[PATCH] D15375: Fix PR #25788: parsing of floating-point constants on non-C locales

[Antoine Pitrou via llvm-commits]

pitrou added a comment.

> I think we'd want APFloat::convertFromDecimalString with rmNearestTiesToEven seeing as how we don't expect to see -nan and the like.


I've looked at this and there are a couple issues:

- convertFromDecimalString() is private
- convertFromDecimalString() doesn't handle leading sign

Also, due to the lexing that occurs before calling APFloat() here, I don't think it can actually be called with `-nan` and friends. I'll upload an updated patch as it turns out the one here doesn't apply anymore.


http://reviews.llvm.org/D15375