[m-rev.] diff: as experiment, generate .int2 files using .int3 algorithm as well

[Zoltan Somogyi]

This adds a way to compare the .int2 files we generate with the
.int2 files we *would* generate if they actually were just fully module
qualified versions of .int3 files. As it happens, the difference is quite big.

I will next see which of these differences are required (and why), and which
are just accidental.

This is a "diff" and not "for review" because (a) it does not change what the
compiler does at all unless you specify an experiment option, and (b) even then,
it only outputs a new file (suffix .int2.via3) which is never read.

Zoltan.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Log.int2via3
Type: application/octet-stream
Size: 990 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20190804/4909a23d/attachment-0002.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: DIFF.int2via3
Type: application/octet-stream
Size: 12149 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20190804/4909a23d/attachment-0003.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: int2diff.gz
Type: application/gzip
Size: 130847 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20190804/4909a23d/attachment-0001.gz>