Fix Linux NGEN/R2R symbols

[brianrob]

As of .NET Core 2.0, crossgen no longer includes the .ni.dll suffix for native images. PerfView still expects the .ni.dll suffix, and so it is not able to resolve any NGEN or R2R image symbols.

This PR fixes the issue in a backwards compatible way, allowing resolution of the image with or without the ni suffix.

[vancem]

Note that there is a break in the tests that looks like it is caused by this change. (ArgumentOutOfRange in the ZipDump test).

at System.String.Substring(Int32 startIndex, Int32 length)
   at Diagnostics.Tracing.StackSources.LinuxPerfScriptEventParser.GetFileNameWithoutExtension(String path, Boolean stripNiSuffix) in C:\projects\perfview\src\PerfView\OtherSources\Linux\LinuxPerfScriptEventParser.cs:line 275
   at Diagnostics.Tracing.StackSources.LinuxPerfScriptMapper.PopulateSymbolMapperAndGuids(ZipArchive archive) in C:\projects\perfview\src\PerfView\OtherSources\Linux\LinuxPerfScriptEventParser.cs:line 703
   at Diagnostics.Tracing.StackSources.LinuxPerfScriptMapper..ctor(ZipArchive archive, LinuxPerfScriptEventParser parser) in C:\projects\perfview\src\PerfView\OtherSources\Linux\LinuxPerfScriptEventParser.cs:line 653
   at Diagnostics.Tracing.StackSources.LinuxPerfScriptEventParser.SetSymbolFile(ZipArchive archive) in C:\projects\perfview\src\PerfView\OtherSources\Linux\LinuxPerfScriptEventParser.cs:line 116
   at Diagnostics.Tracing.StackSources.LinuxPerfScriptStackSource..ctor(String path, Boolean doThreadTime) in C:\projects\perfview\src\PerfView\OtherSources\Linux\LinuxPerfScriptStackSource.cs:line 279
   at Diagnostics.Tracing.StackSources.ParallelLinuxPerfScriptStackSource..ctor(String path, Boolean doThreadTime) in C:\projects\perfview\src\PerfView\OtherSources\Linux\LinuxPerfScriptStackSource.cs:line 19
   at LinuxTracing.Tests.XmlWriting.Write(String source) in C:\projects\perfview\src\LinuxEvent.Tests\XMLWriting.cs:line 20
   at LinuxTracing.Tests.XmlWriting.ZipDump() in C:\projects\perfview\src\LinuxEvent.Tests\XMLWriting.cs:line 28

[vancem]

The assignments of start and end are dead at this point. You should remove them.

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@d457feb). Click here to learn what that means.
The diff coverage is 100%.

@@            Coverage Diff            @@
##             master     #436   +/-   ##
=========================================
  Coverage          ?   17.54%           
=========================================
  Files             ?      213           
  Lines             ?   122889           
  Branches          ?    11839           
=========================================
  Hits              ?    21557           
  Misses            ?   100474           
  Partials          ?      858

Flag	Coverage Δ
#2017	17.54% <100%> (?)
#Debug	17.54% <100%> (?)

Impacted Files	Coverage Δ
...w/OtherSources/Linux/LinuxPerfScriptEventParser.cs	77.96% <100%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d457feb...c2a14c3. Read the comment docs.

[brianrob]

Thanks for the review @vancem. These should be fixed now.

[vancem]

I am surprised this works as you intend because NISuffix =".ni." (that is it has a trailing dot).

I would have expected that typically you would have a path = Foo.ni.dll and name would be Foo.ni (that is no trailing dot), and thus name.IndexOf(NISuffix) would FAIL, and the new code will never do anything.

Are we sure that actually works (it puts both Foo AND Foo.ni in the table)?

[brianrob]

The input path is for the map file and not for the NI. An example path looks like "traces/perfdemo.trace/System.Console.ni.{d2de1c76-d19d-4c52-8ff5-70e88349ea91}.map" Originally, this function reduced this to "System.Console.ni.{d2de1c76-d19d-4c52-8ff5-70e88349ea91}". If you specify true for stripNiSuffix then you get "System.Console.{d2de1c76-d19d-4c52-8ff5-70e88349ea91}". For .NET Core 2.0, the perfinfo-$pid.map file no longer contains the ".ni" and that's where things were breaking.

I have confirmed that it does work, and that we do put both Foo and Foo.ni in the table. You can see where we put both into the table here: https://github.com/Microsoft/perfview/pull/436/files#diff-151b5622877b35646b2ea5e3c9466af6R701