fix underlinking when linking with -Wl,--as-needed

Makefile: fix build on 64 bit

Type __unbounded is deprecated. See:
    http://www.sourceware.org/ml/libc-alpha/2013-02/msg00269.html

Replacing TESTING_ROOT is not needed anymore since we have env.sh for this.

Fixed transformed list. Now works in all cases. Added yet another test script to witness this. Now tracking a new bug in graph display to show final result that will close up the transformation improvements.

- Fixed the transformed list management. There is still one bug where instructions get freed too early. Tracking that one.

-------------

- Fixed problems in the transformation engine that prevented 
the list of transformation instructions from being iterated over while 
rewriting was performed. Fixed problems related to managing the life-time 
of induction variables in the case of nested loop where the transformed
instruction list is involved.

- Added missing evarista files from last commit.

- Added (working) tests for transformation in testsuite/scripting/rewrite*.esh

TODO: Fix a newly introduced bug two commits ago where the list of IR 
instructions fails to be printed in graph once the transformation has 
finished. Looks like we are freeing some expressions too early. I am 
trying to get rid of memleaks so dont blame me ;-)

Julien

-> modified cont command to reflect this too

- Removed subtle bug in debug state machine that would not reinstall breakpoint if step was enabled in the middle
-> modularized code to be able to reuse e2dbg_breakpoint_reinstall()

- Added capability to foreach to record expression-typed values back in iterated list.

- Removed many gcc warnings 

- Added current getpid() value when debugger first takes control

=> The embedded debugger is now working again on Linux/IA32

(successfully traces into multiple breakpoints without ptrace on IA32)

- Fixed eresi-config scripts so that more testsuite build with them

- Fixed e2dbg to aoid using s*printf functions at initialization time as these could use malloc on newer glibc whose symbol is not resolved yet.

- Fixed revm_expr_extend() to add fields to expressions on the fly during transmation (to be cont)

- Added revm_expr_contex_t so that many revm expr engine API is more modular and API take less parameters

- Corrected a one-byte heap overflow in the revm expr engine (fieldname dup should have allocated +1 for NUL byte)

FOUND NEW BUGS: 

- There is a new section type .gnu.hash that is now replacing legacy .hash - need libelfsh support
- e2dbg fails to handle SIGTRAP in-process as before even on linux/x86, maybe a mprotect-related problem, reenabled sigsegv handling

More to come.

- Fixed thorkill's fix of bug #126 . Now it will work. The fix was cutting block for no reason in some conditions. Had to strenghen the check.

- Now allow complex arithmetic in expression initialization
- Added support for logical OR operation in arithmetic
- added commands INTER and UNION for set-based analysis
- fixed bug in PRE/POST command that was forgetting to initialize a hash
- Added support for lazy reflection when given a second parameter to command (avoid reflecting a whole block, you can ask a number of instr now)
- Added API for list append
- Added testsuite for rewriting (rewrite-simple.esh)
- other minor cleanups

- Implemented the arithmetic types for all sparc instructions

- Adapted evarista types to take advantage of this feature

- Incremented evarista backend for sparc

- Fixed many bugs in the source level debugging function. Still not operational but should not crash now

- Added function for intersecting hashes (aka sets)

- Added a libasm option (libasm.output.synth.instrs) to disable synthetic instructions (it makes it easier to translate to IR without them)

- Simplified the sparc backend since we dont have to deal with synthetic instructions anymore

- synthetic instruction support is still here by default, just disable the option if you want to translate to IR and not manually understand the code

- This change fixed translation support for synthetic instructions that were previously badly supported.

Fixed some sparc ELIR translation, backup commit.

- Made ARGCOUNT feature more reentrant (removed 3 global variables)

- Added test for source level debugging in testsuite

- Removed new compilation warnings

- Removed unwanted debugging code

Removed a crash as we were trying to guess the value of a fctptr target, but even the simplest case was not handled by our naive attempt. Removing the attempt altogether also removes the crash.

Fixed a bug in the construction of the call graph in bounded exploration mode (e.g. maxdepth != 0). Now testsuite works nicely with bound = 1.