Hello,

This follows up:
http://sources.redhat.com/ml/gdb-patches/2003-04/msg00111.html
by ``going all the way''.

It modifies "disasm.c" so that it always goes to the target, and never 
the local executable.

The problem is that, regardless of the selected target vector, the 
disassembler can end up disassembling memory that is not in the 
executable - the assumption was wrong.  This is easily seen with:

	disassemble &variable &variable+1

The testsuite part of the patch does just this (fails before, passes after).

As for avoiding target memory read overhead, there is "set 
trust-read-only-sections on" which is a definite improvement.

Andrew

PS: I stumbled across this when trying to convert convert "x/i" to use 
the new disassembler (display.exp mysteriously gained three failures on 
a simulator target).