[PATCH 0/3] Fix some error-printing issues

[Tom Tromey]

This series came out of a race reported to gdb bugzilla.  This is
fixed in patch 1, though perhaps a different approach would be better
for some of it -- I'd appreciate comments on this, I don't mind
redoing it.

Also there's another race in here that I haven't tackled yet, where
bfd_check_format can race with bfd_cache_close_all.

The next two patches clean up the error reporting system a little, to
make it friendlier to gdb.

Let me know what you think.

Tom