This is the mail archive of the
mailing list for the GDB project.
Re: [RFC] File-I/O, target access to host file system via gdb remoteprotocol enhancement
- From: Andrew Cagney <ac131313 at redhat dot com>
- To: Daniel Jacobowitz <drow at mvista dot com>
- Cc: gdb at sources dot redhat dot com
- Date: Wed, 26 Feb 2003 16:21:56 -0500
- Subject: Re: [RFC] File-I/O, target access to host file system via gdb remoteprotocol enhancement
- References: <20021111131354.N10395@cygbert.vinschen.de> <20021112212526.GA28814@nevyn.them.org>
I like this one too; but I'm not sure that I like linking it to the
remote-syscall interface. Implementation-wise, that's the most useful
method for a stub like RedBoot; but for a larger system like gdbserver,
it's useful to allocate the program a normal pseudo-terminal as if it
had a console. And to do character rather than line I/O.
As I'm sure Andrew will remind me, my suggestion doesn't fit the
current remote protocol very well; it's not synchronous enough for one
thing. So don't take this as an objection, and I'll come back to
proper remote console support a little later. This'll be a nice
Just FYI. The protocol doesn't preclude doing character IO and doesn't
preclude the remote end using a pty. The pty problem is that it is more
complex. The debug agent (or GDB, in the case of the MI console) needs
to be able to juggle events from multiple sources (wait and i/o).
Anyway, if the overhead of character i/o proves to be too great then a
refinement can be made.