Which parts of LOGON tree are actually needed for GM regression testing?

LOGON
, , ,
1

Current instructions for Grammar Matrix regression testing say the entire LOGON tree installation is required.

Is that actually true?

If it isn’t, it would be much better if we could install just the part that is actually being used? (And update the instructions accordingly).

2

I would very much like to figure out the answer to this.

I just set up a new VM from scratch yesterday (will update instructions!) and it took several hours to find the right Ubuntu distro + libraries to get things to play nicely (and giving up on VirtualBox). Just downloading the LOGON tree took about an hour.

As far as I can tell looking at run_regression_test.sh, it uses LOGON for two things:

  1. copying the Relations file from ${LOGONROOT}/lingo/lkb/src/tsdb/skeletons/english/Relations
  2. Run this following monster command (which I’m not sure what it does but it looks like it compares profiles?):
{
  options=":error :exit :wait 300"
  echo "(setf (system:getenv \"DISPLAY\") nil)"

  echo "(setf tsdb::*process-suppress-duplicates* nil)"
  echo "(setf tsdb::*process-raw-print-trace-p* t)"

  echo "(setf tsdb::*tsdb-home* \"$tsdbhome\")"
  echo "(tsdb:tsdb :skeletons \"$skeletons\")"

  echo "$profiles"
  echo "$comparison"

} | ${LOGONROOT}/bin/logon -I base -locale no_NO.UTF-8 -qq 2> ${TSDBLOG} > ${TSDBLOG}

So, it seems like we could replace these with a copy of Relations and rtest.py?

3

Hi T.J.,

rtest.py is working really well for me. A few tests do fail currently but probably for good reasons. I updated the gold standard for a few in my branch; those should probably be updated in the trunk as well.

1 Like