[m-dev.] for review: workaround for the large term problem in the debugger

[Zoltan Somogyi]

Erwan wrote:
> Well, I have implemented the same sort of things in
> mercury_trace_external.c. (variable name retrieval, printing just one
> argument).
> Maybe we should rather put common stuff in mercury_trace_util.c and
> write procedures in such way that we can both use it (read "you should
> write procedures in such a way that I can use it" :-)).

I agree with this in general. I have therefore moved the function
MR_trace_materialize_typeinfos into mercury_trace_util.c.

> For example, MR_trace_list_vars should produce a list of string of the
> name of the variable instead of directly printing them (so I can send
> the list to the Opium process). Of course, I can also duplicate your
> code and adapt it to my own need , but I'am not sure that's the way to
> go.

However, in this case, the common code is only the trivial control structure
of a loop, and the differences in the loop body wrt memory management
(none for the internal debugger, alloc and later free of the list
for the external debugger) are reasonably significant.

Zoltan.