementLayout):
(WebKit::WebAutomationSession::getComputedRole):
(WebKit::WebAutomationSession::consumeUserActivation):
(WebKit::WebAutomationSession::getComputedLabel):
(WebKit::WebAutomationSession::selectOptionElement):
(WebKit::WebAutomationSession::setFilesForInputFileUpload):
(WebKit::WebAutomationSession::viewportInViewCenterPointOfElement):
(WebKit::WebAutomationSession::evaluateBidiScript):
(WebKit::WebAutomationSession::takeScreenshot):
* Source/WebKit/UIProcess/WebPageProxy.cpp:
(WebKit::WebPageProxy::webPageIDInProcessForFrame):
* Source/WebKit/UIProcess/WebPageProxy.h:

Canonical link: https://commits.webkit.org/318187@main
https://bugs.webkit.org/show_bug.cgi?id=320136
rdar://183071308

Reviewed by Brady Eidson and Elliott Williams.

This patch moves the install location for relocatable builds of webpushd to the standard /Library/LaunchAgents
directory where it will automatically be loaded by the system. For more info, please see rdar://171543360.

* Source/WebKit/Configurations/webpushd.xcconfig:

Canonical link: https://commits.webkit.org/318188@main
…) dump target in the UI process

https://bugs.webkit.org/show_bug.cgi?id=319764
rdar://182625550

Reviewed by Alex Christensen.

Follow-up to the previous fix in b/317789.
Instead of the web process self-reporting whether it
hosts the local main frame via InjectedBundle::pageHasLocalMainFrame(),
the UI process now determines whether the synchronous notifyDone()/forceImmediateCompletion() message came from the process hosting the main frame and tells WebKitTestRunner via a new WKPageInjectedBundleClient callback.

When the message is from the main-frame process, the injected bundle completes the
dump locally, matching the non-site-isolation behavior. Otherwise the dump is routed
to the process that owns the main frame. These messages are now sent as page-level
synchronous messages so the UI process has the page context needed to make that
determination.

* Source/WebKit/UIProcess/API/C/WKPageInjectedBundleClient.h:
* Source/WebKit/UIProcess/WebPageInjectedBundleClient.cpp:
(WebKit::WebPageInjectedBundleClient::didReceiveSynchronousMessageFromInjectedBundle):
* Source/WebKit/UIProcess/WebPageInjectedBundleClient.h:
* Source/WebKit/UIProcess/WebPageProxy.cpp:
(WebKit::WebPageProxy::handleSynchronousMessage):
* Tools/WebKitTestRunner/InjectedBundle/InjectedBundle.cpp:
(WTR::InjectedBundle::pageHasLocalMainFrame const): Deleted.
* Tools/WebKitTestRunner/InjectedBundle/InjectedBundle.h:
* Tools/WebKitTestRunner/InjectedBundle/TestRunner.cpp:
(WTR::TestRunner::notifyDone):
(WTR::TestRunner::forceImmediateCompletion):
* Tools/WebKitTestRunner/TestController.cpp:
(WTR::TestController::createOtherPlatformWebView):
(WTR::TestController::createWebViewWithOptions):
* Tools/WebKitTestRunner/TestController.h:
* Tools/WebKitTestRunner/TestInvocation.cpp:
(WTR::TestInvocation::didReceiveSynchronousMessageFromInjectedBundle):
(WTR::TestInvocation::resolveNotifyDone):
* Tools/WebKitTestRunner/TestInvocation.h:
* LayoutTests/http/tests/site-isolation/force-immediate-completion-expected.txt: Added.
* LayoutTests/http/tests/site-isolation/force-immediate-completion.html: Added.
* LayoutTests/http/tests/site-isolation/resources/frame-force-immediate-completion.html: Added.
* LayoutTests/http/tests/site-isolation/resources/pointer-lock.html:
* LayoutTests/platform/ios/http/tests/site-isolation/pointer-lock-expected.txt:

Canonical link: https://commits.webkit.org/318189@main
https://bugs.webkit.org/show_bug.cgi?id=320293
rdar://183226312

Reviewed by Pascoe.

SymlinkHeaders.cmake spawned readlink through execute_process for every entry
to check whether an existing symlink already pointed at the right file. The
target that runs it, WebKit_StageFrameworkHeaders, is an add_custom_target with
no output, so ninja re-runs it on every build, over roughly 760 and 68 entries.

Use file(READ_SYMLINK), which needs no subprocess: re-running the script over a
directory of 503 already-staged headers drops from 1.72s to 0.02s.

* Source/cmake/SymlinkHeaders.cmake:

Canonical link: https://commits.webkit.org/318190@main
chicoxyzzy and others added 16 commits August 1, 2026 20:07
https://bugs.webkit.org/show_bug.cgi?id=319894

Reviewed by Yusuke Suzuki.

Regression coverage for i32 extended constant expressions used as data/elem
offsets and global inits (mod 2^32). Production readers already truncate typed
i32 results (see 314444 / 317633@main for memory32 data offsets).

* JSTests/wasm/stress/const-expr-i32-wrap.js: Added.

Canonical link: https://commits.webkit.org/318395@main
…r copies

https://bugs.webkit.org/show_bug.cgi?id=166701

Reviewed by Yusuke Suzuki.

customSections already allocates a new ArrayBuffer via tryCreate(span)
and memcpy. Cover the bug 166701 requirement that mutating a returned
buffer does not affect later customSections results or other returned
buffers, and that successive calls do not reuse the same objects.

* JSTests/wasm/js-api/Module.customSection.js:

Canonical link: https://commits.webkit.org/318396@main
https://bugs.webkit.org/show_bug.cgi?id=319826

Reviewed by Keith Miller.

Accept the optional ExceptionOptions argument on WebAssembly.Exception.
When traceStack is true, capture a stack and store it under a private
name; expose it via a prototype getter (IDL attribute). Default remains
no stack. Set the constructor length to 2 to match the IDL.

Keep Error on CommonIdentifiers::stack. Register stackPrivateName as a
private-only special case (like polyProto) so Exception storage does not
introduce a second public "stack" Identifier.

* Source/JavaScriptCore/wasm/js/WebAssemblyExceptionConstructor.cpp:
* Source/JavaScriptCore/wasm/js/WebAssemblyExceptionPrototype.cpp:
* Source/JavaScriptCore/builtins/BuiltinNames.h:
* Source/JavaScriptCore/builtins/BuiltinNames.cpp:
* JSTests/wasm/stress/exception-trace-stack.js: Added.
* LayoutTests/imported/w3c/web-platform-tests/wasm/jsapi/exception/constructor.tentative.any-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/wasm/jsapi/exception/constructor.tentative.any.worker-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/wasm/jsapi/idlharness.any-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/wasm/jsapi/idlharness.any.worker-expected.txt:

Canonical link: https://commits.webkit.org/318397@main
https://bugs.webkit.org/show_bug.cgi?id=320772
rdar://183774945

Reviewed by Aditya Keerthi.

Expose SPI to define the content environment of a page, a more flexible version of the existing `_editable` WKWebView SPI.

Also refactor WebPage to have a dedicated SPI file.

* Source/WebKit/UIProcess/API/Swift/WebPage+SPI.swift: Added.
(WebPage.setMenuBuilder(_:)):
(WebPage.smartListsEnabled):
(terminateWebContentProcess):
(editorStateSnapshots):
(contentEnvironment):
* Source/WebKit/UIProcess/API/Swift/WebPage.swift:
(editorStateSnapshotsContinuations):
(setMenuBuilder(_:)): Deleted.
(smartListsEnabled): Deleted.
(terminateWebContentProcess): Deleted.
(editorStateSnapshots): Deleted.
* Source/WebKit/WebKit.xcodeproj/project.pbxproj:

Canonical link: https://commits.webkit.org/318398@main
https://bugs.webkit.org/show_bug.cgi?id=320820
rdar://183831879

Reviewed by Justin Michaud.

StringRecursionChecker behavior is not specified in the spec, and rather
causing many random weird behavior. For example, array.join should
finish even if we are using a recursion if we have some getter which
breaks the cycle in the middle. But it does not happen and showing
incorrect behavior.

Let's just focus on stack overflow instead of adding StringRecursionChecker
checks which is incorrect in terms of the spec. This patch removes
StringRecursionChecker. Most of functions are fine as it is
automatically doing error checks through JS -> C++ calls. But some are
not. In those places, we insert stack overflow check and throwing an
error when it is happening.

Also WPT is having a wrong test which relies on non-specified behavior.
This patch fixes it, and upstreaming a change[1].

[1]: web-platform-tests/wpt#61678

Test: JSTests/stress/string-conversion-recursion.js

* JSTests/stress/regress-191731.js:
* JSTests/stress/string-conversion-recursion.js: Added.
(shouldBe):
(shouldThrowRangeError):
(shouldThrowRangeError.get return):
(array.0.toString):
(shouldBe.array.join):
(new.Error):
* JSTests/wasm/v8/regress/regress-769846.js:
(catch):
* LayoutTests/imported/w3c/IndexedDB-private-browsing/keypath-expected.txt:
* LayoutTests/imported/w3c/IndexedDB-private-browsing/keypath.html:
* LayoutTests/imported/w3c/web-platform-tests/IndexedDB/keypath.any-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/IndexedDB/keypath.any.js:
(indexeddb_test):
* LayoutTests/imported/w3c/web-platform-tests/IndexedDB/keypath.any.serviceworker-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/IndexedDB/keypath.any.sharedworker-expected.txt:
* LayoutTests/imported/w3c/web-platform-tests/IndexedDB/keypath.any.worker-expected.txt:
* LayoutTests/js/array-string-recursion-expected.txt:
* LayoutTests/js/array-tostring-and-join-expected.txt:
* LayoutTests/js/dom/cyclic-ref-toString-expected.txt:
* LayoutTests/js/dom/script-tests/cyclic-ref-toString.js:
* LayoutTests/js/kde/crash-1-expected.txt:
* LayoutTests/js/script-tests/array-string-recursion.js:
(catch):
* LayoutTests/js/script-tests/array-tostring-and-join.js:
(Object.prototype.toString):
(plain.toString):
(obj.__proto__.toString): Deleted.
* LayoutTests/js/script-tests/toString-recursion.js:
* LayoutTests/js/toString-recursion-expected.txt:
* Source/JavaScriptCore/JavaScriptCore.xcodeproj/project.pbxproj:
* Source/JavaScriptCore/Sources.txt:
* Source/JavaScriptCore/runtime/ArrayPrototype.cpp:
(JSC::JSC_DEFINE_HOST_FUNCTION):
* Source/JavaScriptCore/runtime/ArrayPrototypeInlines.h:
* Source/JavaScriptCore/runtime/ErrorPrototype.cpp:
(JSC::JSC_DEFINE_HOST_FUNCTION):
* Source/JavaScriptCore/runtime/JSArray.cpp:
(JSC::JSArray::fastToString):
* Source/JavaScriptCore/runtime/RegExpPrototype.cpp:
(JSC::JSC_DEFINE_HOST_FUNCTION):
* Source/JavaScriptCore/runtime/StringRecursionChecker.cpp: Removed.
* Source/JavaScriptCore/runtime/StringRecursionChecker.h: Removed.
* Source/JavaScriptCore/runtime/VM.h:

Canonical link: https://commits.webkit.org/318399@main
…DoNotInterceptHitTesting is a flaky failure on macOS Sequoia bots

https://bugs.webkit.org/show_bug.cgi?id=320732
rdar://183717723

Reviewed by Pascoe.

Failing results suggest that `barCenter` was computed before the HUD bar
subview was not fully laid out, which produced a bad hit test. This
patch attempts to make the test more robust by waiting for not only >=2
HUDs, but also for non-empty bar frames.

* Tools/TestWebKitAPI/Tests/WebKit/WKWebView/UnifiedPDFTests.mm:
(TestWebKitAPI::UNIFIED_PDF_TEST):

Canonical link: https://commits.webkit.org/318400@main
* LayoutTests/platform/glib/TestExpectations:
* LayoutTests/platform/gtk/TestExpectations:
* LayoutTests/platform/wpe/TestExpectations:

Canonical link: https://commits.webkit.org/318401@main
…mory bandwidth

https://bugs.webkit.org/show_bug.cgi?id=320745

Reviewed by Carlos Garcia Campos and Nikolas Zimmermann.

m_contentsOpaque only describes what the layer paints into its own tiles, so a
layer whose contents come from an opaque buffer like a WebGL canvas created with
alpha:false, an opaque video frame or an opaque image, are composited with
source-over and read the destination for every pixel adding read bandwidth usage.
TextureMapper gets the same decision off the buffer flags.

Consider those flags too, restricted to layers with no backing store of their
own, since the flag says nothing about the opacity of the layer's own tiles.

* Source/WebCore/platform/graphics/skia/SkiaCompositingLayer.cpp:
(WebCore::SkiaCompositingLayer::paintContents):

Canonical link: https://commits.webkit.org/318402@main
https://bugs.webkit.org/show_bug.cgi?id=320700

Reviewed by Mike Wyrzykowski.

* Source/JavaScriptCore/inspector/protocol/Canvas.json:
* Source/WebCore/inspector/agents/InspectorCanvasAgent.h:
(WebCore::InspectorCanvasAgent::setShaderProgramDisabled):
(WebCore::InspectorCanvasAgent::isWebGPURenderPipelineDisabled): Added.
* Source/WebCore/inspector/agents/InspectorCanvasAgent.cpp:
(WebCore::InspectorCanvasAgent::setShaderProgramDisabled):
(WebCore::InspectorCanvasAgent::isWebGPURenderPipelineDisabled): Added.
* Source/WebCore/inspector/InspectorShaderProgram.h:
* Source/WebCore/inspector/InspectorShaderProgram.cpp:
(WebCore::InspectorShaderProgram::setDisabled):
* Source/WebCore/inspector/InspectorInstrumentation.h:
(WebCore::InspectorInstrumentation::isWebGPURenderPipelineDisabled): Added.
(WebCore::InspectorInstrumentation::isWebGPURenderPipelineDisabledImpl): Added.
* Source/WebCore/inspector/InspectorInstrumentation.cpp:
(WebCore::InspectorInstrumentation::isWebGPURenderPipelineDisabledImpl): Added.
Expose disabling for `GPURenderPipeline` and throw when attempting the same for `GPUComputePipeline`.

* Source/WebCore/Modules/WebGPU/GPURenderPassEncoder.h:
(WebCore::GPURenderPassEncoder):
* Source/WebCore/Modules/WebGPU/GPURenderPassEncoder.cpp:
(WebCore::GPURenderPassEncoder::setPipeline):
(WebCore::GPURenderPassEncoder::draw):
(WebCore::GPURenderPassEncoder::drawIndexed):
(WebCore::GPURenderPassEncoder::drawIndirect):
(WebCore::GPURenderPassEncoder::drawIndexedIndirect):
(WebCore::GPURenderPassEncoder::executeBundles):
(WebCore::GPURenderPassEncoder::end):
* Source/WebCore/Modules/WebGPU/GPURenderBundleEncoder.h:
(WebCore::GPURenderBundleEncoder):
* Source/WebCore/Modules/WebGPU/GPURenderBundleEncoder.cpp:
(WebCore::GPURenderBundleEncoder::setPipeline):
(WebCore::GPURenderBundleEncoder::draw):
(WebCore::GPURenderBundleEncoder::drawIndexed):
(WebCore::GPURenderBundleEncoder::drawIndirect):
(WebCore::GPURenderBundleEncoder::drawIndexedIndirect):
(WebCore::GPURenderBundleEncoder::finish):
Track the current `GPURenderPipeline` and bail if inspector says it's disabled.

* Source/WebInspectorUI/UserInterface/Models/ShaderProgram.js:
(WI.ShaderProgram.prototype.get supportsDisabling): Added.
(WI.ShaderProgram.prototype.set disabled):
* Source/WebInspectorUI/UserInterface/Views/ShaderProgramTreeElement.js:
(WI.ShaderProgramTreeElement.prototype.constructor):
(WI.ShaderProgramTreeElement.prototype.onattach):
(WI.ShaderProgramTreeElement.prototype.ondetach):
(WI.ShaderProgramTreeElement.prototype.populateContextMenu):
Also allow disabling for ` WI.Canvas.ContextType.WebGPU` (but not `WI.ShaderProgram.ProgramType.Compute`).

* LayoutTests/inspector/canvas/resources/shaderProgram-utilities-webgpu.js:
* LayoutTests/inspector/canvas/setShaderProgramDisabled-webgpu.html: Added.
* LayoutTests/inspector/canvas/setShaderProgramDisabled-webgpu-expected.txt: Added.
* LayoutTests/inspector/canvas/shaderProgram-add-remove-webgpu.html:
* LayoutTests/inspector/canvas/shaderProgram-add-remove-webgpu-expected.txt:
* LayoutTests/platform/glib/TestExpectations:
* LayoutTests/platform/win/TestExpectations:

Canonical link: https://commits.webkit.org/318403@main
https://bugs.webkit.org/show_bug.cgi?id=320784

Reviewed by Adrian Perez de Castro and Carlos Garcia Campos.

This was using the wrong flags to match types. Update docs
and add tests for more types.

Test: Tools/TestWebKitAPI/Tests/WebKit/WKWebView/glib/TestWebsiteData.cpp

* Source/WebKit/UIProcess/API/glib/WebKitWebsiteData.cpp:
(webkit_website_data_get_size):
* Tools/TestWebKitAPI/Tests/WebKit/WKWebView/glib/TestWebsiteData.cpp:
(testWebsiteDataStorage):
(testWebsiteDataDatabases):
(testWebsiteDataDOMCache):
(testWebsiteDataSizes):
(beforeAll):

Canonical link: https://commits.webkit.org/318404@main
https://bugs.webkit.org/show_bug.cgi?id=320809

Reviewed by Carlos Garcia Campos.

This has been enabled on Apple ports for a while.
After the fixes in https://commits.webkit.org/318384@main
it just works.

* LayoutTests/platform/glib/TestExpectations:
* Source/WTF/Scripts/Preferences/UnifiedWebPreferences.yaml:

Canonical link: https://commits.webkit.org/318405@main
https://bugs.webkit.org/show_bug.cgi?id=320724
rdar://182292429

Reviewed by Etienne Segonzac.

Let a spatial portal host more than one <model> child. SpatialPortalController
tracks its children in a map keyed by NodeIdentifier, loading each one and
routing callbacks back to the right child. The Model process keeps one entity
per node, and a new UnloadModel message removes one without tearing down the
player the other models share.

The container transform fits the union of every hosted model's bounding box.
The first model to load still owns stage mode, animation reporting and the
environment map, so the single-model case is unchanged.

Tests: spatial-css/spatial-portal-multi-model-error.html
       spatial-css/spatial-portal-multi-model-insert-remove.html
       spatial-css/spatial-portal-multi-model-move.html
       spatial-css/spatial-portal-multi-model-source-change.html
       spatial-css/spatial-portal-multi-model.html

* LayoutTests/spatial-css/resources/spatial-portal-utils.js: Added.
(const.appendModel):
* LayoutTests/spatial-css/spatial-portal-multi-model-error-expected.txt: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-error.html: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-expected.txt: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-insert-remove-expected.txt: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-insert-remove.html: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-move-expected.txt: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-move.html: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-source-change-expected.txt: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model-source-change.html: Added.
* LayoutTests/spatial-css/spatial-portal-multi-model.html: Added.
* Source/WebCore/Modules/model-element/HTMLModelElement.cpp:
(WebCore::HTMLModelElement::setSourceURL):
* Source/WebCore/Modules/model-element/SpatialPortalController.cpp:
(WebCore::SpatialPortalController::~SpatialPortalController):
(WebCore::SpatialPortalController::hostedModelElement const):
(WebCore::SpatialPortalController::unregisterChildModel):
(WebCore::SpatialPortalController::registerChildModel):
(WebCore::SpatialPortalController::childModelDidChange):
(WebCore::SpatialPortalController::loadChildModelsIfReady):
(WebCore::SpatialPortalController::loadChildModelIfReady):
(WebCore::SpatialPortalController::viewportIntersectionChanged):
(WebCore::SpatialPortalController::deleteModelPlayer):
(WebCore::SpatialPortalController::unloadChildModel):
(WebCore::SpatialPortalController::configureGraphicsLayer):
(WebCore::SpatialPortalController::modelDidFinishLoading):
(WebCore::SpatialPortalController::modelDidFailLoading):
(WebCore::SpatialPortalController::modelDidUnload):
(WebCore::SpatialPortalController::unloadModel): Deleted.
* Source/WebCore/Modules/model-element/SpatialPortalController.h:
(WebCore::SpatialPortalController::numberOfHostedModels const):
* Source/WebKit/ModelProcess/cocoa/ModelProcessModelPlayerProxy.h:
* Source/WebKit/ModelProcess/cocoa/ModelProcessModelPlayerProxy.messages.in:
* Source/WebKit/ModelProcess/cocoa/ModelProcessModelPlayerProxy.mm:
(WebKit::ModelProcessModelPlayerProxy::~ModelProcessModelPlayerProxy):
(WebKit::ModelProcessModelPlayerProxy::loadModel):
(WebKit::ModelProcessModelPlayerProxy::reloadModel):
(WebKit::ModelProcessModelPlayerProxy::unloadModelTimerFired):
(WebKit::ModelProcessModelPlayerProxy::computeTransform):
(WebKit::ModelProcessModelPlayerProxy::updateTransform):
(WebKit::ModelProcessModelPlayerProxy::updateOpacity):
(WebKit::ModelProcessModelPlayerProxy::didFinishLoading):
(WebKit::ModelProcessModelPlayerProxy::didFailLoading):
(WebKit::ModelProcessModelPlayerProxy::load):
(WebKit::ModelProcessModelPlayerProxy::unloadModel):
(WebKit::ModelProcessModelPlayerProxy::clearReportingModelIfNeeded):
(WebKit::ModelProcessModelPlayerProxy::reportingModelScale const):
(WebKit::ModelProcessModelPlayerProxy::findHostedEntityForLoader):
(WebKit::ModelProcessModelPlayerProxy::cancelAllLoaders):
(WebKit::ModelProcessModelPlayerProxy::setUpLoadedEntity):
(WebKit::ModelProcessModelPlayerProxy::sizeDidChange):
(WebKit::ModelProcessModelPlayerProxy::updateForCurrentStageMode):
(WebKit::ModelProcessModelPlayerProxy::updateTransformSRT):
(WebKit::ModelProcessModelPlayerProxy::teardownEntity):
(WebKit::ModelProcessModelPlayerProxy::ensureImmersivePresentation):
(WebKit::ModelProcessModelPlayerProxy::exitImmersivePresentation):
(WebKit::ModelProcessModelPlayerProxy::startAnimating): Deleted.
* Source/WebKit/WebKitSwift/RealityKit/WKRKEntity.h:
* Source/WebKit/WebKitSwift/RealityKit/WKRKEntity.swift:
(WKRKEntity.removeFromParentEntity):
* Source/WebKit/WebProcess/Model/ModelProcessModelPlayer.cpp:
(WebKit::ModelProcessModelPlayer::unload):
* Source/WebKit/WebProcess/Model/ModelProcessModelPlayer.h:
* Tools/TestWebKitAPI/Resources/spatial-portal-two-models-page.html:

Canonical link: https://commits.webkit.org/318406@main
https://bugs.webkit.org/show_bug.cgi?id=201651

Reviewed by Mike Wyrzykowski.

* Source/WebCore/Modules/WebGPU/GPUDevice.idl:
* Source/WebCore/Modules/WebGPU/GPUDevice.h:
(WebCore::GPUDevice::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUDevice::setHasActiveInspectorCanvasCallTracer): Added.
* Source/WebCore/Modules/WebGPU/GPUDevice.cpp:
(WebCore::GPUDevice::GPUDevice):
(WebCore::GPUDevice::createSampler):
(WebCore::GPUDevice::importExternalTexture):
(WebCore::GPUDevice::createBindGroupLayout):
(WebCore::GPUDevice::createAutoPipelineLayout):
(WebCore::GPUDevice::createPipelineLayout):
(WebCore::GPUDevice::createBindGroup):
(WebCore::GPUDevice::createShaderModule):
(WebCore::GPUDevice::createCommandEncoder):
(WebCore::GPUDevice::createRenderBundleEncoder):
(WebCore::GPUDevice::createQuerySet):
* Source/WebCore/Modules/WebGPU/GPUQueue.idl:
* Source/WebCore/Modules/WebGPU/GPUQueue.h:
(WebCore::GPUQueue::create):
* Source/WebCore/Modules/WebGPU/GPUQueue.cpp:
(WebCore::GPUQueue::GPUQueue):
(WebCore::GPUQueue::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUQueue::device const): Added.
(WebCore::GPUQueue::submit):
* Source/WebCore/Modules/WebGPU/GPUBuffer.idl:
* Source/WebCore/Modules/WebGPU/GPUBuffer.h:
* Source/WebCore/Modules/WebGPU/GPUBuffer.cpp:
(WebCore::GPUBuffer::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUBuffer::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUTexture.idl:
* Source/WebCore/Modules/WebGPU/GPUTexture.h:
(WebCore::GPUTexture::create):
* Source/WebCore/Modules/WebGPU/GPUTexture.cpp:
(WebCore::GPUTexture::GPUTexture):
(WebCore::GPUTexture::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUTexture::device const): Added.
(WebCore::GPUTexture::createView):
* Source/WebCore/Modules/WebGPU/GPUTextureView.idl:
* Source/WebCore/Modules/WebGPU/GPUTextureView.h:
(WebCore::GPUTextureView::create):
* Source/WebCore/Modules/WebGPU/GPUTextureView.cpp:
(WebCore::GPUTextureView::GPUTextureView): Added.
(WebCore::GPUTextureView::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUTextureView::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUExternalTexture.idl:
* Source/WebCore/Modules/WebGPU/GPUExternalTexture.h:
(WebCore::GPUExternalTexture::create):
* Source/WebCore/Modules/WebGPU/GPUExternalTexture.cpp:
(WebCore::GPUExternalTexture::GPUExternalTexture): Added.
(WebCore::GPUExternalTexture::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUExternalTexture::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUSampler.idl:
* Source/WebCore/Modules/WebGPU/GPUSampler.h:
(WebCore::GPUSampler::create):
* Source/WebCore/Modules/WebGPU/GPUSampler.cpp:
(WebCore::GPUSampler::GPUSampler): Added.
(WebCore::GPUSampler::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUSampler::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUBindGroupLayout.idl:
* Source/WebCore/Modules/WebGPU/GPUBindGroupLayout.h:
(WebCore::GPUBindGroupLayout::create):
* Source/WebCore/Modules/WebGPU/GPUBindGroupLayout.cpp:
(WebCore::GPUBindGroupLayout::GPUBindGroupLayout): Added.
(WebCore::GPUBindGroupLayout::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUBindGroupLayout::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUBindGroup.idl:
* Source/WebCore/Modules/WebGPU/GPUBindGroup.h:
(WebCore::GPUBindGroup::create):
* Source/WebCore/Modules/WebGPU/GPUBindGroup.cpp:
(WebCore::GPUBindGroup::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUBindGroup::device const): Added.
(WebCore::GPUBindGroup::GPUBindGroup):
* Source/WebCore/Modules/WebGPU/GPUPipelineLayout.idl:
* Source/WebCore/Modules/WebGPU/GPUPipelineLayout.h:
(WebCore::GPUPipelineLayout::create):
* Source/WebCore/Modules/WebGPU/GPUPipelineLayout.cpp:
(WebCore::GPUPipelineLayout::GPUPipelineLayout): Added.
(WebCore::GPUPipelineLayout::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUPipelineLayout::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUShaderModule.idl:
* Source/WebCore/Modules/WebGPU/GPUShaderModule.h:
(WebCore::GPUShaderModule::create):
* Source/WebCore/Modules/WebGPU/GPUShaderModule.cpp:
(WebCore::GPUShaderModule::GPUShaderModule): Added.
(WebCore::GPUShaderModule::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUShaderModule::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUComputePipeline.idl:
* Source/WebCore/Modules/WebGPU/GPUComputePipeline.h:
* Source/WebCore/Modules/WebGPU/GPUComputePipeline.cpp:
(WebCore::GPUComputePipeline::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUComputePipeline::getBindGroupLayout):
* Source/WebCore/Modules/WebGPU/GPURenderPipeline.idl:
* Source/WebCore/Modules/WebGPU/GPURenderPipeline.h:
* Source/WebCore/Modules/WebGPU/GPURenderPipeline.cpp:
(WebCore::GPURenderPipeline::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPURenderPipeline::getBindGroupLayout):
* Source/WebCore/Modules/WebGPU/GPUCommandEncoder.idl:
* Source/WebCore/Modules/WebGPU/GPUCommandEncoder.h:
(WebCore::GPUCommandEncoder::create):
* Source/WebCore/Modules/WebGPU/GPUCommandEncoder.cpp:
(WebCore::GPUCommandEncoder::GPUCommandEncoder):
(WebCore::GPUCommandEncoder::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUCommandEncoder::device const): Added.
(WebCore::GPUCommandEncoder::beginRenderPass):
(WebCore::GPUCommandEncoder::beginComputePass):
(WebCore::GPUCommandEncoder::finish):
* Source/WebCore/Modules/WebGPU/GPUComputePassEncoder.idl:
* Source/WebCore/Modules/WebGPU/GPUComputePassEncoder.h:
(WebCore::GPUComputePassEncoder::create):
* Source/WebCore/Modules/WebGPU/GPUComputePassEncoder.cpp:
(WebCore::GPUComputePassEncoder::GPUComputePassEncoder):
(WebCore::GPUComputePassEncoder::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUComputePassEncoder::device const): Added.
(WebCore::GPUComputePassEncoder::end):
* Source/WebCore/Modules/WebGPU/GPURenderPassEncoder.idl:
* Source/WebCore/Modules/WebGPU/GPURenderPassEncoder.h:
(WebCore::GPURenderPassEncoder::create):
* Source/WebCore/Modules/WebGPU/GPURenderPassEncoder.cpp:
(WebCore::GPURenderPassEncoder::GPURenderPassEncoder):
(WebCore::GPURenderPassEncoder::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPURenderPassEncoder::device const): Added.
(WebCore::GPURenderPassEncoder::end):
* Source/WebCore/Modules/WebGPU/GPURenderBundleEncoder.idl:
* Source/WebCore/Modules/WebGPU/GPURenderBundleEncoder.h:
(WebCore::GPURenderBundleEncoder::create):
* Source/WebCore/Modules/WebGPU/GPURenderBundleEncoder.cpp:
(WebCore::GPURenderBundleEncoder::GPURenderBundleEncoder): Added.
(WebCore::GPURenderBundleEncoder::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPURenderBundleEncoder::device const): Added.
(WebCore::GPURenderBundleEncoder::finish):
* Source/WebCore/Modules/WebGPU/GPURenderBundle.idl:
* Source/WebCore/Modules/WebGPU/GPURenderBundle.h:
(WebCore::GPURenderBundle::create):
* Source/WebCore/Modules/WebGPU/GPURenderBundle.cpp:
(WebCore::GPURenderBundle::GPURenderBundle): Added.
(WebCore::GPURenderBundle::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPURenderBundle::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUCommandBuffer.idl:
* Source/WebCore/Modules/WebGPU/GPUCommandBuffer.h:
* Source/WebCore/Modules/WebGPU/GPUCommandBuffer.cpp:
(WebCore::GPUCommandBuffer::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUCommandBuffer::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUQuerySet.idl:
* Source/WebCore/Modules/WebGPU/GPUQuerySet.h:
(WebCore::GPUQuerySet::create):
* Source/WebCore/Modules/WebGPU/GPUQuerySet.cpp:
(WebCore::GPUQuerySet::GPUQuerySet):
(WebCore::GPUQuerySet::hasActiveInspectorCanvasCallTracer const): Added.
(WebCore::GPUQuerySet::device const): Added.
* Source/WebCore/Modules/WebGPU/GPUPresentationContext.h:
* Source/WebCore/inspector/RecordingSwizzleType.h:
* Source/WebCore/inspector/InspectorCanvasCallTracer.h:
(WebCore::InspectorCanvasCallTracer::processArgument):
(WebCore::InspectorCanvasCallTracer::recordAction):
(WebCore::InspectorCanvasCallTracer::receiverSwizzleType): Added.
(WebCore::InspectorCanvasCallTracer::receiverSwizzleTypeImpl): Added.
* Source/WebCore/inspector/InspectorCanvasCallTracer.cpp:
(WebCore::enabledCanvasAgent):
(WebCore::InspectorCanvasCallTracer::enabledInspectorCanvas):
(WebCore::InspectorCanvasCallTracer::recordAction):
Add `CallTrace` to every WebGPU IDL (including those without any methods today so future API additions automatically get support).
Each object now has a link to the owner `GPUDevice` so the `CallTracer` can check whether recording is active.

* Source/JavaScriptCore/inspector/protocol/Recording.json:
* Source/JavaScriptCore/inspector/scripts/codegen/generator.py:
* Source/WebCore/inspector/InspectorCanvas.h:
* Source/WebCore/inspector/InspectorCanvas.cpp:
(WebCore::InspectorCanvas::hasActiveInspectorCanvasCallTracer): Added.
(WebCore::InspectorCanvas::setHasActiveInspectorCanvasCallTracer): Added.
(WebCore::InspectorCanvas::resetRecordingData):
(WebCore::shouldSnapshotWebGPUAction): Added.
(WebCore::InspectorCanvas::recordAction):
(WebCore::buildActionReceiver): Added.
(WebCore::InspectorCanvas::releaseObjectForRecording):
(WebCore::InspectorCanvas::getContentAsDataURL): Added WebGPU support for device-backed canvases.
(WebCore::InspectorCanvas::appendActionSnapshotIfNeeded):
(WebCore::InspectorCanvas::identifierForRecordingObject): Added.
(WebCore::InspectorCanvas::buildInitialState):
Store explicit receivers as `[identifier, swizzleType]` before the optional snapshot field.

* Source/WebCore/inspector/agents/InspectorCanvasAgent.h:
* Source/WebCore/inspector/agents/InspectorCanvasAgent.cpp:
(WebCore::InspectorCanvasAgent::startRecording):
(WebCore::InspectorCanvasAgent::stopRecording):
(WebCore::InspectorCanvasAgent::didFinishRecordingCanvasFrame):
(WebCore::InspectorCanvasAgent::didCreateWebGPUDevice):
(WebCore::InspectorCanvasAgent::recordAction):
(WebCore::InspectorCanvasAgent::scheduleRecordingCanvasFrame): Added.
(WebCore::InspectorCanvasAgent::consoleStartRecordingCanvas):
(WebCore::InspectorCanvasAgent::consoleStopRecordingCanvas):
(WebCore::InspectorCanvasAgent::reset):
(WebCore::InspectorCanvasAgent::unbindCanvas):
Support explicit and automatic `GPUDevice` recordings.

* Source/WebCore/page/FrameConsoleClient.cpp:
(WebCore::FrameConsoleClient::record):
(WebCore::FrameConsoleClient::recordEnd):
* Source/WebCore/workers/WorkerConsoleClient.cpp:
(WebCore::WorkerConsoleClient::record):
(WebCore::WorkerConsoleClient::recordEnd):
* Source/WebCore/inspector/InspectorInstrumentation.h:
(WebCore::InspectorInstrumentation::consoleStartRecordingCanvas):
(WebCore::InspectorInstrumentation::consoleStopRecordingCanvas):
* Source/WebCore/inspector/InspectorInstrumentation.cpp:
(WebCore::InspectorInstrumentation::consoleStartRecordingCanvasImpl):
(WebCore::InspectorInstrumentation::consoleStopRecordingCanvasImpl):
Support `console.record()` and `console.recordEnd()` for `GPUDevice`.

* Source/WebInspectorUI/UserInterface/Models/Canvas.js:
(WI.Canvas.prototype.get supportsRecording): Deleted.
(WI.Canvas.prototype.recordingProgress):
* Source/WebInspectorUI/UserInterface/Views/CanvasContentView.js:
(WI.CanvasContentView.prototype.initialLayout):
* Source/WebInspectorUI/UserInterface/Views/CanvasSidebarPanel.js:
(WI.CanvasSidebarPanel.prototype._canvasChanged):
(WI.CanvasSidebarPanel.prototype._updateRecordNavigationItem):
Pass the backend recording version when parsing a live recording.

* Source/WebInspectorUI/UserInterface/Models/Recording.js:
(WI.Recording.fromPayload):
(WI.Recording.displayNameForRecordingType):
(WI.Recording.displayNameForReceiver): Added.
(WI.Recording.prototype.get isCanvas):
(WI.Recording.prototype.get isCanvasWebGPU): Added.
(WI.Recording.prototype.createContext):
(WI.Recording.prototype.toJSON):
(WI.Recording.prototype.toHTML):
(WI.Recording.prototype._process):
* Source/WebInspectorUI/UserInterface/Models/RecordingAction.js:
(WI.RecordingAction):
(WI.RecordingAction.fromPayload):
(WI.RecordingAction.isFunctionForType):
(WI.RecordingAction.bitfieldNamesForParameter):
(WI.RecordingAction.constantNameForParameter):
(WI.RecordingAction._prototypeForType):
(WI.RecordingAction.prototype.get receiver): Added.
(WI.RecordingAction.prototype.get contextReplacer): Deleted.
(WI.RecordingAction.prototype.process):
(WI.RecordingAction.prototype.swizzle):
(WI.RecordingAction.prototype.get isCanvasReceiver): Added.
(WI.RecordingAction.prototype.apply):
(WI.RecordingAction.prototype.toJSON):
* Source/WebInspectorUI/UserInterface/Models/RecordingFrame.js:
(WI.RecordingFrame.fromPayload):
(WI.RecordingFrame.prototype.toJSON):
Parse snapshot-based WebGPU recordings without attempting to replay WebGPU commands.
Display receiver identifier and type pairs as lower-camel-cased names.
Drive-by: Include offscreen WebGL(2) recordings when resolving constants, bitfields, and snapshot-based processing.

* Source/WebInspectorUI/UserInterface/Views/RecordingActionTreeElement.js:
(WI.RecordingActionTreeElement._generateDOM):
(WI.RecordingActionTreeElement._classNameForAction):
(WI.RecordingActionTreeElement.prototype.populateContextMenu):
* Source/WebInspectorUI/UserInterface/Views/RecordingActionTreeElement.css:
(body:not(.window-inactive, .window-docked-inactive) .tree-outline:focus-within .item.recording-action.selected:not(.invalid, .initial-state, .name-unknown) > .icon): Renamed from `body:not(.window-inactive, .window-docked-inactive) .tree-outline:focus-within .item.recording-action.selected:not(.invalid, .initial-state, .has-context-replacer, .name-unknown) > .icon`.
(.item.recording-action:not(.invalid, .initial-state, .name-unknown) > .icon): Renamed from `.item.recording-action:not(.invalid, .initial-state, .has-context-replacer, .name-unknown) > .icon`.
Display only explicit receivers and copy actions against the correct `GPUDevice`, `CanvasRenderingContext`, or `<canvas>`.

* Source/WebInspectorUI/UserInterface/Views/RecordingContentView.js:
(WI.RecordingContentView):
(WI.RecordingContentView.prototype._generateContentFromSnapshot):
(WI.RecordingContentView.prototype._updateImageGrid):
Show WebGPU content from snapshots and clear stale preview content when a recording has no initial snapshot.

* Source/WebInspectorUI/UserInterface/Views/TreeOutline.js:
(WI.TreeOutline.prototype.registerScrollVirtualizer):
Drive-by: disable native scroll anchoring when virtualized so recycled recording rows do not cause scrolling to continue on its own.

* Source/WebInspectorUI/Localizations/en.lproj/localizedStrings.js:

* LayoutTests/inspector/canvas/recording.html:
* LayoutTests/inspector/canvas/recording-expected.txt:
* LayoutTests/inspector/canvas/recording-webgpu.html: Added.
* LayoutTests/inspector/canvas/recording-webgpu-expected.txt: Added.
* LayoutTests/platform/glib/TestExpectations:
* LayoutTests/platform/win/TestExpectations:

Canonical link: https://commits.webkit.org/318407@main
…eb3c)

- Drop USE(JSVALUE64) guards from ffi/ and JSDollarVM.cpp: the macro is no
  longer defined so these blocks were compiled out entirely.
- FFIDFG.cpp: remove dead is64Bit() check.
- FFIICStub.cpp: payloadFor -> lowWordFor.
- InspectorDebuggerAgent.cpp: handle BunTranspiledModule/Synthetic in the new
  scriptTypeForScript switch.
- CMakeLists.txt: drop StringRecursionChecker.h from the private headers list
  (removed upstream in f2f2c2d).
@github-actions

github-actions Bot commented Aug 2, 2026

Copy link
Copy Markdown

Preview Builds

Commit Release Date
b9ea4dc5 autobuild-preview-pr-383-b9ea4dc5 2026-08-02 22:33:32 UTC

@claude claude Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No issues found by automated review, but this is a large upstream merge (300 files, 110 JSC/WTF/bmalloc commits) with 17 hand-resolved conflicts and fork-specific fixups for the 32-bit JSValue removal — worth a human pass over the conflict resolutions and ffi/* / InspectorDebuggerAgent adaptations.

Checked: the PRUNE_STALE header-staging block in Source/JavaScriptCore/CMakeLists.txt — still present after the merge, not dropped.
Checked: StringRecursionChecker.h removal from the private headers list matches upstream f2f2c2ddf637.
Bulk of the diff is upstream JSTests / wasm spec-test regeneration and deletions, which carry no fork-specific risk.

Extended reasoning...

Overview

This PR merges upstream WebKit 3722912ff800 into the Bun fork, spanning ~300 files and ~1M characters of diff. The vast majority is upstream-authored: new/regenerated JSTests (stress, microbenchmarks, wasm spec tests), deletion of superseded wasm spec-test harnesses, and JSC/WTF/bmalloc engine changes from 110 upstream commits. The fork-authored portion is comparatively small but high-leverage: 17 manual conflict resolutions preserving USE(BUN_JSC_ADDITIONS) hunks, and follow-up fixups adapting fork-only code (ffi/*, tools/JSDollarVM.cpp, inspector/agents/InspectorDebuggerAgent.cpp, CMakeLists.txt) to upstream's removal of USE_JSVALUE64/USE_JSVALUE32_64, is64Bit(), and payloadFor.

Security risks

None introduced by the merge mechanics themselves. The upstream content includes changes to Yarr, JIT tiers, BigInt arithmetic, and the module loader — all security-relevant surfaces — but those changes are upstream-reviewed WebKit commits, not authored here. The fork-specific fixups are mechanical (guard removal, payloadFor → lowWordFor, adding switch cases) and do not introduce new attack surface.

Level of scrutiny

High. Even though the bulk is upstream code, the 17 hand-resolved conflicts in JSC/WTF and the 32-bit-removal adaptations in fork-only FFI code are exactly where a merge can silently drop a Bun-specific guard or mis-adapt an ABI-sensitive helper. The PR description is thorough and the author verified a debug build + bun run build:local, but conflict-resolution correctness in a JIT/runtime is not something automated review can fully validate.

Other factors

  • The bug-hunting system raised and then ruled out one candidate: that the fork's PRUNE_STALE header-staging feature in CMakeLists.txt was dropped by the merge. Verified it is still present.
  • No prior reviews or comments on the PR.
  • A companion Bun-side PR is required for AbstractModuleRecord / ScriptFetchParameters::Type changes; this PR is not standalone-landable without it.
  • Given the scope, complexity, and criticality (JS engine core), deferring to human review rather than approving.

@sosukesuzuki
sosukesuzuki merged commit e6e37cd into main Aug 3, 2026
43 checks passed
robobun added a commit to oven-sh/bun that referenced this pull request Aug 3, 2026
dylan-conway pushed a commit to oven-sh/bun that referenced this pull request Aug 3, 2026
Upgrades the WebKit fork to upstream WebKit/WebKit@3722912ff800
(2026-08-02) via oven-sh/WebKit#383.

oven-sh/WebKit#383 is merged; `WEBKIT_VERSION` points at the merge
commit `e6e37cda216c0292ae68c30c84a9dc8601d0fba5`.

## Bun-side changes

Upstream `90b2ecf79ae3` keys `m_loadedModules` on `(specifier,
ScriptFetchParameters::Type)` and threads the type through
`ImportEntry`, `ExportEntry`, `StarExportEntry`, and
`ModuleAnalyzer::appendRequestedModule`. Under `bun test --isolate` (the
`BunTranspiledModule` path), Bun'''s synthesized record must agree with
what JSC'''s own `ModuleAnalyzer` would produce from the printed source;
a mismatch fails the BUN_DEBUG record diff in debug and null-derefs
`hostResolveImportedModule` in release.

- `analyze_transpiled_module` / `js_printer`: every `RecordKind` now
carries one trailing `FetchParameters` slot. `add_import_info_*` and
`add_export_info_*` accept it; `finalize()` propagates the source
import'''s slot through the Local->Indirect conversion.
`RequestedModules` dedupes on `(specifier, Type, phase)`.
- `analyze_jsc.rs`: decodes the trailing slot to the JSC `Type` enum and
passes it to every `addImportEntry*` / `addIndirectExport` /
`addNamespaceExport` / `addStarExport`; buffer validation is per-slot so
only the trailing slot accepts the wider `FetchParameters` sentinel
range.
- `BunAnalyzeTranspiledModule.cpp`: all seven `addImportEntry*` /
`add*Export` functions take `uint8_t moduleRequestType` and set
`.moduleRequestType` explicitly; `dumpRecordInfo()` prints `type(N)` for
import/export/star entries; `static_assert` pins the ordinal values
`to_script_fetch_parameters_type()` hardcodes.
- `RuntimeTranspilerCache` `EXPECTED_VERSION` -> 25 (esm_record layout
change).

## WebKit-side changes (oven-sh/WebKit#383)

- 17 merge conflicts resolved in JSC/WTF; fork's
`USE(BUN_JSC_ADDITIONS)` hunks preserved.
- Fork `ffi/` code adapted to upstream's 32-bit removal (`USE_JSVALUE64`
macro deleted, `is64Bit()` removed, `payloadFor` -> `lowWordFor`).
- `InspectorDebuggerAgent.cpp`: handle `BunTranspiledModule`/`Synthetic`
in new `scriptTypeForScript` switch.
- Recorded `01aaa3e0be0c` as ancestor via `-s ours` (oven-sh/WebKit#352
was a squash merge).

## How did you verify your code works?

- `bun run jsc:build:debug` builds and the `jsc` shell runs (`-e
'print(42)'` -> `42`).
- `bun run build:local -p '42'` links and runs against the local merged
WebKit.
- oven-sh/WebKit#383 preview build green on all 38 platform variants.
- `bun bd -p 'process.versions.webkit'` -> `preview-pr-383-b9ea4dc5`.
- New test `test/js/bun/jsc/webkit-upgrade-3722912f.test.ts`: 4/4 pass
with `bun bd test`, 3/4 fail with `USE_SYSTEM_BUN=1 bun test` (old JSC
lacks `Iterator.prototype.includes`, returns `""` for cyclic join,
`WebAssembly.Exception.length === 1`).

## JavaScriptCore/WTF/bmalloc changes since 01aaa3e0be0c (Jul 25)

Upstream range: WebKit/WebKit@01aaa3e0be0c...3722912ff800 (474 commits
total, 110 touching JSC/WTF/bmalloc, Jul 25 → Aug 2 2026).

### Highlights

- `29ceb3c03de3` Remove 32-bit JSValues (JSVALUE32_64 and
`CPU(NEEDS_ALIGNED_ACCESS)` deleted).
- `857bd4334690` Remove ARMv7 JIT support (ARMv7 is CLoop-only now;
drops remaining 32-bit/x86 JIT refs).
- `232cebabc1f3` Remove big-endian support and platforms without
unaligned loads/stores (WTF + JSC).
- `6eaa5ac1f65d` Remove 32-bit libpas support.
- `bfb1b1183bc2` Remove the B3/Air graph-coloring register allocator
(greedy is the only allocator now).
- `0d0080ea539d` Enable WebAssembly Memory64 by default (+ Table64/SIMD
follow-ups).
- `f2f2c2ddf637` Remove `StringRecursionChecker`; cyclic
`toString`/`join` now throws RangeError via stack check (spec-correct).
- `319f94b3db4a` Enable `Iterator.prototype.includes()` by default.
- `90b2ecf79ae3` `hostResolveImportedModule` now honors import-attribute
`type` (module loader behavior change).
- `f5716f6401ed` Add `preserve_most` calling convention to fastMalloc
APIs on ARM64 (perf + ABI of WTF alloc entry points).

### JavaScriptCore

**Runtime / builtins**
- `f2f2c2ddf637` Remove StringRecursionChecker; rely on stack-overflow
checks.
- `cd91e7f128dd` Add fast flag for `ToPrimitive(Object)` calls in
runtime.
- `173a0bd6d937` Use `defaultToPrimitiveFastAndNonObservable` in
`JSObject::toString`.
- `960adeccefcd` Fast operation for `Array#shift`.
- `92c6650c7947` Add `JSArrayIterator::next` C++ helper.
- `cb1c48b3e95f` Extend `Array.from()` Set fast path to
`set.keys()/.values()`.
- `4a2e724d6789` Fix: `Array.from(set.keys()/.values())` fast path
ignored `Symbol.iterator` overrides.
- `73e4c589c1e6` Extend `StringSplitCache` to RegExp separators.
- `656d3c36830f` / `1d355c27ea88` 8-byte SWAR fast paths in
`JSON.stringify` string copy (same-type & upconvert).
- `9f9370cc729f` Fix JSON.stringify regression around `toJSON` check.
- `2eb77e9c9473` BigInt: implement Crandall reduction.
- `39b1bb9cfc85` BigInt: deploy Comba multiplication more broadly.
- `319f94b3db4a` Enable `Iterator.prototype.includes()`.
- `0731b27c1b60` `Iterator.zip`: use null-prototype objects for
options/underlying iterator.
- `90b2ecf79ae3` `hostResolveImportedModule` respects module request
import-attribute `type`.
- `4f54300b848a` Collect diagnostics when `getDirect` returns zero
JSValue in `llint_slow_path_get_by_id`.

**Parser / bytecompiler**
- `c2beca7a439f` Lexer: scan integer tokens in a single pass.
- `72ea806faa21` Use overflow-safe range when choosing a switch jump
table.

**LLInt / DFG / FTL / B3**
- `29ceb3c03de3` Remove 32-bit JSValues.
- `857bd4334690` Remove ARMv7 JIT support.
- `bfb1b1183bc2` Remove B3/Air graph-coloring register allocator.
- `68cde6ba2ad3` Make IRO (Air register allocation) faster.
- `83540481435f` DFG: allocate
`BasicBlock::intersectionOfPastValuesAtHead` only for OSR-entry targets.
- `1566615170ec` DFG fix: `EnumeratorNextUpdateIndexAndMode` must
require original array structure for `InBoundsSaneChain`.
- `e759fa9dd063` LLInt: inline hot path of `op_enter`.
- `9610c2113b45` offlineasm: emit ARM64 register-offset addressing for
BaseIndex operands.
- `4ebed2479144` Speed up `addSortedRange` via binary search.
- `1c006b0b0f62` Fix regex `setLastIndex` on 32-bit.

**WebAssembly**
- `0d0080ea539d` Enable Memory64 feature flag.
- `15aa6fad53e3` Memory64: SIMD support.
- `bf0425598904` Memory64: expand declared memory limits.
- `862994e2cc37` Memory64: validate table import address-type match.
- `184ee4c654bd` Memory64: Table64 in OMG tier.
- `d202bedc5ff6` Memory64: Table64 in BBQ tier.
- `bf6512f84f7d` Support `WebAssembly.Exception` `options.traceStack` (+
`stack` getter, ctor length = 2).
- `24527bbb9ac8` Optimize Wasm JITCallee publication (lock splitting,
icache barrier rework).
- `1803d6109d98` Speed up `WebAssembly.Table` construction.
- `d434a41411a3` BBQ: optimize `br_table` for consecutive same-target
runs.
- `51d3dbaa278e` IPInt: add `DEFINE_IPINT_THUNK_FOR_ENTRY`.
- `244cd98f7986` Fix `generateWasmOpsHeader.py` under non-UTF-8 locales.

**Yarr / RegExp**
- `581f1d958329` Start end-anchored fixed-size regexps at the only
possible position.
- `a458a6c1f0a7` v-mode class-set op loop: stop early when no more
output possible.
- `7c5dbbcba110` Extend `ParenthesesSubpatternTerminal`.
- `e1def8f4e5fd` Don't save sibling/ancestor-sibling frame slots for
`ParenContext`.
- `1e43057f135a` Extend first-character filter further.
- `54916608d7d6` Fix non-BMP advance latch; simplify
`tryReadUnicodeCharImpl`.
- `46a4b17efbe9` `optimizeBOL`: don't filter contents of negative
lookaheads.
- `b128ddd863ab` Fix dot-star-wrapped optimization for sticky patterns.
- `98d0367d2247` Fix: `^` inside a paren that can match empty does not
anchor the pattern.

**Intl / Temporal**
- `09917ef55b0f` Add missing `U_FAILURE(status)` check in
`actualLunisolarMonthLength`.

**Inspector**
- `3722912ff800` / `7be5445e4a22` / `e8c97076834e` / `f3e34dde7c9d`
Canvas: instrument & record WebGPU devices/pipelines.
- `301d6b2b21f7` Associate WebAssembly module scripts with the fetching
resource.
- `28b979b1659b` / `479edb2e4395` Site Isolation: implement
`Network.loadResource` / `Network.getSerializedCertificate`.

### WTF

- `232cebabc1f3` Remove big-endian support and
`CPU(NEEDS_ALIGNED_ACCESS)`.
- `e5fa5c604438` Upgrade fast_float to 8.2.10.
- `a288a8ec809b` Widen `find16`/`find32` SIMD threshold; faster ASCII
case-conversion prefix copy.
- `6cb1077d85c8` `makeStringByReplacingAll()` now uses SIMD-accelerated
`find()`.
- `5590f2e70615` Fix `AdaptiveStringSearcher` good-suffix shift table
off-by-one.
- `f5716f6401ed` Add `preserve_most` to most fastMalloc APIs on ARM64.
- `f7a9d16e1531` Add `removeIf()` to `WeakHashSet` / `WeakListHashSet`.
- `e1fc460b8f1d` Add `removeIf()` to `RobinHoodHashTable`.
- `ff1f31c83dc7` Treat creating/destroying a `CheckedPtr` as no-delete.
- `bf15f00ebe95` Remove 12 unused internal-linkage templates
(TypeTraits/HashTable/Vector/etc.).
- `5b84cf3719fa` Use `__builtin_trap` instead of inline asm under clang
static analyzer.
- `158f737725b7` Add helpers for Darwin temp/cache directories.
- `04e3d47960f3` Limit URL size at IPC boundary (Chrome/Blink parity).
- `5daad377031c` / `a7ea27dd3bb6` Enable `-Wthread-safety` on GTK/WPE
and fix findings.
- `7eb640d408f8` CMake: merge Mac and iOS ports into "Cocoa".
- `cfb222f3c4d1` CMake: run `cleandead` at end of configuration.

### bmalloc

- `6eaa5ac1f65d` Remove 32-bit libpas support.
- `f5716f6401ed` `preserve_most` on fastMalloc APIs (ARM64).
- `8b8b3e5ee16c` Fix inverted `MADV_ZERO` support latch in
`VMAllocate.cpp`.
- `ead6285911f6` libpas: `pas_thread_local_cache_for_all` clobbered its
should-go-again result.
- `90cbe5e85528` libpas: fix benign read from a deallocated TLC.
- `e388877954d1` PGM allocator: fix uninitialized `free_status`
misclassifying OOB as UAF.
- `05c83a6550b7` libpas: fix
`MTE_overrideEnablementForJavaScriptCore=true` incorrectly disabling
MTE.
- `f01297663d40` / `86fb5e3a4eef` / `d18773ec666e` libpas test coverage
(scavenging / zeroing / paged-out pages).

### Breaking/notable for Bun

- **32-bit purge**: `29ceb3c03de3` (JSVALUE32_64 removed),
`857bd4334690` (ARMv7 JIT removed), `6eaa5ac1f65d` (32-bit libpas
removed), `232cebabc1f3` (big-endian + `CPU(NEEDS_ALIGNED_ACCESS)`
removed). Any `#if USE(JSVALUE64)` / `CPU(ADDRESS32)` guards in Bun
patches are now dead.
- **`VM` layout**: `f2f2c2ddf637` deletes
`StringRecursionChecker.{h,cpp}` and the `stringRecursionCheck*` fields
from `VM.h`. Cyclic `Array.prototype.join`/`toString` now throws
`RangeError` instead of returning `""`.
- **Module loader**: `90b2ecf79ae3` changes `hostResolveImportedModule`
to propagate the import-attribute `type` — check Bun's module loader
hook signatures.
- **Register allocator**: `bfb1b1183bc2` removes the B3/Air
graph-coloring allocator and its `Options::` toggle.
- **fastMalloc ABI (ARM64)**: `f5716f6401ed` adds
`__attribute__((preserve_most))` to `fastMalloc`/`fastFree` etc. —
affects anything calling these across the WTF boundary on arm64.
- **BuiltinNames**: `bf6512f84f7d` registers `stackPrivateName` as
private-only; `WebAssembly.Exception` constructor `length` becomes 2 and
gains a `stack` prototype getter.
- **Feature defaults**: `0d0080ea539d` Wasm Memory64 on by default;
`319f94b3db4a` `Iterator.prototype.includes` on by default.
- **Wasm threading**: `24527bbb9ac8` reworks icache barrier / callee
publication and adds `Thread::barrierInstructionCache()` in WTF.

<!-- robobun:evidence:begin -->

---

**[decide:webkit]** gate passed · iteration 2 · 8 files touched

<details><summary>fails on main (without fix)</summary>

```console
ASAN without fix: BUILD FAILED (no junit output)
$ BUN_DEBUG_QUIET_LOGS=1 bun scripts/build.ts --profile=debug --quiet test "--reporter=junit" "--reporter-outfile=/tmp/mechgate.xml" "test/js/bun/jsc/webkit-upgrade-3722912f.test.ts"
ninja: Entering directory `/workspace/bun/build/debug'
[1/182] gen generated_host_exports.rs
generated_host_exports.rs: 94 exports (host=3, lazy=10, generic=81, rust=0); 238 extern-C blocks audited
[2/182] gen cpp.rs (cppbind)
[3/182] gen JSSink.{cpp,h,lut.h,rs}
generated_jssink.rs: 7 sinks, 84 exported symbols
Generating /workspace/bun/build/debug/codegen/JSSink.lut.h from /workspace/bun/build/debug/codegen/JSSink.lut.txt
[4/182] gen JS modules (bundle-modules)
Preprocess modules (8715ms)
Bundle modules (63ms)
Postprocesss modules (24ms)
Bundle Functions (746ms)
Generate Code (12ms)

[9.57s] Bundled "src/js" for development
  2749 kb
  193 internal modules
  13 native modules
  90 internal functions across 19 files
[4/181] cargo bun_bin → libbun_rust.a (--target x86_64-unknown-linux-gnu)

  nightly-2026-07-20-x86_64-unknown-linux-gnu unchanged - rustc 1.99.0-nightly (9f36de775 2026-07-19)

[177/181] cxx obj/unified/UnifiedSource-src_jsc_bindings-0.cpp.o
FAILED: obj/unified/UnifiedSou
... (truncated)

release without fix: all passed
bun test v1.4.0-canary.1 (385f528)

test/js/bun/jsc/webkit-upgrade-3722912f.test.ts:
(pass) WebKit 3722912ff800 upgrade > Iterator.prototype.includes is enabled by default (319f94b3db4a) [0.19ms]
(pass) WebKit 3722912ff800 upgrade > cyclic Array.prototype.join throws RangeError (f2f2c2ddf637) [2.51ms]
(pass) WebKit 3722912ff800 upgrade > WebAssembly.Exception gains options.traceStack and stack getter (bf6512f84f7d) [0.48ms]
(pass) WebKit 3722912ff800 upgrade > typed import attributes resolve through BunTranspiledModule (--isolate) (90b2ecf79ae3) [8.49ms]
(pass) WebKit 3722912ff800 upgrade > indirect, namespace and star re-exports link on the JSC ModuleAnalyzer path (90b2ecf79ae3) [22.09ms]

 5 pass
 0 fail
 12 expect() calls
Ran 5 tests across 1 file. [152.00ms]
__F:0:S:0
```

</details>

<details><summary>passes on PR (with fix)</summary>

```console
ASAN with fix: all passed
$ BUN_DEBUG_QUIET_LOGS=1 bun scripts/build.ts --profile=debug --quiet test "--reporter=junit" "--reporter-outfile=/tmp/mechgate.xml" "test/js/bun/jsc/webkit-upgrade-3722912f.test.ts"
bun test v1.4.0 (59b0de0)

test/js/bun/jsc/webkit-upgrade-3722912f.test.ts:
(pass) WebKit 3722912ff800 upgrade > Iterator.prototype.includes is enabled by default (319f94b3db4a) [13.19ms]
(pass) WebKit 3722912ff800 upgrade > cyclic Array.prototype.join throws RangeError (f2f2c2ddf637) [10.75ms]
(pass) WebKit 3722912ff800 upgrade > WebAssembly.Exception gains options.traceStack and stack getter (bf6512f84f7d) [3.66ms]
(pass) WebKit 3722912ff800 upgrade > typed import attributes resolve through BunTranspiledModule (--isolate) (90b2ecf79ae3) [317.28ms]
(pass) WebKit 3722912ff800 upgrade > indirect, namespace and star re-exports link on the JSC ModuleAnalyzer path (90b2ecf79ae3) [1138.48ms]

 5 pass
 0 fail
 12 expect() calls
Ran 5 tests across 1 file. [3.17s]
__F:0:S:0

release with fix: all passed
$ bun scripts/build.ts --profile=release
[configured] bun-profile → bun (stripped) in 676ms (unchanged)
ninja: Entering directory `/workspace/bun/build/release'
[1/140] gen generated_host_exports.rs
generated_host_exports.rs: 94 exports (host=3, lazy=10, generic=81, rust=0); 238 extern-C blocks audited
[2/140] gen cpp.rs (cppbind)
[3/140] gen JSSink.{cpp,h,lut.h,rs}
generated_jssink.rs: 7 sinks, 84 exported symbols
Generating /workspace/bun/build/release/codegen/JSSink.lut.h from /workspace/bun/build/release/codegen/JSSink.lut.txt
[4/140] gen JS modules (bundle-modules)
Preprocess modules (8727ms)
Bundle modules (51ms)
Postprocesss modules (106ms)
Bundle Functions (687ms)
Generate Code (20ms)

[9.61s] Bundled "src/js" for production
  2559 kb
  193 internal modules
  13 native modules
  90 internal functions across 19 files
[4/139] cargo bun_bin → libbun_rust.a (--target x86_64-unknown-linux-gnu)

  nightly-2026-07-20-x86_64-unknown-linux-gnu unchanged - rustc 1.99.0-nightly (9f36de775 2026-07-19)

^[[1m^[[92m   Compiling^[[0m bun_core v0.0.0 (/workspace/bun/src/bun_core)
^[[1m^[[92m   Compiling^[[0m bun_runtime v0.0.0 (/workspace/bun/src/runtime)
^[[1m^[[92m   Compilin
... (truncated)
```

</details>

<details><summary>diff hotspot</summary>

```
scripts/build/deps/webkit.ts                     |   2 +-
 src/bundler/analyze_transpiled_module.rs         |  38 +++--
 src/bundler/linker_context/postProcessJSChunk.rs |   8 +-
 src/bundler_jsc/analyze_jsc.rs                   | 135 ++++++++++++----
 src/js_printer/lib.rs                            | 191 +++++++++++++++++------
 src/jsc/RuntimeTranspilerCache.rs                |   5 +-
 src/jsc/bindings/BunAnalyzeTranspiledModule.cpp  |  39 +++--
 test/js/bun/jsc/webkit-upgrade-3722912f.test.ts  | 106 +++++++++++++
 8 files changed, 416 insertions(+), 108 deletions(-)
```

</details>

**gate history** · 5 passed · 1 rejected · iteration 2

<details><summary>evidence per changed file</summary>

```
file                                              reads  edits  tests
scripts/build/deps/webkit.ts                          1      1      0
src/bundler/analyze_transpiled_module.rs              3      3      0
src/bundler/linker_context/postProcessJSChunk.rs      1      1      0
src/bundler_jsc/analyze_jsc.rs                       13     15      0
src/js_printer/lib.rs                                 9     16      0
src/jsc/RuntimeTranspilerCache.rs                     2      3      0
src/jsc/bindings/BunAnalyzeTranspiledModule.cpp       8     12      0
test/js/bun/jsc/webkit-upgrade-3722912f.test.ts       2      5      0
```

</details>

<!-- robobun:evidence:end -->

---------

Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
springmin pushed a commit to springmin/bun that referenced this pull request Aug 3, 2026
Upgrades the WebKit fork to upstream WebKit/WebKit@3722912ff800
(2026-08-02) via oven-sh/WebKit#383.

oven-sh/WebKit#383 is merged; `WEBKIT_VERSION` points at the merge
commit `e6e37cda216c0292ae68c30c84a9dc8601d0fba5`.

Upstream `90b2ecf79ae3` keys `m_loadedModules` on `(specifier,
ScriptFetchParameters::Type)` and threads the type through
`ImportEntry`, `ExportEntry`, `StarExportEntry`, and
`ModuleAnalyzer::appendRequestedModule`. Under `bun test --isolate` (the
`BunTranspiledModule` path), Bun'''s synthesized record must agree with
what JSC'''s own `ModuleAnalyzer` would produce from the printed source;
a mismatch fails the BUN_DEBUG record diff in debug and null-derefs
`hostResolveImportedModule` in release.

- `analyze_transpiled_module` / `js_printer`: every `RecordKind` now
carries one trailing `FetchParameters` slot. `add_import_info_*` and
`add_export_info_*` accept it; `finalize()` propagates the source
import'''s slot through the Local->Indirect conversion.
`RequestedModules` dedupes on `(specifier, Type, phase)`.
- `analyze_jsc.rs`: decodes the trailing slot to the JSC `Type` enum and
passes it to every `addImportEntry*` / `addIndirectExport` /
`addNamespaceExport` / `addStarExport`; buffer validation is per-slot so
only the trailing slot accepts the wider `FetchParameters` sentinel
range.
- `BunAnalyzeTranspiledModule.cpp`: all seven `addImportEntry*` /
`add*Export` functions take `uint8_t moduleRequestType` and set
`.moduleRequestType` explicitly; `dumpRecordInfo()` prints `type(N)` for
import/export/star entries; `static_assert` pins the ordinal values
`to_script_fetch_parameters_type()` hardcodes.
- `RuntimeTranspilerCache` `EXPECTED_VERSION` -> 25 (esm_record layout
change).

- 17 merge conflicts resolved in JSC/WTF; fork's
`USE(BUN_JSC_ADDITIONS)` hunks preserved.
- Fork `ffi/` code adapted to upstream's 32-bit removal (`USE_JSVALUE64`
macro deleted, `is64Bit()` removed, `payloadFor` -> `lowWordFor`).
- `InspectorDebuggerAgent.cpp`: handle `BunTranspiledModule`/`Synthetic`
in new `scriptTypeForScript` switch.
- Recorded `01aaa3e0be0c` as ancestor via `-s ours` (oven-sh/WebKit#352
was a squash merge).

- `bun run jsc:build:debug` builds and the `jsc` shell runs (`-e
'print(42)'` -> `42`).
- `bun run build:local -p '42'` links and runs against the local merged
WebKit.
- oven-sh/WebKit#383 preview build green on all 38 platform variants.
- `bun bd -p 'process.versions.webkit'` -> `preview-pr-383-b9ea4dc5`.
- New test `test/js/bun/jsc/webkit-upgrade-3722912f.test.ts`: 4/4 pass
with `bun bd test`, 3/4 fail with `USE_SYSTEM_BUN=1 bun test` (old JSC
lacks `Iterator.prototype.includes`, returns `""` for cyclic join,
`WebAssembly.Exception.length === 1`).

Upstream range: WebKit/WebKit@01aaa3e0be0c...3722912ff800 (474 commits
total, 110 touching JSC/WTF/bmalloc, Jul 25 → Aug 2 2026).

- `29ceb3c03de3` Remove 32-bit JSValues (JSVALUE32_64 and
`CPU(NEEDS_ALIGNED_ACCESS)` deleted).
- `857bd4334690` Remove ARMv7 JIT support (ARMv7 is CLoop-only now;
drops remaining 32-bit/x86 JIT refs).
- `232cebabc1f3` Remove big-endian support and platforms without
unaligned loads/stores (WTF + JSC).
- `6eaa5ac1f65d` Remove 32-bit libpas support.
- `bfb1b1183bc2` Remove the B3/Air graph-coloring register allocator
(greedy is the only allocator now).
- `0d0080ea539d` Enable WebAssembly Memory64 by default (+ Table64/SIMD
follow-ups).
- `f2f2c2ddf637` Remove `StringRecursionChecker`; cyclic
`toString`/`join` now throws RangeError via stack check (spec-correct).
- `319f94b3db4a` Enable `Iterator.prototype.includes()` by default.
- `90b2ecf79ae3` `hostResolveImportedModule` now honors import-attribute
`type` (module loader behavior change).
- `f5716f6401ed` Add `preserve_most` calling convention to fastMalloc
APIs on ARM64 (perf + ABI of WTF alloc entry points).

**Runtime / builtins**
- `f2f2c2ddf637` Remove StringRecursionChecker; rely on stack-overflow
checks.
- `cd91e7f128dd` Add fast flag for `ToPrimitive(Object)` calls in
runtime.
- `173a0bd6d937` Use `defaultToPrimitiveFastAndNonObservable` in
`JSObject::toString`.
- `960adeccefcd` Fast operation for `Array#shift`.
- `92c6650c7947` Add `JSArrayIterator::next` C++ helper.
- `cb1c48b3e95f` Extend `Array.from()` Set fast path to
`set.keys()/.values()`.
- `4a2e724d6789` Fix: `Array.from(set.keys()/.values())` fast path
ignored `Symbol.iterator` overrides.
- `73e4c589c1e6` Extend `StringSplitCache` to RegExp separators.
- `656d3c36830f` / `1d355c27ea88` 8-byte SWAR fast paths in
`JSON.stringify` string copy (same-type & upconvert).
- `9f9370cc729f` Fix JSON.stringify regression around `toJSON` check.
- `2eb77e9c9473` BigInt: implement Crandall reduction.
- `39b1bb9cfc85` BigInt: deploy Comba multiplication more broadly.
- `319f94b3db4a` Enable `Iterator.prototype.includes()`.
- `0731b27c1b60` `Iterator.zip`: use null-prototype objects for
options/underlying iterator.
- `90b2ecf79ae3` `hostResolveImportedModule` respects module request
import-attribute `type`.
- `4f54300b848a` Collect diagnostics when `getDirect` returns zero
JSValue in `llint_slow_path_get_by_id`.

**Parser / bytecompiler**
- `c2beca7a439f` Lexer: scan integer tokens in a single pass.
- `72ea806faa21` Use overflow-safe range when choosing a switch jump
table.

**LLInt / DFG / FTL / B3**
- `29ceb3c03de3` Remove 32-bit JSValues.
- `857bd4334690` Remove ARMv7 JIT support.
- `bfb1b1183bc2` Remove B3/Air graph-coloring register allocator.
- `68cde6ba2ad3` Make IRO (Air register allocation) faster.
- `83540481435f` DFG: allocate
`BasicBlock::intersectionOfPastValuesAtHead` only for OSR-entry targets.
- `1566615170ec` DFG fix: `EnumeratorNextUpdateIndexAndMode` must
require original array structure for `InBoundsSaneChain`.
- `e759fa9dd063` LLInt: inline hot path of `op_enter`.
- `9610c2113b45` offlineasm: emit ARM64 register-offset addressing for
BaseIndex operands.
- `4ebed2479144` Speed up `addSortedRange` via binary search.
- `1c006b0b0f62` Fix regex `setLastIndex` on 32-bit.

**WebAssembly**
- `0d0080ea539d` Enable Memory64 feature flag.
- `15aa6fad53e3` Memory64: SIMD support.
- `bf0425598904` Memory64: expand declared memory limits.
- `862994e2cc37` Memory64: validate table import address-type match.
- `184ee4c654bd` Memory64: Table64 in OMG tier.
- `d202bedc5ff6` Memory64: Table64 in BBQ tier.
- `bf6512f84f7d` Support `WebAssembly.Exception` `options.traceStack` (+
`stack` getter, ctor length = 2).
- `24527bbb9ac8` Optimize Wasm JITCallee publication (lock splitting,
icache barrier rework).
- `1803d6109d98` Speed up `WebAssembly.Table` construction.
- `d434a41411a3` BBQ: optimize `br_table` for consecutive same-target
runs.
- `51d3dbaa278e` IPInt: add `DEFINE_IPINT_THUNK_FOR_ENTRY`.
- `244cd98f7986` Fix `generateWasmOpsHeader.py` under non-UTF-8 locales.

**Yarr / RegExp**
- `581f1d958329` Start end-anchored fixed-size regexps at the only
possible position.
- `a458a6c1f0a7` v-mode class-set op loop: stop early when no more
output possible.
- `7c5dbbcba110` Extend `ParenthesesSubpatternTerminal`.
- `e1def8f4e5fd` Don't save sibling/ancestor-sibling frame slots for
`ParenContext`.
- `1e43057f135a` Extend first-character filter further.
- `54916608d7d6` Fix non-BMP advance latch; simplify
`tryReadUnicodeCharImpl`.
- `46a4b17efbe9` `optimizeBOL`: don't filter contents of negative
lookaheads.
- `b128ddd863ab` Fix dot-star-wrapped optimization for sticky patterns.
- `98d0367d2247` Fix: `^` inside a paren that can match empty does not
anchor the pattern.

**Intl / Temporal**
- `09917ef55b0f` Add missing `U_FAILURE(status)` check in
`actualLunisolarMonthLength`.

**Inspector**
- `3722912ff800` / `7be5445e4a22` / `e8c97076834e` / `f3e34dde7c9d`
Canvas: instrument & record WebGPU devices/pipelines.
- `301d6b2b21f7` Associate WebAssembly module scripts with the fetching
resource.
- `28b979b1659b` / `479edb2e4395` Site Isolation: implement
`Network.loadResource` / `Network.getSerializedCertificate`.

- `232cebabc1f3` Remove big-endian support and
`CPU(NEEDS_ALIGNED_ACCESS)`.
- `e5fa5c604438` Upgrade fast_float to 8.2.10.
- `a288a8ec809b` Widen `find16`/`find32` SIMD threshold; faster ASCII
case-conversion prefix copy.
- `6cb1077d85c8` `makeStringByReplacingAll()` now uses SIMD-accelerated
`find()`.
- `5590f2e70615` Fix `AdaptiveStringSearcher` good-suffix shift table
off-by-one.
- `f5716f6401ed` Add `preserve_most` to most fastMalloc APIs on ARM64.
- `f7a9d16e1531` Add `removeIf()` to `WeakHashSet` / `WeakListHashSet`.
- `e1fc460b8f1d` Add `removeIf()` to `RobinHoodHashTable`.
- `ff1f31c83dc7` Treat creating/destroying a `CheckedPtr` as no-delete.
- `bf15f00ebe95` Remove 12 unused internal-linkage templates
(TypeTraits/HashTable/Vector/etc.).
- `5b84cf3719fa` Use `__builtin_trap` instead of inline asm under clang
static analyzer.
- `158f737725b7` Add helpers for Darwin temp/cache directories.
- `04e3d47960f3` Limit URL size at IPC boundary (Chrome/Blink parity).
- `5daad377031c` / `a7ea27dd3bb6` Enable `-Wthread-safety` on GTK/WPE
and fix findings.
- `7eb640d408f8` CMake: merge Mac and iOS ports into "Cocoa".
- `cfb222f3c4d1` CMake: run `cleandead` at end of configuration.

- `6eaa5ac1f65d` Remove 32-bit libpas support.
- `f5716f6401ed` `preserve_most` on fastMalloc APIs (ARM64).
- `8b8b3e5ee16c` Fix inverted `MADV_ZERO` support latch in
`VMAllocate.cpp`.
- `ead6285911f6` libpas: `pas_thread_local_cache_for_all` clobbered its
should-go-again result.
- `90cbe5e85528` libpas: fix benign read from a deallocated TLC.
- `e388877954d1` PGM allocator: fix uninitialized `free_status`
misclassifying OOB as UAF.
- `05c83a6550b7` libpas: fix
`MTE_overrideEnablementForJavaScriptCore=true` incorrectly disabling
MTE.
- `f01297663d40` / `86fb5e3a4eef` / `d18773ec666e` libpas test coverage
(scavenging / zeroing / paged-out pages).

- **32-bit purge**: `29ceb3c03de3` (JSVALUE32_64 removed),
`857bd4334690` (ARMv7 JIT removed), `6eaa5ac1f65d` (32-bit libpas
removed), `232cebabc1f3` (big-endian + `CPU(NEEDS_ALIGNED_ACCESS)`
removed). Any `#if USE(JSVALUE64)` / `CPU(ADDRESS32)` guards in Bun
patches are now dead.
- **`VM` layout**: `f2f2c2ddf637` deletes
`StringRecursionChecker.{h,cpp}` and the `stringRecursionCheck*` fields
from `VM.h`. Cyclic `Array.prototype.join`/`toString` now throws
`RangeError` instead of returning `""`.
- **Module loader**: `90b2ecf79ae3` changes `hostResolveImportedModule`
to propagate the import-attribute `type` — check Bun's module loader
hook signatures.
- **Register allocator**: `bfb1b1183bc2` removes the B3/Air
graph-coloring allocator and its `Options::` toggle.
- **fastMalloc ABI (ARM64)**: `f5716f6401ed` adds
`__attribute__((preserve_most))` to `fastMalloc`/`fastFree` etc. —
affects anything calling these across the WTF boundary on arm64.
- **BuiltinNames**: `bf6512f84f7d` registers `stackPrivateName` as
private-only; `WebAssembly.Exception` constructor `length` becomes 2 and
gains a `stack` prototype getter.
- **Feature defaults**: `0d0080ea539d` Wasm Memory64 on by default;
`319f94b3db4a` `Iterator.prototype.includes` on by default.
- **Wasm threading**: `24527bbb9ac8` reworks icache barrier / callee
publication and adds `Thread::barrierInstructionCache()` in WTF.

<!-- robobun:evidence:begin -->

---

**[decide:webkit]** gate passed · iteration 2 · 8 files touched

<details><summary>fails on main (without fix)</summary>

```console
ASAN without fix: BUILD FAILED (no junit output)
$ BUN_DEBUG_QUIET_LOGS=1 bun scripts/build.ts --profile=debug --quiet test "--reporter=junit" "--reporter-outfile=/tmp/mechgate.xml" "test/js/bun/jsc/webkit-upgrade-3722912f.test.ts"
ninja: Entering directory `/workspace/bun/build/debug'
[1/182] gen generated_host_exports.rs
generated_host_exports.rs: 94 exports (host=3, lazy=10, generic=81, rust=0); 238 extern-C blocks audited
[2/182] gen cpp.rs (cppbind)
[3/182] gen JSSink.{cpp,h,lut.h,rs}
generated_jssink.rs: 7 sinks, 84 exported symbols
Generating /workspace/bun/build/debug/codegen/JSSink.lut.h from /workspace/bun/build/debug/codegen/JSSink.lut.txt
[4/182] gen JS modules (bundle-modules)
Preprocess modules (8715ms)
Bundle modules (63ms)
Postprocesss modules (24ms)
Bundle Functions (746ms)
Generate Code (12ms)

[9.57s] Bundled "src/js" for development
  2749 kb
  193 internal modules
  13 native modules
  90 internal functions across 19 files
[4/181] cargo bun_bin → libbun_rust.a (--target x86_64-unknown-linux-gnu)

  nightly-2026-07-20-x86_64-unknown-linux-gnu unchanged - rustc 1.99.0-nightly (9f36de775 2026-07-19)

[177/181] cxx obj/unified/UnifiedSource-src_jsc_bindings-0.cpp.o
FAILED: obj/unified/UnifiedSou
... (truncated)

release without fix: all passed
bun test v1.4.0-canary.1 (385f528)

test/js/bun/jsc/webkit-upgrade-3722912f.test.ts:
(pass) WebKit 3722912ff800 upgrade > Iterator.prototype.includes is enabled by default (319f94b3db4a) [0.19ms]
(pass) WebKit 3722912ff800 upgrade > cyclic Array.prototype.join throws RangeError (f2f2c2ddf637) [2.51ms]
(pass) WebKit 3722912ff800 upgrade > WebAssembly.Exception gains options.traceStack and stack getter (bf6512f84f7d) [0.48ms]
(pass) WebKit 3722912ff800 upgrade > typed import attributes resolve through BunTranspiledModule (--isolate) (90b2ecf79ae3) [8.49ms]
(pass) WebKit 3722912ff800 upgrade > indirect, namespace and star re-exports link on the JSC ModuleAnalyzer path (90b2ecf79ae3) [22.09ms]

 5 pass
 0 fail
 12 expect() calls
Ran 5 tests across 1 file. [152.00ms]
__F:0:S:0
```

</details>

<details><summary>passes on PR (with fix)</summary>

```console
ASAN with fix: all passed
$ BUN_DEBUG_QUIET_LOGS=1 bun scripts/build.ts --profile=debug --quiet test "--reporter=junit" "--reporter-outfile=/tmp/mechgate.xml" "test/js/bun/jsc/webkit-upgrade-3722912f.test.ts"
bun test v1.4.0 (59b0de0)

test/js/bun/jsc/webkit-upgrade-3722912f.test.ts:
(pass) WebKit 3722912ff800 upgrade > Iterator.prototype.includes is enabled by default (319f94b3db4a) [13.19ms]
(pass) WebKit 3722912ff800 upgrade > cyclic Array.prototype.join throws RangeError (f2f2c2ddf637) [10.75ms]
(pass) WebKit 3722912ff800 upgrade > WebAssembly.Exception gains options.traceStack and stack getter (bf6512f84f7d) [3.66ms]
(pass) WebKit 3722912ff800 upgrade > typed import attributes resolve through BunTranspiledModule (--isolate) (90b2ecf79ae3) [317.28ms]
(pass) WebKit 3722912ff800 upgrade > indirect, namespace and star re-exports link on the JSC ModuleAnalyzer path (90b2ecf79ae3) [1138.48ms]

 5 pass
 0 fail
 12 expect() calls
Ran 5 tests across 1 file. [3.17s]
__F:0:S:0

release with fix: all passed
$ bun scripts/build.ts --profile=release
[configured] bun-profile → bun (stripped) in 676ms (unchanged)
ninja: Entering directory `/workspace/bun/build/release'
[1/140] gen generated_host_exports.rs
generated_host_exports.rs: 94 exports (host=3, lazy=10, generic=81, rust=0); 238 extern-C blocks audited
[2/140] gen cpp.rs (cppbind)
[3/140] gen JSSink.{cpp,h,lut.h,rs}
generated_jssink.rs: 7 sinks, 84 exported symbols
Generating /workspace/bun/build/release/codegen/JSSink.lut.h from /workspace/bun/build/release/codegen/JSSink.lut.txt
[4/140] gen JS modules (bundle-modules)
Preprocess modules (8727ms)
Bundle modules (51ms)
Postprocesss modules (106ms)
Bundle Functions (687ms)
Generate Code (20ms)

[9.61s] Bundled "src/js" for production
  2559 kb
  193 internal modules
  13 native modules
  90 internal functions across 19 files
[4/139] cargo bun_bin → libbun_rust.a (--target x86_64-unknown-linux-gnu)

  nightly-2026-07-20-x86_64-unknown-linux-gnu unchanged - rustc 1.99.0-nightly (9f36de775 2026-07-19)

^[[1m^[[92m   Compiling^[[0m bun_core v0.0.0 (/workspace/bun/src/bun_core)
^[[1m^[[92m   Compiling^[[0m bun_runtime v0.0.0 (/workspace/bun/src/runtime)
^[[1m^[[92m   Compilin
... (truncated)
```

</details>

<details><summary>diff hotspot</summary>

```
scripts/build/deps/webkit.ts                     |   2 +-
 src/bundler/analyze_transpiled_module.rs         |  38 +++--
 src/bundler/linker_context/postProcessJSChunk.rs |   8 +-
 src/bundler_jsc/analyze_jsc.rs                   | 135 ++++++++++++----
 src/js_printer/lib.rs                            | 191 +++++++++++++++++------
 src/jsc/RuntimeTranspilerCache.rs                |   5 +-
 src/jsc/bindings/BunAnalyzeTranspiledModule.cpp  |  39 +++--
 test/js/bun/jsc/webkit-upgrade-3722912f.test.ts  | 106 +++++++++++++
 8 files changed, 416 insertions(+), 108 deletions(-)
```

</details>

**gate history** · 5 passed · 1 rejected · iteration 2

<details><summary>evidence per changed file</summary>

```
file                                              reads  edits  tests
scripts/build/deps/webkit.ts                          1      1      0
src/bundler/analyze_transpiled_module.rs              3      3      0
src/bundler/linker_context/postProcessJSChunk.rs      1      1      0
src/bundler_jsc/analyze_jsc.rs                       13     15      0
src/js_printer/lib.rs                                 9     16      0
src/jsc/RuntimeTranspilerCache.rs                     2      3      0
src/jsc/bindings/BunAnalyzeTranspiledModule.cpp       8     12      0
test/js/bun/jsc/webkit-upgrade-3722912f.test.ts       2      5      0
```

</details>

<!-- robobun:evidence:end -->

---------

Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
liooil pushed a commit to liooil/poly that referenced this pull request Aug 7, 2026
Upgrades the WebKit fork to upstream WebKit/WebKit@3722912ff800
(2026-08-02) via oven-sh/WebKit#383.

oven-sh/WebKit#383 is merged; `WEBKIT_VERSION` points at the merge
commit `e6e37cda216c0292ae68c30c84a9dc8601d0fba5`.

## Bun-side changes

Upstream `90b2ecf79ae3` keys `m_loadedModules` on `(specifier,
ScriptFetchParameters::Type)` and threads the type through
`ImportEntry`, `ExportEntry`, `StarExportEntry`, and
`ModuleAnalyzer::appendRequestedModule`. Under `bun test --isolate` (the
`BunTranspiledModule` path), Bun'''s synthesized record must agree with
what JSC'''s own `ModuleAnalyzer` would produce from the printed source;
a mismatch fails the BUN_DEBUG record diff in debug and null-derefs
`hostResolveImportedModule` in release.

- `analyze_transpiled_module` / `js_printer`: every `RecordKind` now
carries one trailing `FetchParameters` slot. `add_import_info_*` and
`add_export_info_*` accept it; `finalize()` propagates the source
import'''s slot through the Local->Indirect conversion.
`RequestedModules` dedupes on `(specifier, Type, phase)`.
- `analyze_jsc.rs`: decodes the trailing slot to the JSC `Type` enum and
passes it to every `addImportEntry*` / `addIndirectExport` /
`addNamespaceExport` / `addStarExport`; buffer validation is per-slot so
only the trailing slot accepts the wider `FetchParameters` sentinel
range.
- `BunAnalyzeTranspiledModule.cpp`: all seven `addImportEntry*` /
`add*Export` functions take `uint8_t moduleRequestType` and set
`.moduleRequestType` explicitly; `dumpRecordInfo()` prints `type(N)` for
import/export/star entries; `static_assert` pins the ordinal values
`to_script_fetch_parameters_type()` hardcodes.
- `RuntimeTranspilerCache` `EXPECTED_VERSION` -> 25 (esm_record layout
change).

## WebKit-side changes (oven-sh/WebKit#383)

- 17 merge conflicts resolved in JSC/WTF; fork's
`USE(BUN_JSC_ADDITIONS)` hunks preserved.
- Fork `ffi/` code adapted to upstream's 32-bit removal (`USE_JSVALUE64`
macro deleted, `is64Bit()` removed, `payloadFor` -> `lowWordFor`).
- `InspectorDebuggerAgent.cpp`: handle `BunTranspiledModule`/`Synthetic`
in new `scriptTypeForScript` switch.
- Recorded `01aaa3e0be0c` as ancestor via `-s ours` (oven-sh/WebKit#352
was a squash merge).

## How did you verify your code works?

- `bun run jsc:build:debug` builds and the `jsc` shell runs (`-e
'print(42)'` -> `42`).
- `bun run build:local -p '42'` links and runs against the local merged
WebKit.
- oven-sh/WebKit#383 preview build green on all 38 platform variants.
- `bun bd -p 'process.versions.webkit'` -> `preview-pr-383-b9ea4dc5`.
- New test `test/js/bun/jsc/webkit-upgrade-3722912f.test.ts`: 4/4 pass
with `bun bd test`, 3/4 fail with `USE_SYSTEM_BUN=1 bun test` (old JSC
lacks `Iterator.prototype.includes`, returns `""` for cyclic join,
`WebAssembly.Exception.length === 1`).

## JavaScriptCore/WTF/bmalloc changes since 01aaa3e0be0c (Jul 25)

Upstream range: WebKit/WebKit@01aaa3e0be0c...3722912ff800 (474 commits
total, 110 touching JSC/WTF/bmalloc, Jul 25 → Aug 2 2026).

### Highlights

- `29ceb3c03de3` Remove 32-bit JSValues (JSVALUE32_64 and
`CPU(NEEDS_ALIGNED_ACCESS)` deleted).
- `857bd4334690` Remove ARMv7 JIT support (ARMv7 is CLoop-only now;
drops remaining 32-bit/x86 JIT refs).
- `232cebabc1f3` Remove big-endian support and platforms without
unaligned loads/stores (WTF + JSC).
- `6eaa5ac1f65d` Remove 32-bit libpas support.
- `bfb1b1183bc2` Remove the B3/Air graph-coloring register allocator
(greedy is the only allocator now).
- `0d0080ea539d` Enable WebAssembly Memory64 by default (+ Table64/SIMD
follow-ups).
- `f2f2c2ddf637` Remove `StringRecursionChecker`; cyclic
`toString`/`join` now throws RangeError via stack check (spec-correct).
- `319f94b3db4a` Enable `Iterator.prototype.includes()` by default.
- `90b2ecf79ae3` `hostResolveImportedModule` now honors import-attribute
`type` (module loader behavior change).
- `f5716f6401ed` Add `preserve_most` calling convention to fastMalloc
APIs on ARM64 (perf + ABI of WTF alloc entry points).

### JavaScriptCore

**Runtime / builtins**
- `f2f2c2ddf637` Remove StringRecursionChecker; rely on stack-overflow
checks.
- `cd91e7f128dd` Add fast flag for `ToPrimitive(Object)` calls in
runtime.
- `173a0bd6d937` Use `defaultToPrimitiveFastAndNonObservable` in
`JSObject::toString`.
- `960adeccefcd` Fast operation for `Array#shift`.
- `92c6650c7947` Add `JSArrayIterator::next` C++ helper.
- `cb1c48b3e95f` Extend `Array.from()` Set fast path to
`set.keys()/.values()`.
- `4a2e724d6789` Fix: `Array.from(set.keys()/.values())` fast path
ignored `Symbol.iterator` overrides.
- `73e4c589c1e6` Extend `StringSplitCache` to RegExp separators.
- `656d3c36830f` / `1d355c27ea88` 8-byte SWAR fast paths in
`JSON.stringify` string copy (same-type & upconvert).
- `9f9370cc729f` Fix JSON.stringify regression around `toJSON` check.
- `2eb77e9c9473` BigInt: implement Crandall reduction.
- `39b1bb9cfc85` BigInt: deploy Comba multiplication more broadly.
- `319f94b3db4a` Enable `Iterator.prototype.includes()`.
- `0731b27c1b60` `Iterator.zip`: use null-prototype objects for
options/underlying iterator.
- `90b2ecf79ae3` `hostResolveImportedModule` respects module request
import-attribute `type`.
- `4f54300b848a` Collect diagnostics when `getDirect` returns zero
JSValue in `llint_slow_path_get_by_id`.

**Parser / bytecompiler**
- `c2beca7a439f` Lexer: scan integer tokens in a single pass.
- `72ea806faa21` Use overflow-safe range when choosing a switch jump
table.

**LLInt / DFG / FTL / B3**
- `29ceb3c03de3` Remove 32-bit JSValues.
- `857bd4334690` Remove ARMv7 JIT support.
- `bfb1b1183bc2` Remove B3/Air graph-coloring register allocator.
- `68cde6ba2ad3` Make IRO (Air register allocation) faster.
- `83540481435f` DFG: allocate
`BasicBlock::intersectionOfPastValuesAtHead` only for OSR-entry targets.
- `1566615170ec` DFG fix: `EnumeratorNextUpdateIndexAndMode` must
require original array structure for `InBoundsSaneChain`.
- `e759fa9dd063` LLInt: inline hot path of `op_enter`.
- `9610c2113b45` offlineasm: emit ARM64 register-offset addressing for
BaseIndex operands.
- `4ebed2479144` Speed up `addSortedRange` via binary search.
- `1c006b0b0f62` Fix regex `setLastIndex` on 32-bit.

**WebAssembly**
- `0d0080ea539d` Enable Memory64 feature flag.
- `15aa6fad53e3` Memory64: SIMD support.
- `bf0425598904` Memory64: expand declared memory limits.
- `862994e2cc37` Memory64: validate table import address-type match.
- `184ee4c654bd` Memory64: Table64 in OMG tier.
- `d202bedc5ff6` Memory64: Table64 in BBQ tier.
- `bf6512f84f7d` Support `WebAssembly.Exception` `options.traceStack` (+
`stack` getter, ctor length = 2).
- `24527bbb9ac8` Optimize Wasm JITCallee publication (lock splitting,
icache barrier rework).
- `1803d6109d98` Speed up `WebAssembly.Table` construction.
- `d434a41411a3` BBQ: optimize `br_table` for consecutive same-target
runs.
- `51d3dbaa278e` IPInt: add `DEFINE_IPINT_THUNK_FOR_ENTRY`.
- `244cd98f7986` Fix `generateWasmOpsHeader.py` under non-UTF-8 locales.

**Yarr / RegExp**
- `581f1d958329` Start end-anchored fixed-size regexps at the only
possible position.
- `a458a6c1f0a7` v-mode class-set op loop: stop early when no more
output possible.
- `7c5dbbcba110` Extend `ParenthesesSubpatternTerminal`.
- `e1def8f4e5fd` Don't save sibling/ancestor-sibling frame slots for
`ParenContext`.
- `1e43057f135a` Extend first-character filter further.
- `54916608d7d6` Fix non-BMP advance latch; simplify
`tryReadUnicodeCharImpl`.
- `46a4b17efbe9` `optimizeBOL`: don't filter contents of negative
lookaheads.
- `b128ddd863ab` Fix dot-star-wrapped optimization for sticky patterns.
- `98d0367d2247` Fix: `^` inside a paren that can match empty does not
anchor the pattern.

**Intl / Temporal**
- `09917ef55b0f` Add missing `U_FAILURE(status)` check in
`actualLunisolarMonthLength`.

**Inspector**
- `3722912ff800` / `7be5445e4a22` / `e8c97076834e` / `f3e34dde7c9d`
Canvas: instrument & record WebGPU devices/pipelines.
- `301d6b2b21f7` Associate WebAssembly module scripts with the fetching
resource.
- `28b979b1659b` / `479edb2e4395` Site Isolation: implement
`Network.loadResource` / `Network.getSerializedCertificate`.

### WTF

- `232cebabc1f3` Remove big-endian support and
`CPU(NEEDS_ALIGNED_ACCESS)`.
- `e5fa5c604438` Upgrade fast_float to 8.2.10.
- `a288a8ec809b` Widen `find16`/`find32` SIMD threshold; faster ASCII
case-conversion prefix copy.
- `6cb1077d85c8` `makeStringByReplacingAll()` now uses SIMD-accelerated
`find()`.
- `5590f2e70615` Fix `AdaptiveStringSearcher` good-suffix shift table
off-by-one.
- `f5716f6401ed` Add `preserve_most` to most fastMalloc APIs on ARM64.
- `f7a9d16e1531` Add `removeIf()` to `WeakHashSet` / `WeakListHashSet`.
- `e1fc460b8f1d` Add `removeIf()` to `RobinHoodHashTable`.
- `ff1f31c83dc7` Treat creating/destroying a `CheckedPtr` as no-delete.
- `bf15f00ebe95` Remove 12 unused internal-linkage templates
(TypeTraits/HashTable/Vector/etc.).
- `5b84cf3719fa` Use `__builtin_trap` instead of inline asm under clang
static analyzer.
- `158f737725b7` Add helpers for Darwin temp/cache directories.
- `04e3d47960f3` Limit URL size at IPC boundary (Chrome/Blink parity).
- `5daad377031c` / `a7ea27dd3bb6` Enable `-Wthread-safety` on GTK/WPE
and fix findings.
- `7eb640d408f8` CMake: merge Mac and iOS ports into "Cocoa".
- `cfb222f3c4d1` CMake: run `cleandead` at end of configuration.

### bmalloc

- `6eaa5ac1f65d` Remove 32-bit libpas support.
- `f5716f6401ed` `preserve_most` on fastMalloc APIs (ARM64).
- `8b8b3e5ee16c` Fix inverted `MADV_ZERO` support latch in
`VMAllocate.cpp`.
- `ead6285911f6` libpas: `pas_thread_local_cache_for_all` clobbered its
should-go-again result.
- `90cbe5e85528` libpas: fix benign read from a deallocated TLC.
- `e388877954d1` PGM allocator: fix uninitialized `free_status`
misclassifying OOB as UAF.
- `05c83a6550b7` libpas: fix
`MTE_overrideEnablementForJavaScriptCore=true` incorrectly disabling
MTE.
- `f01297663d40` / `86fb5e3a4eef` / `d18773ec666e` libpas test coverage
(scavenging / zeroing / paged-out pages).

### Breaking/notable for Bun

- **32-bit purge**: `29ceb3c03de3` (JSVALUE32_64 removed),
`857bd4334690` (ARMv7 JIT removed), `6eaa5ac1f65d` (32-bit libpas
removed), `232cebabc1f3` (big-endian + `CPU(NEEDS_ALIGNED_ACCESS)`
removed). Any `#if USE(JSVALUE64)` / `CPU(ADDRESS32)` guards in Bun
patches are now dead.
- **`VM` layout**: `f2f2c2ddf637` deletes
`StringRecursionChecker.{h,cpp}` and the `stringRecursionCheck*` fields
from `VM.h`. Cyclic `Array.prototype.join`/`toString` now throws
`RangeError` instead of returning `""`.
- **Module loader**: `90b2ecf79ae3` changes `hostResolveImportedModule`
to propagate the import-attribute `type` — check Bun's module loader
hook signatures.
- **Register allocator**: `bfb1b1183bc2` removes the B3/Air
graph-coloring allocator and its `Options::` toggle.
- **fastMalloc ABI (ARM64)**: `f5716f6401ed` adds
`__attribute__((preserve_most))` to `fastMalloc`/`fastFree` etc. —
affects anything calling these across the WTF boundary on arm64.
- **BuiltinNames**: `bf6512f84f7d` registers `stackPrivateName` as
private-only; `WebAssembly.Exception` constructor `length` becomes 2 and
gains a `stack` prototype getter.
- **Feature defaults**: `0d0080ea539d` Wasm Memory64 on by default;
`319f94b3db4a` `Iterator.prototype.includes` on by default.
- **Wasm threading**: `24527bbb9ac8` reworks icache barrier / callee
publication and adds `Thread::barrierInstructionCache()` in WTF.

<!-- robobun:evidence:begin -->

---

**[decide:webkit]** gate passed · iteration 2 · 8 files touched

<details><summary>fails on main (without fix)</summary>

```console
ASAN without fix: BUILD FAILED (no junit output)
$ BUN_DEBUG_QUIET_LOGS=1 bun scripts/build.ts --profile=debug --quiet test "--reporter=junit" "--reporter-outfile=/tmp/mechgate.xml" "test/js/bun/jsc/webkit-upgrade-3722912f.test.ts"
ninja: Entering directory `/workspace/bun/build/debug'
[1/182] gen generated_host_exports.rs
generated_host_exports.rs: 94 exports (host=3, lazy=10, generic=81, rust=0); 238 extern-C blocks audited
[2/182] gen cpp.rs (cppbind)
[3/182] gen JSSink.{cpp,h,lut.h,rs}
generated_jssink.rs: 7 sinks, 84 exported symbols
Generating /workspace/bun/build/debug/codegen/JSSink.lut.h from /workspace/bun/build/debug/codegen/JSSink.lut.txt
[4/182] gen JS modules (bundle-modules)
Preprocess modules (8715ms)
Bundle modules (63ms)
Postprocesss modules (24ms)
Bundle Functions (746ms)
Generate Code (12ms)

[9.57s] Bundled "src/js" for development
  2749 kb
  193 internal modules
  13 native modules
  90 internal functions across 19 files
[4/181] cargo bun_bin → libbun_rust.a (--target x86_64-unknown-linux-gnu)

  nightly-2026-07-20-x86_64-unknown-linux-gnu unchanged - rustc 1.99.0-nightly (9f36de775 2026-07-19)

[177/181] cxx obj/unified/UnifiedSource-src_jsc_bindings-0.cpp.o
FAILED: obj/unified/UnifiedSou
... (truncated)

release without fix: all passed
bun test v1.4.0-canary.1 (385f52890)

test/js/bun/jsc/webkit-upgrade-3722912f.test.ts:
(pass) WebKit 3722912ff800 upgrade > Iterator.prototype.includes is enabled by default (319f94b3db4a) [0.19ms]
(pass) WebKit 3722912ff800 upgrade > cyclic Array.prototype.join throws RangeError (f2f2c2ddf637) [2.51ms]
(pass) WebKit 3722912ff800 upgrade > WebAssembly.Exception gains options.traceStack and stack getter (bf6512f84f7d) [0.48ms]
(pass) WebKit 3722912ff800 upgrade > typed import attributes resolve through BunTranspiledModule (--isolate) (90b2ecf79ae3) [8.49ms]
(pass) WebKit 3722912ff800 upgrade > indirect, namespace and star re-exports link on the JSC ModuleAnalyzer path (90b2ecf79ae3) [22.09ms]

 5 pass
 0 fail
 12 expect() calls
Ran 5 tests across 1 file. [152.00ms]
__F:0:S:0
```

</details>

<details><summary>passes on PR (with fix)</summary>

```console
ASAN with fix: all passed
$ BUN_DEBUG_QUIET_LOGS=1 bun scripts/build.ts --profile=debug --quiet test "--reporter=junit" "--reporter-outfile=/tmp/mechgate.xml" "test/js/bun/jsc/webkit-upgrade-3722912f.test.ts"
bun test v1.4.0 (59b0de097)

test/js/bun/jsc/webkit-upgrade-3722912f.test.ts:
(pass) WebKit 3722912ff800 upgrade > Iterator.prototype.includes is enabled by default (319f94b3db4a) [13.19ms]
(pass) WebKit 3722912ff800 upgrade > cyclic Array.prototype.join throws RangeError (f2f2c2ddf637) [10.75ms]
(pass) WebKit 3722912ff800 upgrade > WebAssembly.Exception gains options.traceStack and stack getter (bf6512f84f7d) [3.66ms]
(pass) WebKit 3722912ff800 upgrade > typed import attributes resolve through BunTranspiledModule (--isolate) (90b2ecf79ae3) [317.28ms]
(pass) WebKit 3722912ff800 upgrade > indirect, namespace and star re-exports link on the JSC ModuleAnalyzer path (90b2ecf79ae3) [1138.48ms]

 5 pass
 0 fail
 12 expect() calls
Ran 5 tests across 1 file. [3.17s]
__F:0:S:0

release with fix: all passed
$ bun scripts/build.ts --profile=release
[configured] bun-profile → bun (stripped) in 676ms (unchanged)
ninja: Entering directory `/workspace/bun/build/release'
[1/140] gen generated_host_exports.rs
generated_host_exports.rs: 94 exports (host=3, lazy=10, generic=81, rust=0); 238 extern-C blocks audited
[2/140] gen cpp.rs (cppbind)
[3/140] gen JSSink.{cpp,h,lut.h,rs}
generated_jssink.rs: 7 sinks, 84 exported symbols
Generating /workspace/bun/build/release/codegen/JSSink.lut.h from /workspace/bun/build/release/codegen/JSSink.lut.txt
[4/140] gen JS modules (bundle-modules)
Preprocess modules (8727ms)
Bundle modules (51ms)
Postprocesss modules (106ms)
Bundle Functions (687ms)
Generate Code (20ms)

[9.61s] Bundled "src/js" for production
  2559 kb
  193 internal modules
  13 native modules
  90 internal functions across 19 files
[4/139] cargo bun_bin → libbun_rust.a (--target x86_64-unknown-linux-gnu)

  nightly-2026-07-20-x86_64-unknown-linux-gnu unchanged - rustc 1.99.0-nightly (9f36de775 2026-07-19)

^[[1m^[[92m   Compiling^[[0m bun_core v0.0.0 (/workspace/bun/src/bun_core)
^[[1m^[[92m   Compiling^[[0m bun_runtime v0.0.0 (/workspace/bun/src/runtime)
^[[1m^[[92m   Compilin
... (truncated)
```

</details>

<details><summary>diff hotspot</summary>

```
scripts/build/deps/webkit.ts                     |   2 +-
 src/bundler/analyze_transpiled_module.rs         |  38 +++--
 src/bundler/linker_context/postProcessJSChunk.rs |   8 +-
 src/bundler_jsc/analyze_jsc.rs                   | 135 ++++++++++++----
 src/js_printer/lib.rs                            | 191 +++++++++++++++++------
 src/jsc/RuntimeTranspilerCache.rs                |   5 +-
 src/jsc/bindings/BunAnalyzeTranspiledModule.cpp  |  39 +++--
 test/js/bun/jsc/webkit-upgrade-3722912f.test.ts  | 106 +++++++++++++
 8 files changed, 416 insertions(+), 108 deletions(-)
```

</details>

**gate history** · 5 passed · 1 rejected · iteration 2

<details><summary>evidence per changed file</summary>

```
file                                              reads  edits  tests
scripts/build/deps/webkit.ts                          1      1      0
src/bundler/analyze_transpiled_module.rs              3      3      0
src/bundler/linker_context/postProcessJSChunk.rs      1      1      0
src/bundler_jsc/analyze_jsc.rs                       13     15      0
src/js_printer/lib.rs                                 9     16      0
src/jsc/RuntimeTranspilerCache.rs                     2      3      0
src/jsc/bindings/BunAnalyzeTranspiledModule.cpp       8     12      0
test/js/bun/jsc/webkit-upgrade-3722912f.test.ts       2      5      0
```

</details>

<!-- robobun:evidence:end -->

---------

Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
robobun added a commit that referenced this pull request Aug 10, 2026
…gger/cache/completion switches

#73 added SourceProviderSourceType::BunTranspiledModule so Bun
can attach a pre-computed module record to a SourceProvider and skip JSC's
analyze-phase re-parse. Three sites that branch on the provider's sourceType()
still match only ::Module, so a BunTranspiledModule provider falls through:

- debugger/DebuggerParseData.cpp gatherDebuggerParseDataForSource: falls to the
  default:false arm, Debugger::resolveBreakpoint gets empty pause positions,
  and Debugger.setBreakpoint replies "Could not resolve breakpoint" for every
  line in the script.
- runtime/CachedTypes.cpp CachedSourceProvider encode/decode: would
  RELEASE_ASSERT_NOT_REACHED() if JSC bytecode caching encoded one of these
  providers.
- runtime/Completion.cpp getSourceType: returns ScriptFetchParameters::Type::None
  instead of JavaScript.

(InspectorDebuggerAgent::didParseSource, the fourth site in the original
version of this change, is covered since the upstream merge in #383 by
scriptTypeForScript.)

Currently reachable in Bun via `bun test --isolate` and `--compile` standalone
executables; oven-sh/bun#35605 routes every runtime ESM provider through this
type and is blocked on this.

Each site adds a BUN_JSC_ADDITIONS-gated case that falls through to the Module
arm.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.