Fonzy Bluemers Polyglot Update - 1.4w27

[Charles Browne]

Polyglot 1.4w27 (27 March 2010)

As winboard sends the info strings to fics/ics as well and this can cause unwanted side-effects, polyglot does not send them to winboard unless InfoStrings is set true.
new MultiPVall bool. If set true will show the multipv lines from previous depths too. Otherwise it will only show the pv lines from the current depth


http://www.geenvis.net/pg.html



I'm not actually watching Fonzy's page - honest. I am reinstalling some programs and needed his addition of Polyglot, so I go to his download page and a newly released version.

Already announced on Leo Dijksman's site.

http://wbec-ridderkerk.nl/html/newspage.htm

[H.G.Muller]

t seems to me that Polyglot is trying to fix something that should be fixed in WinBoard. If there is really something wrong, that is. Why would WinBoard send info strings to FICS? As autoKibitz? That should only happen if the last thinking output sent by the engine before the move was in fact an info string rather than normal thinking output. And Polyglot should easily be able to prevent that happening. Not sending the info strngs at all seems a needlessly draconic measure for that. But I think it would be better to count it as a WinBoard bug.

[F. Bluemers]

t seems to me that Polyglot is trying to fix something that should be fixed in WinBoard. If there is really something wrong, that is. Why would WinBoard send info strings to FICS? As autoKibitz? That should only happen if the last thinking output sent by the engine before the move was in fact an info string rather than normal thinking output. And Polyglot should easily be able to prevent that happening. Not sending the info strngs at all seems a needlessly draconic measure for that. But I think it would be better to count it as a WinBoard bug.

Winboard sends them because polyglot shapes them into pv strings.
I guess it has no choice but to send them to the server.
With hindsight,I think the whole idea of sending info strings to winboard(iirc: only for one engine,chestuci)was a bad idea.

[H.G.Muller]

But WinBoard does not send every PV it gets to the ICS. Without -autoKibitz it never sends any PV to the ICS, and with -autoKibitz, only the last line it received before the move. If it happens that the last line of thinking output received was an info line rather than a genuine PV, WinBoard should base the kibitzed line on the genuine PV. If it doesn't, this counts as a bug in the -autoKibitz implementation.

Engines don't print info lines for nothing. They intend them to be shown to the user. WB protocol did not need a special command for this, because the PV field of Thinking Output was defined free format especially with this in mind. The way we let Polyglot send the info lines to WB (with time = nodes = score = 0) should make them easily to recognize by WinBoard, and indeed WinBoard already tests for this to alter the layout of such a line in the Engine-Output display (suppressing the dummy zeros). This same test should also disqualify the line for the purpose of -autoKibitz.

From looking at the code, it seems that info lines could also erroneously be used for deriving the /pgnExtendedInfo. I fixed that now, and the fix will go into the 4.4.3 release, which we hope to do this week.