Fix SimpleDiagnostic GetHashCode is more specific than Equals

[pdelvo]

Fixes #57.

I hope it is okay that I just went ahead and fixed it.

[pdelvo]

Btw. Why are the additional locations not used to determine equality?

[theoy]

I hope it is okay that I just went ahead and fixed it.
-- @pdelvo

It is absolutely okay, in fact, awesome. Thanks for offering a fix - we're almost ready to start blasting through pull requests. We're just in the process of replacing our source control internally with Git on both sides so that we can fast-track all the subsequent pull requests 😄 Assignining to @mavasani since he has the corresponding bug too.

[pdelvo]

@AdamSpeight2008
I don't think that the foreach is expensive. foeach on Arrays are getting turned into a for loop by the compiler if Im not mistakten.

@mavasani Create is calling the constructor which is checking messageArgs for null. It is using an empty array instead if it is. Hash.Combine is checking for null so this should not be a problem if items in messageArgs are null. If I ignore null items in the array new[] {null, "foo"} would have the same hash as new[] { "foo"}

@pharring I did not know about Hash.CombineValues. Is the jit compiling the foreach loop in there into a for loop too?

I came up with a solution which should be faster and more readable.

[pharring]

Nit: Assignment to zero is not necessary since it's going to be overwritten on the next line.

[pdelvo]

@pharring I fixed that and I fixed a possible NRE if the location is null

[AdamSpeight2008]

Doesn't foreach get converted into a while loop?

var en = foos.GetEnumerator();
while( en.MoveNext() )
{

}

[pdelvo]

I can't compile roslyn on my machine this is why this stupid mistake could happen >.<
@AdamSpeight2008 Roslyn does optimize that.

Edit: Im going to squash that into one commit when this is okay

[pharring]

Just an observation: While using the local variable undoubtedly makes the code easier to read, it does introduce a slight performance overhead. The original form's IL is shorter and (obviously) eliminates a local variable slot. The Roslyn compiler (currently) doesn't try to eliminate the local. I don't know whether the various JIT compilers will do so either.

[pdelvo]

I hope this is what you are looking for. If it is Im going to squash these commits together.

[pdelvo]

I squashed this into one commit.

[pdelvo]

The failed test semms to be unrelated to this

[mavasani]

@pdelvo Agreed, failing test "HandleSolutionProjectTypeSolutionFolder" seems to be unrelated to this change. I'll merge your PR. Thanks again!

[jaredpar]

@pdelvo

Could you take a minute and sign the .NET Foundation CLA for the contributed code?

https://cla2.dotnetfoundation.org/

This should have been done before we merged the PR but there was some confusion on the process.

[pdelvo]

Done

[jaredpar]

@pdelvo thank you!