Revert "Switch to new lower-overhead spell checking protocol"

[RikkiGibson]

#68426 is suspected of introducing RPS regressions.

Last known good: https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/477008
First known bad: https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/477037
Revert validation: TODO

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 @CyrusNajmabadi @dibarbet.

It looks like this PR is improving perf, so most likely we want to be seeking an exception. In that case, this PR tracks the status of getting that exception.

Regressed counters

❌ Test Run FAILED
There were 10 regressions, please review the results below.

🕳 View Performance Details on PIT
PR build 33807.538.dn-bot.230608.055657.477037 VS. Baseline CI build main-33807.538

Performance results from Target run and Baseline run

🚩 Regressions
Check if there is a known noise issue in these counters

Found in Details Next steps
BuildInfoVS64.Test
0001.BuildInfo
Build_Ngen_InvalidAssemblyCount
Regressed: 6 Count (35.29%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
WinformsVS64.Designer
9990.Totals
CLR_AdjustedExceptions_Count_Total_devenv
Regressed: 6 Count (17.71%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.SolutionManagement
9990.Totals
CLR_AdjustedExceptions_Count_Total_devenv
Regressed: 5 Count (5.22%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.Typing
9990.Totals
CLR_AdjustedExceptions_Count_Total_devenv
Regressed: 6 Count (6.01%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.RebuildSolution
9990.Totals
CLR_AdjustedExceptions_Count_Total_devenv
Regressed: 5 Count (5.09%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
WebToolsVS64.SolutionManagement
9990.Totals
CLR_AdjustedExceptions_Count_Total_devenv
Regressed: 6 Count (22.9%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
UWP64.ProjectManagement
9990.Totals
CLR_AdjustedExceptions_Count_Total_devenv
Regressed: 5 Count (1.28%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
CPlusPlusVS64.SolutionManagement
9990.Totals
CLR_AdjustedExceptions_Count_Total_devenv
Regressed: 6 Count (11.63%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
CPlusPlusVS64.Debugging
0400.Stop Debugging
CLR_BytesAllocatedLOH_Total_devenv
Regressed: 695,168 Bytes (6.71%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
CPlusPlusVS64.Coding
0400.MemberList
CLR_BytesAllocatedLOH_Total_devenv
Regressed: 611,187 Bytes (6.38%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView

✅ 14 Improvements found
Found in Details Links
ManagedLangsVS64.Debugging
0200.Start Debugging
CLR_BytesAllocated_devenv
Improved: -3,881,378 Bytes (-23.06%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
WinformsVS64.Designer
9990.Totals
CLR_MethodsJitted_Total_devenv
Improved: -514 Count (-10.66%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.SolutionManagement
9990.Totals
CLR_MethodsJitted_Total_devenv
Improved: -491 Count (-11.62%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.RebuildSolution
9990.Totals
CLR_MethodsJitted_Total_devenv
Improved: -510 Count (-10.9%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
UWP64.ProjectManagement
9990.Totals
CLR_MethodsJitted_Total_devenv
Improved: -865 Count (-11.43%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
CPlusPlusVS64.SolutionManagement
0400.Change Solution Configuration - Warm
Duration_TotalElapsedTime
Improved: -615 ms (-20.72%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.SolutionManagement
0100.Open Solution
RefSet_Image_Delta_devenv
Improved: -3,178,496 Bytes (-3.12%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.SolutionManagement
0200.Close Solution
VM_AdjustedImagesInMemory_Total_devenv
Improved: -2 Count (-0.59%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.AddNewProject
0100.Add New Project
VM_AdjustedImagesInMemory_Total_devenv
Improved: -2 Count (-0.56%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.Debugging
0600.Stop Debugging
VM_AdjustedImagesInMemory_Total_devenv
Improved: -1 Count (-0.24%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
ManagedLangsVS64.Typing
0200.Typing
VM_AdjustedImagesInMemory_Total_devenv
Improved: -1 Count (-0.28%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
WebToolsVS64.TypingInWebforms
0200.Typing CSharp in Webforms
VM_AdjustedImagesInMemory_Total_devenv
Improved: -2 Count (-0.39%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
WebToolsVS64.Debugging
0300.Stop Debugging
VM_AdjustedImagesInMemory_Total_devenv
Improved: -1 Count (-0.18%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView
UWP64.ProjectManagement
0200.Xaml Designer Load
VM_AdjustedImagesInMemory_Total_devenv
Improved: -1 Count (-0.22%) 🕳 View it in PIT
📂 Open test outputs
📈 Compare in PerfView

[CyrusNajmabadi]

@veler ?

[CyrusNajmabadi]

I don't see what we could have done to CPlusPlusVS64.Coding. That's C++.

Also, getting ngen problems for us referencing the latest VS binaries doesn't seem acceptable. We' going to need to move to them no matter what.

[veler]

I don't see what we could have done to CPlusPlusVS64.Coding. That's C++.

Also, getting ngen problems for us referencing the latest VS binaries doesn't seem acceptable. We' going to need to move to them no matter what.

Thanks for pinging me here. I will look into it. Indeed, the C++ regression is odd. This one might be unrelated.

[beccamc]

I also need LSP version 17.7.7-preview (or newer) to go into VS (issue). Cyrus updated the version in #68426. It doesn't look like this PR reverts the LPS version. Does that change stay?

[RikkiGibson]

I left the package versions alone mainly because there were conflicts in the revert (other changes also changed those package versions).

It definitely feels like if RPS regressions are occurring simply from updating our versions of VS packages, that means the team that owns those packages owns the regression.

It looks like this PR won't actually build anyway with the given package versions. If we decide to move forward with reverting we'll want to push a commit to this branch to get it into a buildable state.

However, it feels like there shouldn't be a need to avoid adopting new VS package versions, or to revert the indicated PR here. Mainly there's a need to investigate, confirm that the regressions are something we're comfortable with, and move forward with the exception process.

[genlu]

The insertion was completed last week, I believe this is no longer needed. @RikkiGibson

[CyrusNajmabadi]

Thanks so much!