Re: gdb remote stub for proprietary operating system?

[Daniel Jacobowitz <drow@false.org>]

On Tue, Jan 23, 2007 at 05:45:36PM -0500, Dominique Toupin (QA/EMC) wrote:
> Hi,
> 
> I have this info
> http://sourceware.org/gdb/current/onlinedocs/gdb_18.html#SEC165, when
> they say we can ignore the details I was hopping we could ignore the
> communication protocol.
> Do we have documentation other than sparc-stub.c?

That is not a good place to look for an OS stub.  Look at the appendix
which documents the communication protocol - I think it's Appendix D. 
You will almost certainly need to write your own implementation of
that, connected to your proprietary debug interface.  And it will give
you a good idea of what is expected of a remote stub.

If you have threads, you'll probably want vCont.  You'll need at least
g/G and m/M (or X).  If you have code in non-writeable or shared
memory, you'll need the z/Z packets.  And so forth.

> If you have to choose between 3 months, 6 months, 1 year, 2 years, what
> would be the guess for how long it would take 1 person to implement the
> remote stub to support all the GDB/MI commands?

It depends on the questions I asked, and the experience of the people
working on the project.  It is probably more like a six month project
than like a one year project.

Of course, if you need a better estimate than that, there are a number
of companies who can do this sort of project :-)

-- 
Daniel Jacobowitz
CodeSourcery