Change Details

The card deck is shuffled in shuffled() (in dealer.cpp) using a random number generator that, according to the comments above, is identical to the Freecell one. I'm not familiar with the details of the Windows implementation, but the kpat implementation uses arithmetic operators on signed values which according to the C11 standard (quoted at https://wiki.sei.cmu.edu/confluence/display/c/INT13-C.+Use+bitwise+operators+only+on+unsigned+operands) can give implementation defined results. What I have observed when recompiling kpat recently (possibly after a switch to GCC 7.3.0) is the 'seed' value after a few iterations of the loop becoming negative due to integer overflow. Then 'rand' also becomes negative (despite the '& 0x7fff'!), 'rand % i' again is negative and the result.swap() below segfaults. Changing the type of the 'seed' variable to unsigned appears to avoid the problem, although I don't know whether this will be incompatible with the Freecell implementation.

The card deck is shuffled in shuffled() (in dealer.cpp) using a random number generator that, according to the comments above, is identical to the Freecell one. I'm not familiar with the details of the Windows implementation, but the kpat implementation uses bitwise operators on signed values which according to the C11 standard (quoted at https://wiki.sei.cmu.edu/confluence/display/c/INT13-C.+Use+bitwise+operators+only+on+unsigned+operands) can give implementation defined results. What I have observed when recompiling kpat recently (possibly after a switch to GCC 7.3.0) is the 'seed' value after a few iterations of the loop becoming negative due to integer overflow. Then 'rand' also becomes negative (despite the '& 0x7fff'!), 'rand % i' again is negative and the result.swap() below segfaults. Changing the type of the 'seed' variable to unsigned appears to avoid the problem, although I don't know whether this will be incompatible with the Freecell implementation.

The card deck is shuffled in shuffled() (in dealer.cpp) using a random number generator that, according to the comments above, is identical to the Freecell one. I'm not familiar with the details of the Windows implementation, but the kpat implementation uses arithmeticbitwise operators on signed values which according to the C11 standard (quoted at https://wiki.sei.cmu.edu/confluence/display/c/INT13-C.+Use+bitwise+operators+only+on+unsigned+operands) can give implementation defined results. What I have observed when recompiling kpat recently (possibly after a switch to GCC 7.3.0) is the 'seed' value after a few iterations of the loop becoming negative due to integer overflow. Then 'rand' also becomes negative (despite the '& 0x7fff'!), 'rand % i' again is negative and the result.swap() below segfaults. Changing the type of the 'seed' variable to unsigned appears to avoid the problem, although I don't know whether this will be incompatible with the Freecell implementation.