nb:  unexpected failures 195->187

Also start to add support for the other iterators, but it does not work yet.

The idea is that
- when it is not in the memory, it is considered as true
- it is set to false after each step is executed

test: 203->198 failures

test: 238->203 failures

nb: 20 more tests passes.

Always use Soc2cPredef.gen_assign to generate var assignments.

Also, use affection for assigning structures.

nb: 21 more tests passes.

Also do not crash if the soc contains reference to extern types.

To do that, I've used the (un)expand_profile function defined in SocExec
(and put the in a new module with a more neutral name: SocVar).

expected passes		1116 -> 1121
unexpected failures	282 -> 277

of expected passes		1073 -> 1116
of unexpected failures	325 -> 282

Indeed, Soc with memory can be used several times; hence we mark the
set of visited soc while printing the soc typedef.

 of expected passes		1052 -> 1073
 of unexpected failures	346 -> 325

46 ok, 159 ko...

121 ok, 110 ko. Basically, ko programs contains arrays arrays and
structures.

Define the ctx typedef a good order (w.r.t. typedef dependancies).

- Generate code for most predef operators.
- Generate a loop.c file (in a similar manner as ec2c -loop does)

It generates C code that compile on at least one example.

Merge branch 'master' of git+ssh://scm.forge.imag.fr/var/lib/gforge/chroot/scmrepos/git/lustre/lus2lic

Add the modes3x2 exemple of Pascal (its version + a la V6 version).

Fix the doc wrt the merge exemple (the tex was refering to a
non-existant file).

Work in progress...

cf rdbg git 66d01f567940e06eab9a49604663d43c29fdfe01

Instead, print a warning and consider the input as unbound.
Is it really the right thing to do?

cf git sha 55d3d77c470f416befd16a1b7d47f751eb2171fc of rdbg

cf git sha 142783a77cad1a0f7ef91972b8376f0b0e44b878 of rdbg

(cf 55d3d77c470f416befd16a1b7d47f751eb2171fc)

Some generated intermediary variables where declared as int instead
of bool.  Funnily, ecexe was ok with that...

More specifically, replace @ by rev_append and flatten+map by fold_left.

On a file that has 11000 equations (normal.ec), I win a factor 4: 60s -> 15s

The main difference is that I use a map instead of list to store
visited nodes.  The compilation time of normal.ec is divided by 2!!
(35s->15s)

We do not really see the difference in the time of the non-reg tests
since the execution of normal.ec still exceeds the 10s timeouts.

Also add support to rdbg to perform nonreg test. I do not plug yet it
as it's twice slower. but most of it seems due patch_ecexe. Indeed,
lus2lic -interface compile the whole program just to print the node
interface.

nb: it has detected a couple of existing bugs.