On May 21, 11:33*pm, Brian Drummond <brian_drumm...@btconnect.com>
wrote:
> Try running Translate from the command line with the exact command line given in
> the ".cmdlog" file - including the "-intstyle ISE" flag and see if that gives
> the same result as the GUI flow.

Thanks for your reply, Brian.

Unfortunately, the cmdlog indicates that the GUI flow uses a different
executable - unwrapped/ngdbuild.exe which has the problem I described
with loading constraints when run from the command line. So one issue
right at the start is that the GUI uses different binaries.

> Indeed an apparently successful run through the GUI tools produces a
> non-functional bitfile! I have not had time to explore deeply enough to
> find out why.

I don't think it's safe to assume the GUI and command-line flows have
the same results. From what I've seen, the two have diverged.
Fortunately, by manually including the UCF file (and therefore
deliberately disregarding the -i flag that cmdlog said we should use)
we are able to produce an identical final bit-file, excluding
timestamp in header, even though the intermediate files are all
different.

That's with our current inputs at least. Who knows whether this will
be the case tomorrow, or next month?

-- David.