Suggest separating the GDB stuff out (native, target, remote) and using
separate e-mail threads to discuss each.
I was going to but it's easier this way.
It isn't for the person doing the review -> the easier the reviewer's life is, the quicker (well, ok, in theory :-/) the reviewer should be. I'd strongly focus on getting the cross debugger (--target=i386-unknown-nto) integrated into GDB, and then worry above the native.
The native nto-procfs.c makes use
of some of the code in remote-qnx.c and remote-qnx-<target>.c (we still have
four more targets.) If you really feel it's necessary I could do the work
but I had started on it and concluded it would either lead to a lot of
duplicated code or an explosion of files.
That looks like a design problem. The common code should likely be moved to something like nto-tdep.c. What exactly is common?