Eloi Sanfelix b4b909ec1a - Fixed breakpoint behaviour
* Find recoil
	* Set pc to addr-recoil properly
	* Fixed bp_restore: per bp, check if handled, else do via io.write

- Improved debug_gdb
	* Added 'read all regs to buffer'
	* Added breakpoint callback: check if supported first time, else return FALSE

- Maybe something else...
2011-05-11 20:08:19 +02:00
..
2011-05-11 20:08:19 +02:00
2011-05-11 20:08:19 +02:00
2010-05-26 18:25:35 +02:00
2010-06-30 01:13:09 +02:00
2010-06-30 01:13:09 +02:00

libr.bp
=======

Breakpoint API

- Manages list of defined breakpoints
- Determines if a stop is caused by a breakpoint
- Owns a database of multiple types of breakpoints
  - arch and os based ones
  - Supports endianness
  - r_bp_get should return a buffer and a length
- Manages conditional breakpoints expressions
- Types of breakpoints
  - software (traps)
  - conditional traps
  - hardware (registers)
  - mmu (changes page protections)
- All non-native operations are translated into evaluable expressions
  by other modules. Like changing register values and so on
  - Do we should place some callbacks for this kind of ops?
- We need to make this work also remotely
  - r_debug can handle the remoteness of the debugger backend.
  - r_io can do it also
- Watchpoints and its exception should be handled here
  - watchpoint expressions should be handled by using the r_num stuff
- Hardware breakpoints require access to registers, or pid/tid
  this is... the debugger backend. For those, the debugger backend
  should fill a callback to manage them.
  - if the debugger breakpoint handler does not manages the breakpoint
    type, r_bp must do it with r_io storing and loading bp bytes.

* Do we need the plugin API to define new breakpoints and so on?