[m-rev.] for review: a pass for computing type representations

[Julien Fischer]

Hi Zoltan,

On Tue, 30 Jan 2018, Zoltan Somogyi wrote:

> Peter, will you please review this diff? You were the last person
> to work on type representations when you implemented argument packing,
> and you added Mantis feature request 432, which this diff prepares for,
> so you are the logical person.

>From the looks of it, this change has broken something since the mdbcomp
library no longer builds in the csharp grade.

     ...
     Making Mercury/css/mdbcomp.shared_utilities.cs
     Making Mercury/css/mdbcomp.slice_and_dice.cs
     Making Mercury/css/mdbcomp.sym_name.cs
     Making Mercury/css/mdbcomp.trace_counts.cs
     Making Mercury/css/mdbcomp.feedback.automatic_parallelism.cs
     Making mer_mdbcomp.dll
     ** Error making `mer_mdbcomp.dll'.
     Mercury/css/mdbcomp.feedback.automatic_parallelism.cs(3657,44): error
     CS0426: The nested type `Var_name_table_0' does not exist in the type
     `mercury.mdbcomp__program_representation'
     Mercury/css/mdbcomp.feedback.automatic_parallelism.cs(3662,39): error
     CS0426: The nested type `Var_name_table_0' does not exist in the type
     `mercury.mdbcomp__program_representation'
     Compilation failed: 2 error(s), 0 warnings
     gmake[1]: *** [libmer_mdbcomp.install] Error 1

Julien.