[PATCH 00/16] clean up remote.c state

[Tom Tromey <tromey@redhat.com>]

I was toying with David Taylor's idea of multi-target support, and I
came up with this preliminary series to do some cleanups of remote.c.

Most of this cleanup amounts to moving various global variables into
struct remote_state.  I did each variable (or in some cases, a few
clearly related variables) as a separate patch, to make the series
maximally understandable.  Each patch can essentially be read (though
not applied) in isolation.

I found the globals using nm.  Then I went through them by hand to
find ones that clearly must be global.


This patch does not fix all the globals in remote.c.  It is just a
start.  For example:

* I suspect the remote_g_packet_data stuff must be per-remote
* remote_protocol_packets has some per-remote state
* I think the client-stop notification code needs some update
* Maybe signal handling needs some fixing up, but what else is new


After this series goes in, I propose that all future remote.c changes
be reviewed to ensure that remote state is in remote_state and not a
new global variable.


This series also does a couple of other minor cleanups to remote.c.
It removes the gdb-specific CRC code in favor of libiberty's; and it
fixes an oddity I noticed with sizeof_pkg.


I built and regression tested this series on x86-64 Fedora 18 using
the native-gdbserver target board.

Tom