Le Lundi 27 Mars 2006 14:25, Gunnar Farneb�ck a écrit : > Alain wrote: > > I have some unexpected results in regressions tests in 3.7.9. > > I run it with a standard gnugo 3.7.9 compiled with ggc4.0 > > Is it normal (some results have not been updated) or is it > > my compiler/computer fault ? > > How do you run the regressions? Test files not listed in any batch at > the bottom of regressions/Makefile.am, e.g. the scoring files, are not > considered part of the official regressions and are not regularly > updated. Ok. I had run once all the tests with my twinpatch, and build a test list from it, where i kept all testfiles with a breakage, without checking what kind of test it was <:o) (there are no genmove there so twin should not break them). That's ok now i just remove this two non official tests from my list.
> > > Is there a tool to automatically adjust this so i can test my > > twin more easyly ? > > If you run the regressions with "make all_batches" or regress.pike, > unofficial test files are excluded. If you're comparing the results to > a non-release CVS version with pre-existing breakage you may find > analyze_results.pike at > http://trac.gnugo.org/gnugo/wiki/RegressionScripts useful. > > /Gunnar > Nice this seems to be what i was looking for. Thanks Alain _______________________________________________ gnugo-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnugo-devel
