http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45841
--- Comment #4 from David Krauss <potswa at mac dot com> 2010-09-30 09:52:48 UTC --- Ah, I didn't even notice the target line before. This is unbuffered I/O, which can be challenging to the codecvt. But this isn't a failure I would anticipate. If the codecvt fails, there should be an exception; if it succeeds, the result is cached so the second sgetc() returns the same result. (Thus, it's not *really* unbuffered.) I don't see how the patch could change this particular behavior. Definitely need that trace. And I presume I'll need to upgrade gdb to use it. Which entails building GDB for that architecture from source, on my machine? Or do you guys happen to have a setup I can ssh into? :v)
