xboard Problem with multi pv analysis

[BBauer]

Trying some multi pv analysis with xboard shows only the last level.
I would like to see all analysis just like arena.
BTW, is the super issue solved?
Kind regards
Bernhard

[hgm]

I think this is Polyglot issue. Recent Polyglots do not clear the Engine-Output window between iterations. The Superchess problem should be solved, at least w.r.t. occurrence of the word 'super'. XBoard will now not go on the event tag for any variant that was introduced after I implemented an explicit variant tag. Event names line "crazyhousewives championship", "Tourney at Arnocastle, VA." etc. will continue to give the same problems as they would in 4.2.7. The patch is in the savannah repository.

[BBauer]

Ok, it is a polyglot problem, but I do not have it in console mode.
I have those ugly problems only with xboard.

The other patch is somewhere. I got

Suchergebnisse für xboard (in Patch):
Objekt-ID Objektzusammenfassung Gruppe Eingesandt von Datum
#6757 Titanic size default doesn't fit XBoard rpeterso Fr 27 Feb 2009 05:10:10 UTC
#6182 Add DESTDIR support to the Makefile, XBoard 4.2.7 XBoard None Mo 03 Sep 2007 21:50:55 UTC
#6108 Preliminary dgt-support for xboard XBoard None Fr 20 Jul 2007 12:20:07 UTC

Perhaps I should give up xboard.
Sorry
Bernhard

[hgm]

Ok, it is a polyglot problem, but I do not have it in console mode.
I have those ugly problems only with xboard.

The problems occur because Polyglot re-sends PVs of lower depth after it has aready sent PVs of a higher depth. For XBoard this is a signal to clear the EngineOutput window. It would not make sense to re-send lines unless you want to clear the window. Why would you want a line printed that was already in the display?

So engines should either send all lines only once, in ascending order of depth, so they all remain displayed (this is presumably the mode you want), or they should send them in batches low to high, each batch clearing the window before it is displayed, because it starts with a lower depth as that was already seen. (The engine can send dummy empty lines of low depth to force the clearing in an invisible way.)

I am pretty sure that you will not see a monotonous increase in the PVs sent by Polyglot in console mode. So Polyglot requests the display to be cleared many times during the search, and XBoard dutifully obeys. If that is not what you want, then don't let Polyglot request it...

The other patch is somewhere. I got

Suchergebnisse für xboard (in Patch):
Objekt-ID Objektzusammenfassung Gruppe Eingesandt von Datum
#6757 Titanic size default doesn't fit XBoard rpeterso Fr 27 Feb 2009 05:10:10 UTC
#6182 Add DESTDIR support to the Makefile, XBoard 4.2.7 XBoard None Mo 03 Sep 2007 21:50:55 UTC
#6108 Preliminary dgt-support for xboard XBoard None Fr 20 Jul 2007 12:20:07 UTC

Perhaps I should give up xboard.
Sorry
Bernhard

I don't understand what you are doing here. Are you using the official sources from http://git.savannah.gnu.org/cgit/xboard.git (in particular the master-20100221 tar ball)?

[BBauer]

Here is what polyglot does in console mode for mulipv=3

Code: Select all

12 -133 752 3926793 Re2 Qxe2 Bxe2 bxa6 
12 -141 752 3926793 axb5 Ne4 Bxe4 fxe4 
12 -246 752 3926793 cxb5 Ne4 Bxe4 fxe4 
0 -121 889 5260502 
13 -121 889 5260502 Re2 Qxe2 Bxe2 bxa6 
0 -117 933 5861157 
13 -117 933 5861157 axb5 Ne4 Bxe4 fxe4 
13 -121 933 5861157 Re2 Qxe2 Bxe2 bxa6 
0 -117 986 6625931 
13 -117 986 6625931 axb5 Ne4 Bxe4 fxe4 
13 -121 986 6625931 Re2 Qxe2 Bxe2 bxa6 
13 -242 986 6625931 cxb5 Ne4 Bxe4 fxe4 
[code]

Arena can do it.

My last part was related to your sentence:[quote]
The patch is in the savannah repository[/quote]

In other words I could not find thw patch.

kind regards
Bernhard

[Michel]

The latest versions of PG do not clear the display anymore.
HGM provided a patch for that. The output in console mode now looks like this.

Code: Select all

2 +11 0 41 Nf3 Nf6
2 +7 0 62 Nc3 Nf6
2 +0 0 84 e4 Nc6
2 -5 0 105 d4 Nf6
3 +60 0 224 Nf3 Nf6 Nc3
3 +60 0 296 Nc3 Nf6 Nf3
3 +45 0 417 e4 Nf6 Nc3
3 +41 0 505 d4 Nf6 Nf3
4 +11 0 649 Nc3 Nf6 Nf3 Nc6
4 +11 0 753 Nf3 Nf6 Nc3 Nc6
4 -3 0 857 e4 Nf6 Nc3 Nc6
4 -1 1 974 d4 Nf6 Nf3 Nc6
5 +43 1 1916 Nc3 Nf6 Nf3 Nc6 d4
5 +43 1 2145 Nf3 Nf6 Nc3 Nc6 d4
5 +43 1 2681 d4 Nf6 Nf3 Nc6 Nc3
5 +17 1 3011 e4 Nf6 e5 Nd5 d4
5 +29 1 3622 e3 Nf6 Nf3 Nc6 Nc3
6 +11 1 4857 d4 Nf6 Nf3 Nc6 Nc3 d5
6 +11 2 5446 Nc3 Nf6 Nf3 Nc6 d4 d5
6 +11 2 5934 Nf3 Nf6 Nc3 Nc6 d4 d5
6 +0 2 6733 e3 Nf6 Nf3 Nc6 Nc3 d5
6 +15 2 7065 e4 Nf6 e5 Nd5 Nc3 Nxc3 dxc3
7 +15 2 10929 e4 Nf6 Nc3 Nc6 Nf3 d5 Bd3
7 +29 2 13202 d4 Nf6 Nf3 Nc6 Nc3 d5 Bf4
7 +29 2 14128 Nc3 Nf6 Nf3 Nc6 d4 d5 Bf4
7 +29 2 14774 Nf3 Nf6 Nc3 Nc6 d4 d5 Bf4
7 +31 2 18208 e3 Nf6 Nf3 e6 Nc3 Nc6 Bd3
8 +23 3 22738 e3 Nf6 Nf3 e6 d4 d5 Bb5+ c6 Bd3
8 +23 3 31159 d4 Nf6 e3 d5 Nf3 e6 Bb5+ c6 Bd3
8 +11 3 33783 Nc3 Nf6 Nf3 Nc6 d4 d5 Bf4 Bf5
8 +17 3 35620 Nf3 Nf6 d4 Nc6 c4 e6 Nc3 Bd6

[hgm]

In other words I could not find thw patch.

Well, obvously. But what were you loking for, and where? Why don't you simply load the most recent version of the master of the master branch? That includes all patches.

The patch you are looking for is called "Do not recognize non-ICS variants from PGN event tag", and is currently still listed on the overview page of the reposiory, at the very bottom. No idea what one would have to query to get the response that you posted. The patches that you list seem to have no logical connection to each other or to the tag-recognition problem. In fact I doubt if they are even in the Savannah repository: I get no matches at all when I search for 'Titanic' there.

Code: Select all

12 -133 752 3926793 Re2 Qxe2 Bxe2 bxa6 
12 -141 752 3926793 axb5 Ne4 Bxe4 fxe4 
12 -246 752 3926793 cxb5 Ne4 Bxe4 fxe4 
0 -121 889 5260502 
13 -121 889 5260502 Re2 Qxe2 Bxe2 bxa6 
0 -117 933 5861157 
13 -117 933 5861157 axb5 Ne4 Bxe4 fxe4 
13 -121 933 5861157 Re2 Qxe2 Bxe2 bxa6 
0 -117 986 6625931 
13 -117 986 6625931 axb5 Ne4 Bxe4 fxe4 
13 -121 986 6625931 Re2 Qxe2 Bxe2 bxa6 
13 -242 986 6625931 cxb5 Ne4 Bxe4 fxe4 

From the Polyglot output you can see that it is as I diagnosed. The Polyglot you are using inserts dummy lines of 0 depth, for no other reason than to instruct the GUI to clear the display. Of course XBoard obeys that. If you don't want the display cleared, you should not send clear commands...

[BBauer]

It simply dosn't work.
I downloaded
master-20100221 master-20100221.tar.gz
and installed it, but
xboard: Variant super not supported by Stockfish 1.6.3s(dc) 64bit


Perhaps you could be more specific.
What should I install, where can I get it.
Installing something that doesn't work is not funny.
Kind regards
Bernhard

[hgm]

Sorry, this was my mistake, I should not have mentioned the tar ball. This is from Feb 21, and the event-tag patch was made after that. So you really should load the most recent commit directy from git. (Or in any case one later than the event-tag commit, but there is no reason to leaave out any of the other fixes.)

To do that, go to the page I mentioned ( http://git.savannah.gnu.org/cgit/xboard.git ), the section just below the tar balls. (For which the left column header says "Age"). Click there on the most recent commit message, tagged "master" (currently "added logo files for xboard"). Then you get to a page where near the top there are four links with very long gibberish names. Clicking the lowest one ( "download") should give you the tar ball for the sources after that commit.

You could do the same for the v4.4.x branch (after clicking v4.4.x in the "branch" column left in the upper section of the main page), if the master version is too modern for your taste. Beware, though, that master has a little better multiPV support than v4.4.x: it sorts the lines with the same depth by score, even if the engine sends them in another order. v4.4.x always displays them in the order the engine sent them.

[BBauer]

Thank you.
Now I can view and analyse the linares games.

Multipv behavior has not changed.
From "about Xboard" I wanted to copy the version but that is not possible.
Cut and paste seams not to work for me.
How do you work with xboard?
Anyway
kind regards
Bernhard