* remove MONO_ASMCTX_REFONLY

* Also remove mono_image_open_from_data_internal refonly arg

* Remove console-null console-win32 dead icalls

* Remove shadow copying remains

No more MonoDomain shadow_serial field

* No MonoDomain socket assembly fields

* remove MonoRuntimeInfo arg from mono_assembly_load_corlib

* delete useless allocations

* Update dependencies from https://github.com/dotnet/arcade build 20210205.5

Microsoft.DotNet.XUnitExtensions , Microsoft.DotNet.VersionTools.Tasks , Microsoft.DotNet.Build.Tasks.TargetFramework.Sdk , Microsoft.DotNet.Build.Tasks.Packaging , Microsoft.DotNet.Build.Tasks.Installers , Microsoft.DotNet.Build.Tasks.Feed , Microsoft.DotNet.Build.Tasks.Archives , Microsoft.DotNet.Arcade.Sdk , Microsoft.DotNet.ApiCompat , Microsoft.DotNet.CodeAnalysis , Microsoft.DotNet.XUnitConsoleRunner , Microsoft.DotNet.GenFacades , Microsoft.DotNet.GenAPI , Microsoft.DotNet.RemoteExecutor , Microsoft.DotNet.Helix.Sdk , Microsoft.DotNet.SharedFramework.Sdk
 From Version 6.0.0-beta.21101.7 -> To Version 6.0.0-beta.21105.5

* Fix xunit analyzer warning in SharedArrayBufferTests.cs
Co-authored-by: Ndotnet-maestro[bot] <dotnet-maestro[bot]@users.noreply.github.com>
Co-authored-by: NStephen Toub <stoub@microsoft.com>

* Cleanup the assembly loading code.

* Remove the 'problematic' assembly code.
Co-authored-by: NAleksey Kliger (λgeek) <alklig@microsoft.com>

* Remove the remainder of the appdomain unload code.

* Restore public API functions.

* Add crossgen2 automated SuperPMI collection of Core_Root libraries

Linux is excluded; there are uninvestigated crashes doing crossgen2 Linux collections.

I added logic to copy the runtime ".dotnet" directory to Helix machines to use as the
driver for crossgen2, but that doesn't work for cross-platform scenarios (e.g., Windows
x64 .dotnet, x86 Core_Root), so it's commented out for now and the same-Core_Root corerun.exe
is used as the crossgen2 driver.

Add `-collection_type` argument to superpmi-setup.py

Allow copying of .json files to Helix; needed by crossgen2.

Changed superpmi pipeline to run with DEBUG level logging from superpmi.py so it's
easier to see what happened when looking at Helix logs (without having to find and separately
download the superpmi.log file).

Changed superpmi.py to write stdout/stderr files to log on failure, so we can see why
crossgen2 failed.

Fix creation of log file directories when `-log_file` argument specified

Excluded many more native files from superpmi collection setup.

Fixes #47541

Extra:
1. If `-temp_dir` is specified to superpmi.py, normalize it to
an absolute path.
2. Add a superpmi.py collect `-exclude` option to avoid collecting
assemblies known to fail (and cause pop-up failures: see #47552).
(This isn't used by the automated collection, but is generally useful.)
3. Stop logging too-verbose message about CORINFO_SIG_INST handle
array illegal values, since it's "by design" for crossgen2
4. Force lower-casing of JIT-EE version string, for consistency.

Remove unused fields/functions.

* Remove unused fields from ref.emit classes.

* Remove MonoDynamicImage->run/save, they are always TRUE/FALSE.

* Remove more code.

- Enables MemoryExtensions.AsSpan(T[], Range) to be trimmed in a default Blazor wasm app.
- Use of span.Slice(int) is more efficient than span[int..]
- Use of AsSpan(int, int) is more efficient than AsSpan(Range)

We'd removed this almost a year ago, but it came back almost immediately and accidentally as a result of two PRs that raced with each other.

- Use TryEnsureSufficientExecutionStack rather than EnsureSufficientExecutionStack, allowing the latter to be trimmed in a default Blazor wasm app.
- Use ValueTuple instead of Tuple, allowing the latter to be trimmed in a default Blazor wasm app
- Clean up awaiter usage in RunOnEmptyStackCore

The former just checks the input, and if it's null, calls the latter.

All of our call sites to it pass in a const StringComparison; GetHashCode will then turn around and decide which property on StringComparer to use, and then call GetHashCode on the relevant instance... we may as well just pick the instance directly, which not only saves the lookup and enables devirtualization, it also enables better trimming, as string.GetHashCode(StringComparison) ends up rooting all of the StringComparer properties in case you pass in the appropriate StringComparison.

For a rare path, GetCanonicalPath is removing one character from a string and then inserting  a replacement character.  Thanks to span, we can replace the two string allocations with a single char[] allocation and avoid these Remove/Insert calls, which are the only ones keeping string.Remove and string.Insert from being trimmed out of a default Blazor wasm app.

Fixes NameHeaderValue validation to be in line with RFC.

Resolves #1504

* [reflection] Remove sre-save and DISABLE_REFLECTION_EMIT_SAVE

* remove old sre-save entrypoint declarations too

* mono: remove ENABLE_NETCORE and associated framework code

* Remove g_assert_not_netcore

* No more locales.h

* No more filewatcher.h

* Fix Windows build

* Remove g_assert_netcore

I don't see anything calling this

* Add now-needed include

* Set diagnostic severity

* Fix violations in System.Private.Xml

* Revert .editorconfig

* Remove a couple LINQ usages in Microsoft.Extensions

* Respond to PR feedback.

Change ParameterDefaultValue to use GetUninitializedObject instead of Activator.CreateInstance (#47722)

* Change ParameterDefaultValue to use GetUninitializedObject instead of Activator.CreateInstance

Activator.CreateInstance has the ability to call a parameterless constructor (if one is defined) on a value type. This is incorrect for ParameterDefaultValue. It is more correct to call GetUninitializedObject, which is the same as using `default(T)`.

* Add test for struct with default constructor

* Resolve ILLink warnings in System.Linq.Expressions (Round 2)

Contributes to #45623

* Suppress ToString warning in ExpressionStringBuilder

Fix typo in browser sample so it doesn't log an error

* Remove more unused files from netcore/.

* Remove unused gen-xunit-runner.

The previous logic would not call `mono_gc_thread_detach` for a number of scenarios. One example:
```
start_wrapper
- mono_thread_info_attach - thread is now live in mono threads layer
- start_wrapper_internal
-- mono_thread_attach_internal - mono attached to vm thread layer. GC handle is set via mono_thread_info_set_internal_thread_gchandle
-- mono_thread_detach_internal - detached from vm thread layer. GC handle cleared via call to mono_thread_info_unset_internal_thread_gchandle
- mono_thread_info_exit
-- mono_thread_info_detach
--- unregister_thread
---- thread_detach callback - checks if gc handle is valid and returns if not via mono_thread_info_try_get_internal_thread_gchandle. We've already cleared above so we never call mono_gc_thread_detach.
```

This change ensures `mono_gc_thread_detach` is always called even the GC handle for the thread has already been cleared.

There is a problem in the
ClassLoader::ValidateMethodsWithCovariantReturnTypes that results in
failed verification of valid override in case the return type of the
method being overriden is generic in canonical form and it is defined
in an ancestor class that is not the parent.

The problem is that we attempt to use instantiation of the parent class
instead of the ancestor class that contains definition of the method
being overriden.

This change fixes it by locating the proper ancestor MethodTable and
using it.

* Limit SNI to 63 chars

* Split SNI to more DNS labels

We always need to run the profile data phase so that jit stress can inject
random profile counts if it so chooses.

Also, clean up a few dumping nits -- don't dump the profile query status until
we get around to trying to incorporate counts; summarize schema records before
asserting that we must have block counts, etc.

Closes #47839