Incremental build wouldn't consider things like `OptimizationPreference` property changing as a thing that should trigger a rebuild. I feel like this is more a MSBuild bug, but it has been like this for a long time.

Turns out we can use `WriteIlcRspFileForCompilation` target as a sentinel:

* Run the target always. We only actually write out the file if it's different (`WriteOnlyWhenDifferent` is already set to `true`).
* ILC execution already specifies the RSP as one of its inputs. So if the RSP is more recent than the output, it will trigger a build.

Fixes #88725.

Co-authored-by: NKatelyn Gadd <kg@luminance.org>
Co-authored-by: NLarry Ewing <lewing@microsoft.com>

* Update dependencies from https://github.com/dotnet/roslyn build 20230802.9

Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
 From Version 4.8.0-1.23402.8 -> To Version 4.8.0-1.23402.9

* Update dependencies from https://github.com/dotnet/roslyn build 20230802.10

Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
 From Version 4.8.0-1.23402.8 -> To Version 4.8.0-1.23402.10

* Update dependencies from https://github.com/dotnet/roslyn build 20230803.1

Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
 From Version 4.8.0-1.23402.8 -> To Version 4.8.0-1.23403.1

---------
Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>

* [wasm] Update WBT AppTests to use the common infrastructure for building, and running apps

* [wasm] WasmApp.targets - add $(WasmAssemblyExtension)

* AppTests should respect UseWebcil

* Fix tests to work for no-webcil case

* Generate compact unwind sequence for methods with frame and no saved registers

* Store prolog length in LSDA data when compact unwinding is used

* Remove prolog length, detect prologs from instruction stream

* Update comments

* Tie llvm to emsdk's version
This avoids incoherency that will cause the .NET build drop to be too large to fit on the staging pipeline disk.

* Update to latest emsdk flow

---------
Co-authored-by: NLarry Ewing <lewing@microsoft.com>

See https://github.com/dotnet/runtime/issues/89921 for the test failures.

Additionally, the Android build was failing while compiling an AOT test due to this error:

```
/__w/1/s/artifacts/bin/Android.Device_Emulator.Aot.Test/Release/net8.0/android-x64/AppBundle/modules.c:51:6: error: no previous prototype for function 'register_aot_modules' [-Werror,-Wmissing-prototypes]
void register_aot_modules ()
     ^
```

This is because we turned on `-Werror=missing-prototypes` in https://github.com/dotnet/runtime/pull/89197 but the MonoAOTCompiler didn't emit a function prototype in modules.c

Fixes https://github.com/dotnet/runtime/issues/89566

GC thread doing gcdump will dump the EventPipe events after world
has restarted but before releasing GC lock. There was one case
where we logged a bulk type during that face where a type didn't
have its finalizer data initialized and at the same time main thread
running interpreter held loader lock and tried to acquire GC lock,
that triggers a deadlock since the GC thread (still holding the GC lock)
would trigger logic to initialize the finalizer data, but that in turn
requires the GC lock.

Fix delays the fire of GC dump events until after we completed GC.
All GC dump events have been cached into a temp file and will be
written into EventPipe, the only potential issue with this is that
we keep vtable pointers in cache that will be resolved when emitting
EventPipe event, after releasing GC lock, but since we currently
won't unload vtables, that is not an issue, but needs to be addressed
if/when we implement ability to unload vtables. We would then need
to root the vtables while stored in temporary cache.

Commit also enable GC dump test on Mono platforms.

* Fix handling ThreadAbortException at the end of catch

After the catch handler for ThreadAbortException exits, the
exception needs to be rethrown immediatelly. This change
makes that work.

* arm64 stuff

* Enable arm

* Fix macOS x64 and enable one ControlledExecutionTest path for Unix

* Update src/coreclr/vm/amd64/redirectedhandledjitcase.S
Co-authored-by: NAdeel Mujahid <3840695+am11@users.noreply.github.com>

* Fix macOS x64

The non-local label was preventing unwinding through \stub\()_RspAligned

---------
Co-authored-by: NAdeel Mujahid <3840695+am11@users.noreply.github.com>

These functions use memcmp that can internally be optimized
by vectorization. It can thus read more bytes than necessary
for the purpose of this function. In edge cases it can end up
reading from unmapped memory and crashing with access violation.

The fix is to compare the data byte by byte and early out as soon
as a mismatch is found.

Fixes #88992. Explicit specify callback calling convention on Windows x86.

CoreCLR runtime build uses stdcall while Mono uses cdecl (default). If a native runtime function takes a reverse pivnoke callback, this difference will cause issues for Mono on Windows x86 since reverse pinvoke callbacks have stdcall on Windows x86, but callbacks have been prototyped to use the calling convention setup in build.

Since CoreCLR already uses stdcall on Windows x86 this change won't change anything, while on Mono it will fixes the crashes observed in #8892.

- Mono defines HOST_WIN32, not CLR_CMAKE_HOST_WIN32
- Follow same pattern from src\native\eventpipe\CMakeLists.txt

* [wasm] Fix condition to use latest chrome for testing
* [wasm] CI: Trigger WBT on changes in ProvisioningVersions.props

* [wasm] runtime: Fix creating the stack trace for a ManagedError

With the latest chrome (`115.*`) the following code in
`runtime/marshal.ts` fails because `this.superStack.value` is no longer
available:

```js
    getSuperStack() {
        if (this.superStack) {
            return this.superStack.value;
        }
        return super.stack; // this works on FF
    }
```

This causes the final error to not have the original managed error
message, and also have a `"undefined"` at the end of the string.

Truncated error missing the native part of the stack, and the message:
```
   at System.Runtime.InteropServices.JavaScript.Tests.JavaScriptTestHelper.ThrowFromJSExport(String message)
   at System.Runtime.InteropServices.JavaScript.Tests.JavaScriptTestHelper.__Wrapper_ThrowFromJSExport_271731536(JSMarshalerArgument* __arguments_buffer)
undefined
```

With the fix:
```
   at System.Runtime.InteropServices.JavaScript.Tests.JavaScriptTestHelper.ThrowFromJSExport(String message)
   at System.Runtime.InteropServices.JavaScript.Tests.JavaScriptTestHelper.__Wrapper_ThrowFromJSExport_817705034(JSMarshalerArgument* __arguments_buffer)
Error: -t-e-s-t-
    at sr (http://127.0.0.1:60345/_framework/dotnet.runtime.js:3:33284)
    at Br (http://127.0.0.1:60345/_framework/dotnet.runtime.js:3:42679)
    at http://127.0.0.1:60345/_framework/dotnet.runtime.js:3:40825
    at Module.catch1stack (http://127.0.0.1:60345/JavaScriptTestHelper.mjs:132:9)
    at http://127.0.0.1:60345/_framework/dotnet.runtime.js:3:36627
    at mr (http://127.0.0.1:60345/_framework/dotnet.runtime.js:3:37821)
    at do_icall (http://127.0.0.1:60345/_framework/dotnet.native.wasm:wasm-function[221]:0x19711)
    at do_icall_wrapper (http://127.0.0.1:60345/_framework/dotnet.native.wasm:wasm-function[108]:0x157bc)
    at mono_interp_exec_method (http://127.0.0.1:60345/_framework/dotnet.native.wasm:wasm-function[101]:0x9c92)
    at interp_runtime_invoke (http://127.0.0.1:60345/_framework/dotnet.native.wasm:wasm-function[141]:0x16cd7)
```

Thanks to @kg for the fix.

* We don't throw PNSE for unguarded HW intrinsic use. We could translate illegal instruction to PNSE but it feels too big of a hammer. We could do something more sophisticated such as disassemble the instruction and compare with "known problematic" ones. But it doesn't feel worth the effort.
* Non-zero lower bounds are not supported.
* Corner case weird enums like `enum Foo : double { }` are not supported.

Noticed this because lines were off with `--ildump`.

* Run the tests in debug mode to get more information.

* Adding log messages

* Fix build

* Adding more messages

* Adding more logs

* Removing task.delay.

* Reducing maxParallelDownloads

* Adding more messages

* Fix compilation

* Trying to fix CI.

* Setting the maxParallelDownloads to 10.

This was found in Pri0 testing. The LDTOKEN data structures didn't have enough information for generic method instantiation arguments. The Pri0 test was around generic virtual methods, but this is a more general issue so adding an extra test for that.

I'm also changing how `NodeFactory.ReflectedMethod` works - it will now require being called with canonical things explicitly (and asserts if it isn't) instead of canonicalizing behind the scenes. This was obscuring the problem.

* Update dependencies from https://github.com/dotnet/roslyn build 20230802.2

Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
 From Version 4.8.0-1.23401.10 -> To Version 4.8.0-1.23402.2

* Update dependencies from https://github.com/dotnet/roslyn build 20230802.3

Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
 From Version 4.8.0-1.23401.10 -> To Version 4.8.0-1.23402.3

* Update dependencies from https://github.com/dotnet/roslyn build 20230802.4

Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
 From Version 4.8.0-1.23401.10 -> To Version 4.8.0-1.23402.4

* Update dependencies from https://github.com/dotnet/roslyn build 20230802.8

Microsoft.CodeAnalysis , Microsoft.CodeAnalysis.CSharp , Microsoft.Net.Compilers.Toolset
 From Version 4.8.0-1.23401.10 -> To Version 4.8.0-1.23402.8

---------
Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>

[ComInterfaceGenerator] Guaranteed unmarshal before managed call in unmanaged to managed stubs (#89583)

Move guaranteed unmarshalling to the correct place in unmanaged to managed stucs

* Create a temporary directory for crossgen2 SuperPMI collections to hold a JIT binary

* Print superpmi jit path for logging purposes

* Added fallback

* Ignore cleanup errors

* Do not cleanup in handling an the exception

* Using existing temp location

* Feedback

* Remove extra logging

Fixes #89767

* Add `CompressSymbols` opt-out option in NativeAOT

* Report what went wrong in dwarfdump test

* Update src/tests/nativeaot/SmokeTests/DwarfDump/Program.cs
Co-authored-by: NJan Kotas <jkotas@microsoft.com>

* Updating the Roslyn toolset compiler to v4.8.0-1.23378.8

* Changing signatures from `in` to `ref readonly`

* Changing signatures from `ref` to `in`

* Changing signatures from `ref` to `ref readonly`

* Fix case missed for Unix build

* Ensure an xml comment works for .NET or .NET Framework

* Ensure tests properly handle `ref readonly` changes

Adds a number of test interfaces to ComInterfaceGenerator tests.

* Update prepare-artifacts.proj

Don't pass AssetsTemporaryDirectory

* Remove TmpWorkingDir

And PublishingVersion, per PR feedback

[main] Update dependencies from dotnet/runtime dotnet/llvm-project dotnet/arcade dotnet/xharness dotnet/runtime-assets dotnet/emsdk (#89810)
Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>