importante du mecanisme de UnifyType :
- fait :
  * definition de Eff.poly_match
  * TypeVar type_var au lieu de Any/Overload
    avec type_var = Any | AnyNum
    (pour l'instant et pour longtemps ?!)
- à faire :
  * revoir UnifyType pour qu'il rende un poly_match
  * stocker là où c'est nécessaire les poly_match
    calculés lors du type check
    (pour les CALL et peut-être les sargs ?)

parametric nodes (or predef iterators) are nested.

Before, nested iterators were handles as follows : a source-level
(i.e., not compiled) alias node was created on-the-fly, and then
compiled. But this way to proceed was buggy for static args coming
from nested iterators or nested parametric nodes.

Now, I rely on Inline.iterators to get rid of iterators (the iterator
inlining is now recursive).

not use the module name. Otherwise, names with "::" are created, and
sometimes the lhs of the "::" does not even correpond to an existing
module.

'+' replaced by "p"
(see "correct_num_string" in LicDump.ml)

'-' replaced by 'm', '.' by 'd',
(see "correct_num_string" in LicDump.ml)

message if the constant is abstract ("an integer is expected whereas an int is
provided" was not a great message...).

not expanded.

info is attached to all  val_exp. Moreover, this allow to perfrom all
the  type checking  in  EvalType.f, instead  of  doin part  of it  in
GetEff.translate_val_exp.

maintaining (ugly and error-prone) hash tables.

constants that  are not  on the base  clock, add the  suitable "when"
statement if it was not present  in the source code (the lic (and the
ec/lv4) backend  does not  perform such kind  of clock  inference for
constants.

instead of silently generate  (syntactically) wrong lustre v4 (or ec).

instanciated by a polymorphic operator.

iterated, was still printed using the prefix/long name method (e.g.,
"Lustre__not(a)"  instead   of  "not  a").   Moreover,  equality  was
inconsistently represented as "eq" and "equal". I choose the former.

still   printed   using    the   prefix/long   name   method   (e.g.,
"Lustre__and(a,b)" instead of "a and b".

expansed) type  definitions that were containing struct  or arrays in
the expand_structs mode.

var only! Hence, I tranform "#(a)" into "a" in the lv4 mode.

as extern constants.

Also  fix  a  bug   in  the  struct/array  expanser:  when  expanding
structured constants  into val_exp,  I need to  add a entry  into the
type and clock val_exp tables.

 - expr such as "current x,y" should be written "current (x,y)"
 - abstract   struct  or   array   types  were   handled  as   extern
   types, which  prevent the struct  and array expanser to  work them
   out. In order to fix that, I have added an Abstract_type_eff variant
   to Eff.type_ which contains the concerte type.

the moment, I just don't generate such operator anymore (i generate the
polymorphic version Lustre::lt instead). I don't really know what is the
better.

fixing the v4 backend : print "+" instead of "Lustre::plus" (and ditto for the others).

(i.e., tuple of size one).

with a non-empty body (instead of just the first node).