[gnugo-devel] Regression Bug(s?)

[Richard Outerbridge]

Version: 3.6 distribution
Area: Regression testing
Repeatable: Yes

Summary: Missing games are unreported and result in spurious
regression test results.

Hi... I'm compiling gnugo 3.6 under Mac OS X, and wanted to make
sure I hadn't broken anything.  So I ran the regression-test
"make all_batches" test suite.  Everything passed except one test
(140) in the very last test-suite, kgs.tst.

To make a long story short, kgs.tst looks for the following files
which are not in the 3.6 distribution (at least, not under these
names):

games/kgs/20040507-GnuGoCVS-read.sgf
games/kgs/20040516-GoBucks-GnuGoCVS.sgf
games/kgs/20040517-siRNA-GnuGoCVS.sgf
games/kgs/2004-04-10-GnuGoCVS-inuyasha34.sgf

and so the AWK script carries forward the last result it received
from gnugo.  As most of the tests are expected failures, most of
the time this will (as expected) fail (that is, pass) the test.

The regression tests don't actually prove very much unless they can
test that the testing is actually being carried out.

outer