mjmckp;7366 wrote:I see, thanks. Do you think it would be straightforward to reproduce the bug outside NCrunch so that it can be reported to Microsoft?
It depends upon the nature of the problem. There is a possibility that the type provider results in a valid IL sequence that NCrunch is somehow mangling with its instrumentation (this is what I hope is happening), in which case I'll be able to solve the problem with a code fix.
If it really is a bug in Roslyn itself, then reporting it to MS would be the normal procedure .. though implementing a workaround would also be a preference.
The problem with this issue is that because it's being reported down the line and very obscurely, it's likely to be very hard to actually understand it and pin it down. We're in real luck that MS have open sourced Roslyn. Without this, I'm not sure if trying to solve the problem would be feasible.