* Use "read" to fetch misaligned 64bit values

* remove read_direct

* rename read() -> read_byte()

If we have a non-inline candidate call with generics context, save
the context so it's available for late devirtualization.

Fixes a missing devirtualization reported in #63283.

I am deliberately leaving `LateDevirtualizationInfo` more general than
necessary as it may serve as a jumping-off point for enabling late
inlining.

Enable llvm for arm perf runs.

[mono] Make some icalls pass/return object references using ObjectHandleOnStack/QCallTypeHandle (#62141)

* [mono] Make some icalls return objects using an extra ObjectHandleOnStack argument.

* Fix the passing of scalar vtypes on wasm.

* Convert RuntimeTypeHandle icalls to receive a QCallTypeHandle.

Also do some other cleanups:
* Convert some icalls which don't receive/return objects any more and don't
  return an error to NOHANDLES.
* Implement IsGenericVariable in managed code.
* Add a separate GetMetadataToken icall to avoid the REUSE_WRAPPER stuff.
* Sync the implementation of IsTypeDefinition with coreclr.

* Convert some RuntimeType icalls to use QCallTypeHandles/ObjectHandleOnStack.

* Fix RuntimeType.GetPacking () to work with dynamic types.

* Convert internal_from_name to use QCallTypeHandles/ObjectHandleOnStack.

* Convert more icalls.

* Update src/libraries/System.Private.CoreLib/src/System/Runtime/CompilerServices/QCallHandles.cs
Co-authored-by: NAleksey Kliger (λgeek) <akliger@gmail.com>

* Remove unused argument from RuntimeTypeHandle:internal_from_name ().

* Add support for FRAME_TYPE_IL_STATE to the metadata stack walker code.

* Update coding style in QCallTypeHandle:.ctor ().

* Revert "Convert more icalls."

* Convert some Delegate icalls.

* Convert some Enum icalls.

* Convert some Marshal icalls.

* Avoid creating RuntimeType objects while AOTing.

* Convert some RuntimeType icalls.
Co-authored-by: NAleksey Kliger (λgeek) <akliger@gmail.com>

* Fix bugs with generation for ctor param default values

* Address review feedback

* Address feedback

Co-authored-by: NTomas <trylek@microsoft.com>

* [mono] Add back HAVE_SYS_ZLIB support

In particular this re-enables embedded PDB support on iOS and Android

* [mono] Reorganize zlib dependency logic

Instead of deciding whether features are enabled based on the presence or
absense of zlib, add switches to control features (embedded PDB support, and
compressed log profiler output) and based on that require zlib.

On some platforms we use the in-tree copy of zlib from
src/native/external/zlib, otherwise we want to link against the system zlib.

Previously, if zlib was missing the build would quietly succeed, but the
feature support would be missing.  Now, if the feature isn't disabled but zlib
is missing, we fail the build.

If both features are disabled, don't depend on zlib at all.

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

* Style nits

- reverse if() logic to avoid negation

- use cmakedefine without a value in config.h.in, same as CoreCLR
Co-authored-by: NAdeel Mujahid <3840695+am11@users.noreply.github.com>

A FriendlyName in Windows' length is a DWORD, an unsigned integer. We
however marshal it as a signed integer, so a friendly name with a
length greater than int.MaxValue would wrap around to negative. This
in turn would be "below" the stackalloc threshold, and attempt to stackalloc
a negative value. stackalloc treats this value as unsigned, so it results
in allocating too much on the stack.

When Microsoft.Extensions.Configuration.UserSecrets is referenced transitively, for example through Extensions.Hosting, the UserSecretsIdAttribute isn't getting added to the built assembly. This is because the logic is contained in a `build` folder which is excluded by the dependency from Extensions.Hosting.

The fix is to move the logic to `buildTransitive`, so it gets picked up by NuGet. One wrinkle is that the logic now conflicts with the netstandard compatibility logic added by the library infrastructure. In order to make progress, disable that logic for this package and rely on its dependencies to raise the compat error.

Contributes to #63246

In XUnitWrapperGenerator we need to initialize the stopwatch
variable using Stopwatch.StartNew(), not just new Stopwatch(),
otherwise the stopwatch never starts to run, that's why all the
test case times showed as zero. (Alternatively we could use
new Stopwatch() followed by stopwatch.Start() but I believe that
Stopwatch.StartNew() is generally considered to be more
performant.)

After applying this fix, some short test times started reporting
their running time in scientific notation (e.g. 5.5e-5) so I
modified the formatting to F6 - in the Methodical suite I see
a test running for 55 microseconds.

Thanks

Tomas

This caused the disabled ArrayTests to run on windows, causing the builds to fail. With this change, the tests should correctly get skipped on windows.

* Fixing how we test and the implementation after #62627

* Passing null to name.

* Use ishst instead of ish

* Do not contain address of volatile fields

* Do not contain address only for Arm64

* Remove ishst

Move `gtStubCallStubAddr` out of the union it shares with class probe,
inline, and guarded devirt information, so that we no longer need to
save and restore the stub address when we want to probe, guardedly
devirtualize, or inline calls.

Delete all the code that did saving and restoring, and the field in the
inline info struct that held the saved copy.

* Throw a HttpListenerException if the session is null

* Add similar check to BeginGetContext

* Address PR feedback

* Remove trailing spaces

The wrong module was being passed to the remote unwinder because the load bias for shared modules
was being calculated incorrectly.

This method was not taking into account the fact that string ctors
don't have dummy this argument after a recent change. Stack walking
in cases where prestub for the ctor was invoked resulted in a wrong
unwinding of ESP for ctors with more than one argument.

This change fixes it by removing "has this" flag from the signature
before getting the stack arguments size.

This has been unused ever since crossgenning framework moved to build.proj.

* Remove unnecessary duplicate check in DeflateManagedStream ctor

* Remove duplicate null check in JsonSerializer.Deserialize

* Delete duplicate check in XmlSerializationWriterILGen.WriteQualifiedNameElement

* Delete duplicate rule check in XmlTreeGen.AutoGenerated

Exact same pattern was previously changed in OleDbCommandBuilder.

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

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

Set BROWSER_PATH variable only when HELIX_CORRELATION_PAYLOAD variable is set.

Fixes change in the https://github.com/dotnet/runtime/pull/62779.

This is now shared unit test project that is not AOT/JIT specific.

The recently added MarshalUtilsTests.cs created a dependency on these.

It's after validation on the HttpResponseStream corresponding methods, and having it before the validation means it'll null ref if a null buffer is erroneously provided.

Reverts #62549 as the configuration parameter is needed to let xharness know where the iOS/tvOS/MacCatalyst test app is.

For example, a debug test app will be under Debug-iphoneos and a release one will be under Release-iphoneos. Xharness won't be able to tell unless provided the right info.