@schellap - Can you see why the Nuget feed does not have the latest package from \jkotas1\drops\corert.external\gkhanna? This package has a CPPSDk subfolder while the one that gets restored does not.

using System;

namespace WebApplication106
{
    public class Startup
    {
        public static void Main(string[] args)
        {
            var identity = new Func<int, int>(Identity);
        }

        private static int Identity(int arg)
        {
            return arg;
        }
    }
}

I get compilation errors in the generated Cpp

McgIntrinsic attribute should be renamed because of it is not MCG-specific. Also, it should be per-method attribute, not per-type attribute.

See TODO https://github.com/dotnet/corert/blob/master/src/Common/src/TypeSystem/IL/ILProvider.cs#L27

getenv and putenv do not play well together.

We need to figure out the right thing to do.

In the wiki and in the build.cmd it refers to the corert.external folder which I do not have locally. Should I manually create it? What is the purpose of that folder?

The generated Address methods on MDArrays should have a type check (for reference type arrays).

Provide actual implementation of getSystemVAmd64PassStructInRegisterDescriptor in CorInfoImpl.cs.

Paste this program into "repro" solution and compile it using ILToNative.

using System;

class Program
{
    class MyG<T>
    {
        override public string ToString()
        {
            return null;
        }
    }

    static void Main(string[] args)
    {
        new MyG<Program>().ToString();
    }
}

\src\System.Private.CoreLib\src\System\Environment.EnvironmentVariables.Unix.cs has a workaround to return null for LANG environment variable. This is to prevent System.Console from taking complicated code paths that we don't support yet.

The required feature is either interface calls or constrained calls (don't remember which one...).

We should enable the following:

1. Badges for nightly builds on the repo for all platforms
2. Signing the binaries before pushing them to the daily nuget feed.

We have a bunch of structs that are just a bundle of values with no logic - we are using private fields + properties in some places, and public readonly fields in other places.

Would be nice to use readonly fields everywhere we can.

        private static bool Init()
        {
            if (s_InvariantCultureInfo == null)
            {
                CultureInfo temp = new CultureInfo("", false);
                temp._isReadOnly = true;
                s_InvariantCultureInfo = temp;
            }

            s_userDefaultCulture = GetUserDefaultCulture();
            return true;
        }

Notice this will call "new CultureInfo(...)" which is defined as:

        internal CultureInfo(String name, bool useUserOverride)
        {
            if (name == null)
            {
                throw new ArgumentNullException("name",
                    SR.ArgumentNull_String);
            }

#if CORERT
            // CORERT-TODO CultureInfo
            m_cultureData = s_InvariantCultureInfo.m_cultureData;
            m_name = _sortName = _nonSortName = name;
            m_isInherited = false;
#else
            ...

The constructor is trying to access s_InvariantCultureInfo which is not yet initialized because we're in the process of initializing it.

To hit the bug, undo workaround #117 and compile hello world. Make sure you have the contents of pull request #125 or you'll hit a different bug.

When I try to use the System.Console in my ASP.NET sample I get this:

1>c:\corefxlab\src\system.buffers\src\system\buffers\bufferpool.cs(17193): error C2733: 'FormatMessageW': second C linkage of overloaded function not allowed
1>  c:\corefxlab\src\system.buffers\src\system\buffers\bufferpool.cs(7454): note: see declaration of 'FormatMessageW'
1>c:\corefxlab\src\system.buffers\src\system\buffers\bufferpool.cs(21051): error C2556: 'uint32_t GetLastError(void)': overloaded function differs only by return type from 'int32_t GetLastError(void)'
1>  c:\corefxlab\src\system.buffers\src\system\buffers\bufferpool.cs(17199): note: see declaration of 'GetLastError'
1>c:\corefxlab\src\system.buffers\src\system\buffers\bufferpool.cs(21051): error C2371: 'GetLastError': redefinition; different basic types

/cc @jkotas

Looks like the issue is with System.Buffers?

Turning off line numbers shows:

1>c:\dev\git\corert\src\iltonative\repronative\repro.cpp(30841): error C2733: 'FormatMessageW': second C linkage of overloaded function not allowed
1>  c:\dev\git\corert\src\iltonative\repronative\repro.cpp(21102): note: see declaration of 'FormatMessageW'
1>c:\dev\git\corert\src\iltonative\repronative\repro.cpp(34699): error C2556: 'uint32_t GetLastError(void)': overloaded function differs only by return type from 'int32_t GetLastError(void)'
1>  c:\dev\git\corert\src\iltonative\repronative\repro.cpp(30847): note: see declaration of 'GetLastError'
1>c:\dev\git\corert\src\iltonative\repronative\repro.cpp(34699): error C2371: 'GetLastError': redefinition; different basic types

It would help if the cpp codegen preserved (maybe in a comment) where it found the DLL exports. That would help identify the problem.

See #98

We should be able to use CLI command to support corss-compiling for different architectures on a given platform.

Just built CoreRT on OS X. Got error messages about installing an SDK of reference assemblies into the GAC.

/Users/rlander/git/corert/packages/Microsoft.Build.Mono.Debug.14.1.0.0-prerelease/lib/Microsoft.Common.CurrentVersion.targets(1090,5): warning MSB3644: The reference assemblies for framework ".NETPortable,Version=v4.5,Profile=Profile7" were not found. To resolve this, install the SDK or Targeting Pack for this framework version or retarget your application to a version of the framework for which you have the SDK or Targeting Pack installed. Note that assemblies will be resolved from the Global Assembly Cache (GAC) and will be used in place of reference assemblies. Therefore your assembly may not be correctly targeted for the framework you intend. [/Users/rlander/git/corert/src/ILToNative.DependencyAnalysisFramework/tests/ILToNative.DependencyAnalysisFramework.Tests.csproj]
/Users/rlander/git/corert/packages/Microsoft.Build.Mono.Debug.14.1.0.0-prerelease/lib/Microsoft.Common.CurrentVersion.targets(1090,5): warning MSB3644: The reference assemblies for framework ".NETPortable,Version=v4.5" were not found. To resolve this, install the SDK or Targeting Pack for this framework version or retarget your application to a version of the framework for which you have the SDK or Targeting Pack installed. Note that assemblies will be resolved from the Global Assembly Cache (GAC) and will be used in place of reference assemblies. Therefore your assembly may not be correctly targeted for the framework you intend. [/Users/rlander/git/corert/src/ILToNative.DependencyAnalysisFramework/tests/ILToNative.DependencyAnalysisFramework.Tests.csproj]
/Users/rlander/git/corert/packages/Microsoft.Build.Mono.Debug.14.1.0.0-prerelease/lib/Microsoft.Common.CurrentVersion.targets(1090,5): warning MSB3644: The reference assemblies for framework ".NETPortable,Version=v4.5,Profile=Profile7" were not found. To resolve this, install the SDK or Targeting Pack for this framework version or retarget your application to a version of the framework for which you have the SDK or Targeting Pack installed. Note that assemblies will be resolved from the Global Assembly Cache (GAC) and will be used in place of reference assemblies. Therefore your assembly may not be correctly targeted for the framework you intend. [/Users/rlander/git/corert/src/ILToNative.TypeSystem/tests/CoreTestAssembly/CoreTestAssembly.csproj]
/Users/rlander/git/corert/packages/Microsoft.Build.Mono.Debug.14.1.0.0-prerelease/lib/Microsoft.Common.CurrentVersion.targets(1090,5): warning MSB3644: The reference assemblies for framework ".NETPortable,Version=v4.5" were not found. To resolve this, install the SDK or Targeting Pack for this framework version or retarget your application to a version of the framework for which you have the SDK or Targeting Pack installed. Note that assemblies will be resolved from the Global Assembly Cache (GAC) and will be used in place of reference assemblies. Therefore your assembly may not be correctly targeted for the framework you intend. [/Users/rlander/git/corert/src/ILToNative.TypeSystem/tests/CoreTestAssembly/CoreTestAssembly.csproj]
CoreTestAssembly -> /Users/rlander/git/corert/bin/Product/OSX.x64.Debug/CoreTestAssembly.dll
/Users/rlander/git/corert/packages/Microsoft.Build.Mono.Debug.14.1.0.0-prerelease/lib/Microsoft.Common.CurrentVersion.targets(1090,5): warning MSB3644: The reference assemblies for framework ".NETPortable,Version=v4.5,Profile=Profile7" were not found. To resolve this, install the SDK or Targeting Pack for this framework version or retarget your application to a version of the framework for which you have the SDK or Targeting Pack installed. Note that assemblies will be resolved from the Global Assembly Cache (GAC) and will be used in place of reference assemblies. Therefore your assembly may not be correctly targeted for the framework you intend. [/Users/rlander/git/corert/src/ILToNative.TypeSystem/tests/TypeSystem.Tests.csproj]
/Users/rlander/git/corert/packages/Microsoft.Build.Mono.Debug.14.1.0.0-prerelease/lib/Microsoft.Common.CurrentVersion.targets(1090,5): warning MSB3644: The reference assemblies for framework ".NETPortable,Version=v4.5" were not found. To resolve this, install the SDK or Targeting Pack for this framework version or retarget your application to a version of the framework for which you have the SDK or Targeting Pack installed. Note that assemblies will be resolved from the Global Assembly Cache (GAC) and will be used in place of reference assemblies. Therefore your assembly may not be correctly targeted for the framework you intend. [/Users/rlander/git/corert/src/ILToNative.TypeSystem/tests/TypeSystem.Tests.csproj]

Compile "hello world" using JIT codegen:

at ILToNative.DependencyAnalysisFramework.DependencyAnalyzer`2.ComputeMarkedNodes() in C:\corert\src\ILToNative.DependencyAnalysisFramework\src\DependencyAnalyzer.cs:line 254

Good warnings and error reporting are of great help to tool developers or general developers alike.

The warning system should not be a burden in environments where warnings are undesired/unactionable - it should get out of the way without impacting perf. At the same time we want the warning system be flexible enough to enable PEVerify-like scenarios (dotnet/coreclr/#295) for the type system in the future.

The error system should be based on standardized exceptions that provide context with programmatic access (e.g. having an object-typed Pertinent property that points to a type system object associated with the exception).

Mac build fails with the following error:

In file included from /Users/gkhanna/Github/gkhanna79/corert/src/Native/Runtime/EHHelpers.cpp:26:
/Users/gkhanna/Github/gkhanna79/corert/src/Native/Runtime/stressLog.h:377:9: error: unused typedef 'C_ASSERT' [-Werror,-Wunused-local-typedef]

This happens on Apple Clang 7+ (part of XCode 7+) and is no repro on Apple Clang 6.1 (part of XCode 6.1+).

When computing the size of the environment variable we need to guard against the very rare case where the size of the envvar is > 2GB. Since in manged code we cannot have a larger string than 2GB we need to handle that case.

#104 (it was not perfect before, but #104 made it significantly worse) regressed where the bits are dropped by VS projects. E.g

• Building in VS drops System.Private.CoreLib.dll to C:\corert\bin\System.Private.CoreLib.dll and C:\corert\bin\System.Private.CoreLib\System.Private.CoreLib.dll
• build.cmd drops it to C:\corert\bin\Product\Windows_NT.x64.Debug\System.Private.CoreLib.dll.

We need to get these back in sync to maintain development loop efficiency.

Implement C++ JIT interface wrapper that will catch the managed exception and re-throw it as unmanaged exception so that we can throw managed exceptions from JitInterface and not crash afterwards on Linux.

The compiler needs a warning/error infrastructure that provides experience developers in the 21st century got used to.

This includes:

• Errors that can unambiguously distinguish between bad inputs and internal compiler errors. For bad inputs, they should be actionable and provide good context.
• Warnings that have warning levels associated with them
• Warnings that can be selectively disabled
• Errors and warnings that are testable. We have seen diagnostics silently regressing far too often in the past

This should work together with the error/warning infrastructure in the type system (#71).

GetEnvironmentVariable and ExpandEnvironment are implemented like this:

public static String GetEnvironmentVariable(String variable)
{
    if (variable == null)
        throw new ArgumentNullException("variable");

    // Environment variable accessors are not approved modern API.
    // Behave as if the variable was not found in this case.
    return null;
}

We need real implementations.

This will allow managed components to be built by default as well.

I'm not sure if this is my setup but I'm unable to run ILToNative.exe from the command line (although I'm pretty sure it worked last week). It fails because it wants to load the following assembly:

Unhandled Exception: System.IO.FileNotFoundException: Could not load file or assembly 'System.Runtime.InteropServices.RuntimeInformation, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a' or one of its dependencies. The system cannot find the file specified.
at ILToNative.Program.Run(String[] args)
at ILToNative.Program.Main(String[] args) in C:\work\Microsoft\corert\src\ILToNative\src\Program.cs:line 180

I can run it just fine using CoreCLR or from VisualStudio.

Likely this change d56f145#diff-7d246f2f1ad0f9ce7a6ffcf4005175bbR167

Similar to Windows, we should be able to run CoreCLR tests on Linux and Mac.

1. It should be used by the test infrastructure
2. There should be a well defined flow for setting it up to perform compilation

It'll help if we can grab all of the bits from a single place until we have aot implementations available elsewhere (packages)

/cc @jkotas @gkhanna79

We should refactor the existing static/instance field layout type system tests to be easily extensible to other platform. Currently, the expected offsets are hardcoded to a specific number and the architecture is hardcoded to x64 (aka AMD64 or x86-64).

What we want is probably a data structure that has expected field offsets for each architecture and a driver that iterates over all architectures and for each architecture tests that the computed offsets match the expected values.

I didn't do it initially because the tests are a line-by-line port from the tests for the .NET Native NUTC compiler and we don't support anything but x64 right now anyway.

• Convert Cpp code gen to use dependency analysis
• Use only 1 Relocation structure
• Remove unused AsmWriter
• Generate target in jit interface as dependency analysis nodes directly instead of having translation step

Currently, InstantiatedType.GetMethods uninstantiates the type before doing signature comparisons with the signature passed in. This doesn't work for the case where the signature in question was instantiated since the instantiated type doesn't match the generic parameter. When fixing this, please remove the workaround in FindMatchingVirtualMethodOnTypeByNameAndSig in VirtualFunctionResolution.cs

Check https://github.com/dotnet/corert/blob/master/src/ILToNative.Compiler/src/Compiler/DependencyAnalysis/EETypeNode.cs on what is there today.

Rough list of sub-workitems:

• Various bits on EEType
• GCDescs
• Interface and interface dispatch structures
• GIDs (refactored; we don't want what's in ProjectN)

Similar to how we build and run tests, as part of the product build on Windows, we need to enable the same for Linux and Mac as well.

Revise things like uses of IsInstanceOfArray to make sure we don't use them with the assumption that the type being looked at is a vector.

See TODOs in #84

Need implementations in InstantiatedType.Interfaces.cs

Review/design/refactor interface between driver .exe and compiler .dll. See #51.

Currently, __managed_main is the entry point, but we want to add more known entry points for other MethodCode nodes.

Note: @davidwrighton suggested in pull #144 to raise an issue for now.

"Hello world" crashes during shutdown with debug assertion violation:

Expression: '!IsStateSet(TSF_Detached)`

File: c:\corert\src\native\runtime\thread.cpp, Line: 951

reproNativeCpp.exe!PalDebugBreak() Line 918   C++
reproNativeCpp.exe!Assert(const char * expr, const char * file, unsigned int line_num, const char * message) Line 97  C++
reproNativeCpp.exe!Thread::SetDetached() Line 952        C++
reproNativeCpp.exe!Thread::Destroy() Line 405  C++
reproNativeCpp.exe!ThreadStore::DetachCurrentThreadIfHomeFiber() Line 247  C++
reproNativeCpp.exe!FiberDetach(void * lpFlsData) Line 267          C++

/src/Native/Runtime/amd64/WriteBarriers.s is stubbed out because I really didn't feel like translating WriteBarriers.asm to AT&T assembler.

Make name mangling to be robust against collisions with built-in C++ identifiers or our internal identifiers (see #31)