Don Syme
Don Syme
Thanks for the detailed info, that makes sense.
The access modifier should be taken into account, thanks for the report.
This is related to #6338 and would be fixed at the same time
> @forki Yeah the other implementations like Xamarin don't do anything with tailcalls, the arm compiler code thats generated by AOT just ignores it. My understanding this is inaccurate and...
Something of interest is that the F# compiler _does_ have logic to eliminate emit of "unnecessary" tailcall instructions even when `--tailcalls` is enabled. You can search for * [MakesNoCriticalTailcalls](https://github.com/Microsoft/visualfsharp/search?q=MakesNoCriticalTailcalls&unscoped_q=MakesNoCriticalTailcalls) *...
> The jit may be able to artificially increase the size of the root method's argument area via a special prolog so that fast tail calls to callees with more...
Agree this is a bug. The compiler generated ``patternInput`` is either internally considered public or is somehow otherwise triggering the check
Ugh, thanks for the bug report. We have an extensive systematic test suite somewhere for almost exactly this case and I can't quite believe this one slipped through, but eveidently...
There is one case not fixed by #12395. I'll leave this open to track this, though won't immediately seek to fix it ``` module rec Test12384d = type Node =...
Yrs, an explicit Main is required, and this is currently considered by design, for better or worse. Adding a warning on referencing such a component may well be feasible.