JIT: Fix bug in finally cloning caused by unsound callfinally reordering

[AndyAyersMS]

We need to make sure that if we reorder a callfinally during finally cloning
that the callfinally is actually the one being targeted by the last block in
the try range.

Closes #18332. Linked issue has some more detailed notes.

[AndyAyersMS]

@BruceForstall PTAL
cc @dotnet/jit-contrib

Don't expect any FX diffs from this. Will verify.

[AndyAyersMS]

Well, there are FX diffs.... possibly requires exceptions to see the impact of this bug, so maybe not caught in any testing.

Completed Crossgen Diffs for System.Private.CoreLib.dll, framework assemblies in 273.73s
Analyzing diffs...
Found files with textual diffs.
Summary:
(Note: Lower is better)
Total bytes of diff: 169 (0.00% of base)
    diff is a regression.
Total byte diff includes 0 bytes from reconciling methods
        Base had    0 unique methods,        0 unique bytes
        Diff had    0 unique methods,        0 unique bytes
Top file regressions by size (bytes):
         123 : Microsoft.CodeAnalysis.dasm (0.02% of base)
          30 : System.Net.WebSockets.dasm (0.07% of base)
          13 : System.Net.Http.dasm (0.00% of base)
           6 : System.Net.HttpListener.dasm (0.00% of base)
           5 : NuGet.Protocol.Core.v3.dasm (0.00% of base)
Top file improvements by size (bytes):
          -6 : System.Net.Requests.dasm (-0.01% of base)
          -3 : System.Runtime.Extensions.dasm (0.00% of base)
          -2 : System.Security.AccessControl.dasm (0.00% of base)
          -1 : System.Collections.Concurrent.dasm (0.00% of base)
          -1 : System.Private.CoreLib.dasm (0.00% of base)
12 total files with size differences (5 improved, 7 regressed), 118 unchanged.
Top method regessions by size (bytes):
         123 : Microsoft.CodeAnalysis.dasm - CommonCompiler:RunCore(ref,ref,struct):int:this
          30 : System.Net.WebSockets.dasm - <ReceiveAsyncPrivate>d__65`2:MoveNext():this (3 methods)
           5 : System.Net.Http.dasm - <ConnectAsync>d__2:MoveNext():this
           4 : System.Net.HttpListener.dasm - HttpListener:RemovePrefix(ref):bool:this
           3 : NuGet.Protocol.Core.v3.dasm - <StartWithTimeout>d__0`1:MoveNext():this (3 methods)
Top method improvements by size (bytes):
          -7 : System.Net.Requests.dasm - FtpWebRequest:GetResponse():ref:this
          -3 : System.Runtime.Extensions.dasm - Environment:get_WorkingSet():long
          -1 : System.Collections.Concurrent.dasm - InternalPartitionEnumerable:GrabChunk_Buffered(ref,int,byref):bool:this
          -1 : System.Private.CoreLib.dasm - NativeOrStaticEventRegistrationImpl:RemoveAllEventHandlers(ref)
          -1 : System.Security.AccessControl.dasm - Win32:GetSecurityInfo(int,ref,ref,int,byref):int
21 total methods with size differences (6 improved, 15 regressed), 142199 unchanged.
Completed analysis in 112.09s

Am also trying out PMI based diffs on this across all tests, but amhitting some issues with (I think) marker file collision. So need to tweak that some...

[AndyAyersMS]

Here's an all-up PMI diff for this change using the still somewhat-in-flight PMI diffs I've been working on over in jitutils.

Not surprisingly the new test case shows up.

Completed PMI Diffs for System.Private.CoreLib.dll, framework assemblies, assemblies in D:\repos\coreclr\bin\tests\Windows_NT.x64.Release in 1262.86s
Analyzing diffs...
Found files with textual diffs.
Summary:
(Note: Lower is better)
Total bytes of diff: 136 (0.00% of base)
    diff is a regression.
Total byte diff includes 0 bytes from reconciling methods
        Base had    0 unique methods,        0 unique bytes
        Diff had    0 unique methods,        0 unique bytes
Top file regressions by size (bytes):
          77 : Microsoft.CodeAnalysis.dasm (0.01% of base)
          31 : JIT\Regression\JitBlue\GitHub_18332\GitHub_18332\GitHub_18332.dasm (6.22% of base)
          14 : NuGet.Protocol.Core.v3.dasm (0.01% of base)
          12 : System.Collections.Concurrent.dasm (0.01% of base)
           6 : System.Net.HttpListener.dasm (0.00% of base)
Top file improvements by size (bytes):
          -6 : System.Net.Requests.dasm (-0.01% of base)
          -3 : System.Runtime.Extensions.dasm (0.00% of base)
          -1 : JIT\HardwareIntrinsics\X86\Avx\CompareScalar_ro\CompareScalar_ro.dasm (-0.02% of base)
11 total files with size differences (3 improved, 8 regressed), 2802 unchanged.
Top method regessions by size (bytes):
          77 : Microsoft.CodeAnalysis.dasm - CommonCompiler:RunCore(ref,ref,struct):int:this
          31 : JIT\Regression\JitBlue\GitHub_18332\GitHub_18332\GitHub_18332.dasm - GitHub_18332:Aargh():ref
          12 : NuGet.Protocol.Core.v3.dasm - <StartWithTimeout>d__0`1:MoveNext():this (4 methods)
          12 : System.Collections.Concurrent.dasm - InternalPartitionEnumerable:GrabChunk_Buffered(ref,int,byref):bool:this (4 methods)
           4 : System.Net.HttpListener.dasm - HttpListener:RemovePrefix(ref):bool:this
Top method improvements by size (bytes):
          -3 : System.Net.Requests.dasm - FtpWebRequest:GetResponse():ref:this
          -3 : System.Net.Requests.dasm - FtpWebRequest:BeginGetResponse(ref,ref):ref:this
          -3 : System.Runtime.Extensions.dasm - Environment:get_WorkingSet():long
          -2 : System.Net.Http.dasm - <DrainAsync>d__15:MoveNext():this
          -1 : System.Net.Http.dasm - <ConnectAsync>d__2:MoveNext():this
19 total methods with size differences (7 improved, 12 regressed), 342431 unchanged.
Completed analysis in 286.99s

[sandreenko]

LGTM

[AndyAyersMS]

OSX failure is some kind of infrastructure/networking hiccup. Will retry

@dotnet-bot retest OSX10.12 x64 Checked Innerloop Build and Test

[AndyAyersMS]

@RussKeldorph this is a candidate for 2.1 servicing.