Fixes #9, Internal error in FSI: FS0192: binding null type in envBindTyp...

[KevinRansom]

Fixes #9, internal error: binding null type in envBindTypeRef
Fixes #10, internal error: binding null type in envBindTypeRef

Both issues were caused by the same underlying issue. RefEmit automagically applies escaping to names that
contain a ',' So , becomes \,.

When we tried to create a ref to a nested type we didn't add the ',' escaping so we couldnot bind to the created type.

The fix is very straightforward, change BasicQualifiedName to correctly escape names with comma's.

[latkin]

1-line fix FTW!

Banging on this, I see some related issues with pickling/de-pickling quotations containing types with commas. They don't seem to be caused by this change, though, so I'll open as separate bugs.

Have you checked how this affects type providers? They rely on comma-encoding in the type name to handle static parameters, so might be worth a careful look.

For the tests - the bug was with FSI only, but current tests won't be run through FSI (at least not usually). Rather than add a new core area, it might be a better fit to expand existing backtick stuff at fsharpqa\source\Conformance\LexicalAnalysis\IdentifiersAndKeywords. You can add 2 lines to the env.lst file, one that runs via FSC, one via FSI (add FSIMODE=EXEC).

[KevinRansom]

I'm looking at some type provider stuff right now, so I can move right on to the new bugs.
I was in two minds about re-using an existing test area. Although I concluded that item names have a bunch of test holes, they may not be terribly important, for example there are a number of banned characters that are not yet validated. I was thinking of adding a bug to get those test holes fixed.
Good call on the fsimode-exec.

[dsyme]

Nice work!

[KevinRansom]

Closed Fixes #9 and Fixes #10

[ovatsus]

Hi, I started hitting this FS0192 problem recently, even when testing on latest release of 4.0:

fsprojects/FSharp.Formatting#272

[latkin]

@ovatsus, looks like everything is cleared up in that issue, can you confirm? Turns out to be unrelated?

[ovatsus]

It's still a very bad error message for missing a reference, but yes, it is fixed now. Thanks