Revert "Make solution options global"

[RikkiGibson]

#59168 is suspected of introducing RPS regressions.

Last known good: https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/379782
First known bad: https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/379787
Revert validation: https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/379865

Regressed counters:

• BuildInfoVS64.Test - RegressedBeyondThreshold / 0001.BuildInfo / Build_Ngen_InvalidAssemblyCount / Regressed: 3.00 Count (50.00%)
• WebToolsVS64.Debugging - RegressedBeyondThreshold / 0300.Stop Debugging / VM_AdjustedImagesInMemory_Total_devenv / Regressed: 3.80 Count (0.75%)
• WebToolsVS64.SolutionManagement - RegressedBeyondThreshold / 0100.Open Solution / IO_AdjustedFileReads_devenv / Regressed: 35.40 Count (9.41%)
• WebToolsVS64.SolutionManagement - RegressedBeyondThreshold / 0100.Open Solution / RefSet_Image_Delta_devenv / Regressed: 5296128.00 Bytes (3.81%)
• WebToolsVS64.SolutionManagement - RegressedBeyondThreshold / 0300.Close Solution / VM_AdjustedImagesInMemory_Total_devenv / Regressed: 4.00 Count (1.01%)
• WebToolsVS64.TypingInRazorView - RegressedBeyondThreshold / 0100.Typing in Razor View / VM_AdjustedImagesInMemory_Total_devenv / Regressed: 4.60 Count (1.09%)

Please advise whether the failures are plausibly connected with this PR, and if so, whether we should revert, or take a follow-up PR to fix, or seek an exception. @genlu @tmat @dibarbet.

[genlu]

BuildInfoVS64.Test - RegressedBeyondThreshold / 0001.BuildInfo / Build_Ngen_InvalidAssemblyCount / Regressed: 3.00 Count (50.00%)

The ngen error is definitely caused by this change:

02/14/2022 10:44:44.177 [4060]: 1>Warning: System.TypeLoadException: Could not load type 'Microsoft.CodeAnalysis.ExternalAccess.Razor.IRazorDocumentExcerptService' from assembly 'Microsoft.CodeAnalysis.ExternalAccess.Razor, Version=4.2.0.0, Culture=neutral, PublicKeyToken=31bf3856ad364e35'. while resolving 0x200000d - Microsoft.CodeAnalysis.Razor.DocumentExcerptServiceBase.

[genlu]

Those assembly load regression seem to be caused by ProjectSystem reporting erros. For example:

Name
 Image Load wer.dll (C:\Windows\System32\)
+ ntoskrnl!?
 + ntdll!NtMapViewOfSection
  + ntdll!LdrpMinimalMapModule
   + ntdll!LdrpMapDllWithSectionHandle
    + ntdll!LdrpMapDllNtFileName
     + ntdll!LdrpMapDllSearchPath
      + ntdll!LdrpProcessWork
       + ntdll!LdrpLoadDllInternal
        + ntdll!LdrpLoadDll
         + ntdll!LdrLoadDll
          + kernelbase!LoadLibraryExW
           + clr!CLRLoadLibraryExWorker
            + clr!CLRLoadLibraryEx
             + clr!LocalLoadLibraryHelper
              + clr!??NDirect::LoadLibraryModule
               + clr!NDirect::NDirectLink
                + clr!NDirect::GetStubForILStub
                 + clr!GetStubForInteropMethod
                  + clr!??MethodDesc::DoPrestub
                   + clr!PreStubWorker
                    + clr!ThePreStub
                     + microsoft.visualstudio.projectsystem.implementation.ni!HResult.Verify
                      + microsoft.visualstudio.projectsystem.implementation.ni!WatsonErrorReport.Prepare
                       + microsoft.visualstudio.projectsystem.implementation.ni!WatsonErrorReport.CreatePreparedNonFatalReport
                        + microsoft.visualstudio.projectsystem.implementation.ni!ProjectErrorReporting.SubmitErrorReport
                         + microsoft.visualstudio.projectsystem.implementation.ni!Microsoft.VisualStudio.ProjectSystem.ProjectErrorReporting+<>c__DisplayClass10_0.<SubmitErrorReport>b__1(System.Exception)
                          + guardmethodassembly!GuardMethodClass.GuardMethod(class System.Func`1,class System.Func`2,class System.Func`2)
                           + clr!ExceptionTracker::CallHandler
                            + clr!ExceptionTracker::CallFilterFunclet
                             + clr!ExceptionTracker::ProcessManagedCallFrame
                              + clr!ExceptionTracker::ProcessOSExceptionNotification
                               + clr!ProcessCLRException
                                + ntdll!RtlpExecuteHandlerForException
                                 + ntdll!RtlDispatchException
                                  + ntdll!RtlRaiseException
                                   + kernelbase!RaiseException
                                    + clr!RaiseTheExceptionInternalOnly
                                     + clr!IL_Throw
                                      + mscorlib.ni!ExceptionDispatchInfo.Throw
                                       + microsoft.visualstudio.projectsystem.ni!CommonProjectSystemTools.Rethrow
                                        + microsoft.visualstudio.projectsystem.implementation.ni!Microsoft.VisualStudio.ProjectSystem.ExceptionFilter+<>c__DisplayClass2_0.<Guard>g__Action|0()
                                         + guardmethodassembly!GuardMethodClass.GuardMethod(class System.Func`1,class System.Func`2,class System.Func`2)
                                          + microsoft.visualstudio.projectsystem.implementation.ni!Microsoft.VisualStudio.ProjectSystem.Designers.ProjectFaultHandlerService+<HandleErrorAsync>d__45.MoveNext()
                                           + microsoft.visualstudio.projectsystem.implementation.ni!System.Runtime.CompilerServices.AsyncTaskMethodBuilder.Start[Microsoft.VisualStudio.ProjectSystem.Designers.ProjectFaultHandlerService+<HandleErrorAsync>d__45](<HandleErrorAsync>d__45 ByRef)
                                            + microsoft.visualstudio.projectsystem.implementation.ni!ProjectFaultHandlerService.HandleErrorAsync
                                             + microsoft.visualstudio.projectsystem.managed.ni!FaultExtensions.ReportFaultAsync
                                              + microsoft.visualstudio.projectsystem.managed.ni!Microsoft.VisualStudio.ProjectSystem.FaultExtensions+<>c__DisplayClass6_0.<RegisterFaultHandlerAsync>b__0(System.Threading.Tasks.Task)

And by checking the exceptions thrown in the trace around the time of the call stack above, it's also caused by the razor ExternalAccess change:

[tmat]

We should instead insert new Razor build. I thought it's already inserted but apparently not.

[tmat]

Let's make sure Razor is inserted instead of reverting this.

[tmat]

Razor insertion is now in progress.

[RikkiGibson]

Could you please include a link to it?

[dibarbet]

Could you please include a link to it?

@RikkiGibson this is the release that is generating the insertion, unfortunately it doesn't have your improvements and it has to clone VS so it may take a few :)
https://dev.azure.com/dnceng/internal/_releaseProgress?releaseId=40624&_a=release-environment-logs&environmentId=121442

[dibarbet]

PR link - https://devdiv.visualstudio.com/DevDiv/_git/VS/pullrequest/379899?_a=overview

[RikkiGibson]

We're now dual inserting in https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/379897

[RikkiGibson]

We're now dual inserting in https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/379897

RPS results are in for this. Unfortunately, we're still regressing a counter:

• BuildInfoVS64.Test - RegressedBeyondThreshold / 0001.BuildInfo / Build_Ngen_InvalidAssemblyCount / Regressed: 1.00 Count (16.67%)

It appears the same counter is regressed in the insertion containing only aspnetcore changes. https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/379899

cc @genlu @tmat

The revert validation does have clean RPS results. If I don't hear from you guys tonight with a sure-fire "change X and the assembly count will be fixed, guaranteed", or "we need an exception for the new assembly, guaranteed", I'm going to merge this revert, so that we can hopefully have a healthy insertion in the morning.

It's very important that our release branches stay insertable, because if more PRs get merged that break RPS while this is also going on, it will be even more difficult to triage the problems and get un-stuck. Please ensure that if your change depends on a recent change in another component, that you verify that the other component is up to date before merging to Roslyn.

[tmat]

@NTaylorMullen There seems to be version mismatch with Razor assemblies. Missing binding redirect?

02/14/2022 23:51:57.434 [5464]: Failed to load dependency Microsoft.AspNetCore.Razor.Language of assembly Microsoft.WebTools.Languages.Razor.Core, Version=17.0.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a because of the following error : The located assembly's manifest definition does not match the assembly reference. (Exception from HRESULT: 0x80131040)
02/14/2022 23:51:57.637 [5464]: 1>    Compiling assembly C:\VisualStudio\Common7\IDE\Extensions\Microsoft\Web Tools\Languages\Razor\v4.0\Microsoft.WebTools.Languages.Razor.Core.dll (CLR v4.0.30319) ...
02/14/2022 23:51:57.934 [5464]: 3>    Compiling assembly C:\VisualStudio\Common7\IDE\Extensions\ukdqr4lk.ss4\MSTestv2IntelliTestExtension.dll (CLR v4.0.30319) ...
02/14/2022 23:51:58.028 [5464]: 1>Warning: System.IO.FileLoadException: Could not load file or assembly 'Microsoft.AspNetCore.Razor.Language, Version=6.0.2.0, Culture=neutral, PublicKeyToken=adb9793829ddae60' or one of its dependencies. The located assembly's manifest definition does not match the assembly reference. (Exception from HRESULT: 0x80131040) ---> System.IO.FileLoadException: Could not load file or assembly 'Microsoft.AspNetCore.Razor.Language, Version=6.0.0.0, Culture=neutral, PublicKeyToken=adb9793829ddae60' or one of its dependencies. The located assembly's manifest definition does not match the assembly reference. (Exception from HRESULT: 0x80131040)
   --- End of inner exception stack trace ---02/14/2022 23:51:58.028 [5464]: 1> while resolving 0x1000031 - Microsoft.AspNetCore.Razor.Language.RazorCodeDocument.

[tmat]

We should insert Razor separately and make sure it's good.