Incorrect interpretation of engine search time

[Dan]

Hi all

I compiled a version of Olithink 2.2.1 which didn't originally output duration since the beginning of its search, and I edited it to make it output centiseconds, but when xboard displays the time usage in the engine output window, it doesn't properly transcribe the time to, what seems to me, the normal mm:ss:cs format. (Minutes,Seconds,Centiseconds)

Input:
1. -23 0 3 a6
1& -22 0 5 a5
1& -20 0 10 b5
1& -18 0 13 c5
1& 0 0 15 d6
1& 6 0 17 d5
2. -24 0 61 d5 d4
3. 6 0 340 d5 d4 e5
4. -19 0 860 d5 d4 e5 Nf3
5- -378 2 7657 d5 Qh5 g6 Qe5 Be6 Qxc7
5& -264 4 14282 e5 Qg4 f5 Qxg7 d5
5& -16 10 35546 f5 Qg4 f5 Qxg7 d5
5& 6 11 40245 Nf6 Qg4 f5 Qxg7 d5
6. -5 17 60896 Nf6 f4 d5 d4 Nc6 Nf3
7. 6 47 172920 Nf6 Nc3 Ne4 Nxe4 d5 d4 e5
8. 1 124 529748 Nf6 f4 d5 d3 Nc6 Nc3 Nb4 Nb5

Engine window output:
8 +0.01 529748 2:05.00 Nf6 f4 d5 d3 Nc6 Nc3 Nb4 Nb5
7 +0.06 172920 0:48.00 Nf6 Nc3 Ne4 Nxe4 d5 d4 e5
6 -0.05 60896 0:21.00 Nf6 f4 d5 d4 Nc6 Nf3
5 +0.06 40245 0:17.00 Nf6 Qg4 f5 Qxg7 d5
5 -0.16 35546 0:16.00 f5 Qg4 f5 Qxg7 d5
5 -2.64 14282 0:11.00 e5 Qg4 f5 Qxg7 d5
5 -3.78 7657 0:09.00 d5 Qh5 g6 Qe5 Be6 Qxc7
4 -0.19 860 0:07.00 d5 d4 e5 Nf3
3 +0.06 340 0:07.00 d5 d4 e5
2 -0.24 61 0:07.00 d5 d4
1 +0.06 17 0:07.00 d5

I ran 2 instances to copy the text, so the times don't match exactly, but here 2:05.00 should be actually 2.05 seconds or displayed as 00:02.05 (as well as all the other times being too much)

Have any idea how to fix this or ... ?

[H.G.Muller]

The problem is that the engine prints a character other than space behind the depth. This is a non-compliant form of thinking output (WB protocol prescribes it should start with 4 numbers), used by GNU Chess, and therefore recognized by WinBoard (which is unreasonably forthcoming towards GNU Ches' many non-compliancies)..

But GNU Chess also reports in seconds, rather than centi-seconds. So in this non-compliant format, the time field is interpreted as seconds. So what you see is that WinBoard translates the 124 sent at 8 ply to 2:05, which is about rights. (As you said these were really two different runs.)

The way to fix it is not print that extra character after the depth.

[Dan]

OK, thanks for the info, HG! I will correct this.

Thanks!

Edit: works like a charm!

[Michel]

The problem is that the engine prints a character other than space behind the depth. This is a non-compliant form of thinking output (WB protocol prescribes it should start with 4 numbers), used by GNU Chess, and therefore recognized by WinBoard (which is unreasonably forthcoming towards GNU Ches' many non-compliancies)..

But GNU Chess also reports in seconds, rather than centi-seconds. So in this non-compliant format, the time field is interpreted as seconds. So what you see is that WinBoard translates the 124 sent at 8 ply to 2:05, which is about rights. (As you said these were really two different runs.)

The way to fix it is not print that extra character after the depth.

Note that this is only true for the FSF versions of GnuChess (more specifically 5.07 and 5.08, I don't think 6 suffers from this problem).

My version is compliant (it was one of the first things I fixed).