Gdb does not support target mips-mips-rtems

[gregory.menke at gsfc.nasa.gov]

 > > 
 > > Hi All,
 > > Please disregard my previous email.  I have discovered and corrected the
 > > configuration problems and I now have GDB-5.1.1 with the Insight Tck GUI
 > > front-end properly configured for mips-rtems and the Cygwin/XFree86
 > > environment.  I am now downloading over Ethernet and executing code on my
 > > NEC DDB-VRC5476 (NEC VR5432 CPU) development system using my newly
 > > constructed tools.  This was a major hindrance to my STB project because I
 > > was using the supplied Cygnus Pro tools circa 1999 (mips64vr5xxxel-elf-gcc,
 > > etc) to build rtems.  There are still some small issues, but when the task
 > > and thread awareness becomes stable enough I will make it available on my
 > > website.
 > > 
 > > :)
 > 
 > The current snapshots have a thread-aware gdb serial stub for the mips.
 > I worked with Greg Menke to get it to work.  
 > 
 > FWIW this work is architected so that it should be relatively painless 
 > to use the thread specific part on any CPU.

The stub is only tested on an R3000 MIPS, though the R4000 <should>
work.  The core stub code is shared for all MIPS, but in order to
enable it, a bsp has to initialize the gdb stub- specifically, it has
to tell the stub a little bit about the memory map of the board.  The
Mongoosev bsp has a representative implementation at the moment.

On our board, we have the stub working in C and C++ over the 2nd
serial port of our breadboard.  To use the stub, a bsp must supply a
getchar and putchar routine to support the gdb protocol.  Hooking this
up to ethernet would probably be the biggest hurdle in using the stub,
but I imagine its been done well enough before.

The stub is thread aware and can step code compiled at -O0.  Its
breakpoint support could be enhanced a little, but works reasonably
well.  Our biggest problem at this point is a mysteriously incorrect
link map...

The latest version of the stub should be available in the next RTEMS
snapshot.

Gregm