xboard 4.7.3 crashes on HGMs ICS

[Volker Annuss]

Yesterday I tried to play some games with Arminius using my linux machine on HGMs ICS before the monthly tournament started. I had to switch to a windows machine because xboard crashed.

xboard is version 4.7.3 on Ubuntu 14.04.

It looks like the bug is already described here: [/url]https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=726247[/url]

Here is a call stack dump from the core file:

Code: Select all

(gdb) where
#0  0x00007fa2a1b38f41 in XtWidgetToApplicationContext () from /usr/lib/x86_64-linux-gnu/libXt.so.6
#1  0x00007fa2a1b3fc6c in XtGetValues () from /usr/lib/x86_64-linux-gnu/libXt.so.6
#2  0x00007fa2a2cfbe02 in ?? ()
#3  0x00007fa2a2ce9e9f in ?? ()
#4  0x00007fa2a2ccc05b in ?? ()
#5  0x00007fa2a1b47817 in XtAppProcessEvent () from /usr/lib/x86_64-linux-gnu/libXt.so.6
#6  0x00007fa2a1b3c89d in XtAppMainLoop () from /usr/lib/x86_64-linux-gnu/libXt.so.6
#7  0x00007fa2a2c9d34d in ?? ()
#8  0x00007fa2a0cddec5 in __libc_start_main (main=0x7fa2a2c9ca30, argc=10, argv=0x7fff616f54c8, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>,
    stack_end=0x7fff616f54b8) at libc-start.c:287
#9  0x00007fa2a2c9d95d in ?? ()


And here are the last lines from the debug file:

Code: Select all

load 8x8 board
parseboard 2, castling = 7 0 4 7 0 4
accepted move Nf6 from ICS, parse it.
moveNum = 2
board = 0-8 x 8
Move parsed to 'Nf6 (0:00)'
45118 >first : time 30000
45118 >first : otim 30000
book hit = (NULL)
45118 >first : usermove 45118 >first : g8f6
nps: w=-1, b=-1
Display title 'Arminius (39) gegen Lozza (39) {5 1}, gameInfo.variant = 0'
45228 <first : move g1f3
>ICS: g1f3\015\012
>ICS: kibitz !!! +0,00/0 (-0,01 sec, 0 nodes, 0 knps) PV=\015\012
nodes = 0, 0
<ICS: \012\015Lozza(1544)(C) kibitzes: !!! -0.08/14 (25.01 sec, 11693898 nodes, 468 knps) PV=Nf6 Nc3 e6 e4 Bb4 Bd3 Nc6 Nf3 d5 exd5 Nxd5 O-O Nxc3 bxc3\012\015fics% \012\015<12> rnbqkb-r pppppppp -----n-- -------- ---P---- -----N-- PPP-PPPP RNBQKB-R B -1 1 1 1 1 2 2 Arminius Lozza -1 5 1 39 39 300 300 2 N/g1-f3 (0:00) Nf3 0 1\012\015fics% kibitzed to 1.\012\015fics%
ics input 3, castling = 7 0 4 7 0 4


Strange about this bug is, that xboard always crashes after 1 or 2 moves when playing with Arminius (local) against Lozza (remote) or against some other engines I don't remember. On the other hand I could successfully play some games against other engines, for example rpiFruit. I also could play games against Lozza with a Windows machine using winboard 4.7.0.

[H.G.Muller]

I remember there was a problem (reported by the author of ZuriChess) when the opponent was kibitzing, and the window where the kibitz was supposed to go was not open. IIRC the problem manifested itself when autoKibitz was on, so probably the Engine Output window has to be open to avoid it. Apparently a recent patch does not properly test for that, acessing a non-existent widget as a result, and I never noticed in testing, as I always run with that window open.

Note that rpiFruit does not kibitz...

I will have a look at it.

[Volker Annuss]

Thank you for looking.

so probably the Engine Output window has to be open to avoid it.

I am quite sure the engine output window was always open.

[H.G.Muller]

I just fixed a very similar error reported by the ZuriChess author, where the Xaw build did crash after receiving a kibitz message from the opponent. This could be traced to an attempt changing the color of a button in a non-existent dialog. After he opened the ICS Console/Chat window the crash went away. I now pushed a version to the master branch of the hgm.nubati.net repository that tests for the existence of the dialog.

I am not 100% sure the error you suffered is the same, though. In the Ubuntu bug report the crash also happens after receiving a kibitz, but it occurs in the X-server routine XtGetValues instead of XtSetValues, so it was attempting to do something else in response to the kibitz. (But that could be appending the text to a memo rather than coloring a button.) In your case the crash does not seem to occur on receiving a kibitz message from the ICS.

Can you test the new version to see if the fix also solves your problem?