Ok, I have run the version from the financial hacker blog and also the last one posted here where the WinPayout and LossPayout variables are set inside the loop.
I did set(LOGFILE) so that log files are beeing produced.
I have uploaded the full log files here:
Log.zip I have compared the train logs. They are identical. I have compared a few generated parameter files. They are identical.
I made a file compare with the Windows utility program FC file1 file2 between the test logs as they are different and attached the result to this post.
The working test log has a filesize of 17 287 150 bytes while the buggy version log file has a size of 17 096 468 bytes.
There are a lot of price differences like so:
***** BinaryBugtest.log
[98749: Tue 08.05.12 20:35] (1.30068)
[EUR/USD::L4901] Long 1@1.30064 at 20:35:00
***** BINARYWORKINGTEST.LOG
[98749: Tue 08.05.12 20:35] (1.30068)
[EUR/USD::L4901] Long 1@1.30063 at 20:35:00
*****
***** BinaryBugtest.log
[98750: Tue 08.05.12 20:40] +0 +1 1/0 (1.30064)
[EUR/USD::L4901] Sell 1@1.30063: -1.00 at 20:40:00
***** BINARYWORKINGTEST.LOG
[98750: Tue 08.05.12 20:40] +0 +1 1/0 (1.30064)
[EUR/USD::L4901] Sell 1@1.30059: -1.00 at 20:40:00
And maybe some other differences but there were too many of them so the tool did not show all of them.