GRC now confused about source/sink sizes

[This is with latest master: 3.7.6git-57-g029079cf]

If you pull simple_ra from cgran:

svn co https://www.cgran.org/svn/projects/simple_ra

go into simple_ra/trunk

And do a “make”

It will fail.

If you then:

PYTHONPATH=.:$PYTHONPATH
export PYTHONPATH

(add ‘.’ so that it will correctly pick up simple_ra_utils)

then manually pull simple_ra_receiver.grc into gnuradio-companion,
you’ll find that it will error claiming that a virtual source of size 0
is connecting
to sink of some other size. Open the offending block, click “OK”,
and the error leaves. Exactly which source/sink blocks are involved
seems
to be somewhat non-deterministic. But when you open the block, and
“OK”, I guess it re-evaluates the dependency graph or something, and
everything is just fine from that point onwards, then just build and
save, and you’re on your merry way.

But this prevents the makefile from working, since grcc is just
non-deterministically erroring-out.

I was seeing similar behaviour with tnt’s eval-cache patch, so I wonder
if this is similar, but due to Sebastian’s recent work on GRC?