Engine output bugs in analysis mode+error message+request

Discussions about the WinBoard protocol. Here you can also report bugs and request new features.

Moderators: hgm, Andres Valverde

Engine output bugs in analysis mode+error message+request

Postby velocidrom » 11 Feb 2014, 18:20

I was analyzing a position with stockfish (8/7p/5P1k/7P/8/4p1p1/4P1P1/3K2b1 w - - 0 1).
Bug: The Engine Output will display the tbhits and hide the time when I left click on them, as expected. However it would never updates the "not shown: tbhits" and the time would still appear as shown while it was disabled. In other words the line "dep score nodes time (not shown: tbhits knps seldep)" is never updated while the real displayed values are updated as I want.

Second bug, as you can see in the screenshot, I received the message
Code: Select all
Gtk-Message: (for origin information, set GTK_DEBUG): failed to retrieve property `GtkRange::activate-slider' of type `gboolean' from rc file value "((GString*) 0x2a41560)" of type `GString'

and now the Engine Output is never updated anymore, stockfish "stuck" at depth 59 on the position which is in fact not true (100% cpu usage and many minutes stuck on that move.). To me it seems that the engine output has stopped to updates itself when I got the error message in the terminal.

Request: I would like to show tbhits by default and hide time by default. I have done what you (H.G.M.) suggested me on talkchess (i.e. modify engineoutput.c and change line 95 from
Code: Select all
static int  columnMask = 0xF0;
to
Code: Select all
static int  columnMask = 0xE0;
), recompiled. But what it seems to do is that tbhits are still disabled by default and now require a double-click on them to display them instead of a single click.
Also I do not know how to hide the time by default, in the Engine Output.

Another request: Would it be possible to allow analysis mode in FICS when examining a game? Right now it is only possible to use analysis mode when observing a game. IMO it would be extremely helpful (to me at least) if it was possible to use the analysis mode when examining a game.

Thank you!

Link to screenshot: http://imgur.com/zzA4jeB
velocidrom
 
Posts: 44
Joined: 25 Sep 2012, 01:20

Re: Engine output bugs in analysis mode+error message+reques

Postby H.G.Muller » 15 Feb 2014, 11:30

velocidrom wrote:I was analyzing a position with stockfish (8/7p/5P1k/7P/8/4p1p1/4P1P1/3K2b1 w - - 0 1).
Bug: The Engine Output will display the tbhits and hide the time when I left click on them, as expected. However it would never updates the "not shown: tbhits" and the time would still appear as shown while it was disabled. In other words the line "dep score nodes time (not shown: tbhits knps seldep)" is never updated while the real displayed values are updated as I want.

Does "never" here also include the case when you start a new search? That it does not update the line during the current search is normal. It also does not update the already-printed output lines of lower depth.

Second bug, as you can see in the screenshot, I received the message
Code: Select all
Gtk-Message: (for origin information, set GTK_DEBUG): failed to retrieve property `GtkRange::activate-slider' of type `gboolean' from rc file value "((GString*) 0x2a41560)" of type `GString'

and now the Engine Output is never updated anymore, stockfish "stuck" at depth 59 on the position which is in fact not true (100% cpu usage and many minutes stuck on that move.). To me it seems that the engine output has stopped to updates itself when I got the error message in the terminal.

Well, in the message field above the board it also gives the same line as the most-recent one in the Engine Output pane, so if Stockfish did send more, XBoard must have stopped completely. Was it still responsive to user input in thi ssituation?

The GTK error message is strange. I suppose this only is printed rarely? I don't expect that the reported error could hang XBoard. It seems to be an internal GTK problem, where it tries to operate a scroll bar to get the displayed text in view (so presumably in the Engine Output window). I could imagine that an error here shuts off the displaying in the Engine-Output window, but it should never be able to affect the message field above the board.

Request: I would like to show tbhits by default and hide time by default. I have done what you (H.G.M.) suggested me on talkchess (i.e. modify engineoutput.c and change line 95 from
Code: Select all
static int  columnMask = 0xF0;
to
Code: Select all
static int  columnMask = 0xE0;
), recompiled. But what it seems to do is that tbhits are still disabled by default and now require a double-click on them to display them instead of a single click.
Also I do not know how to hide the time by default, in the Engine Output.

Well, I would have to try it out myself then. The columnMask contains one bit per item, which is set to 1 if the corresponding item should be hidden. The default setting F0 is 11110000 in binary, and thus hides 4 items and displays 4 others. By changing the initially set bits you can control that. But it could be that the initial header line should also be modified to make this work properly.

Another request: Would it be possible to allow analysis mode in FICS when examining a game? Right now it is only possible to use analysis mode when observing a game. IMO it would be extremely helpful (to me at least) if it was possible to use the analysis mode when examining a game.

I will have a look at this. The ICS-analysis code was written by Daniel Mehrmann, and I have no idea how it works. So I cannot say off-hand if this will be a trivial change (like just refusing to do it in fewer cases) or that it would involve much more.
User avatar
H.G.Muller
 
Posts: 3080
Joined: 16 Nov 2005, 12:02
Location: Diemen, NL

Re: Engine output bugs in analysis mode+error message+reques

Postby velocidrom » 06 Mar 2014, 19:17

Sorry for the long delay H.G.M.
I would like to point out that I was wrong with the fix that you provided for tbhits to appear by default. They do appear by default with your suggestion, although it shows "dep score nodes time (not shown: tbhits knps seldep)".



Another totally unrelated "bug" (I think it is a bug but I may be wrong) is that very rarely it happens that Stockfish thinks it is its turn to play when it is not. So when it is pondering, it is pondering on an impossible move (and its eval is generally very high, because it wants to play 2 moves in a row!). I cannot reproduce the bug. It happens probably in less than 10,000 moves, but since I have seen it 2 to 4 times already (including today), I felt like mentioning it.
velocidrom
 
Posts: 44
Joined: 25 Sep 2012, 01:20


Return to WinBoard development and bugfixing

Who is online

Users browsing this forum: No registered users and 1 guest

cron