A unique bug probably never before seen

[fivemack]

If you are sieving, say, 8^225+7^225 and 8^240+7^240 (IE two numbers with the same algebraic polynomial), it is very important to set the N you're factoring, not only consistently in msieve.fb, worktodo.ini and the top of msieve.dat, but also to the number you're trying to factor rather than to the other one.

Otherwise msieve finishes the linear algebra and then starts complaining the dependencies aren't products of squares.

[jasonp]

Originally Posted by fivemack

If you are sieving, say, 8^225+7^225 and 8^240+7^240 (IE two numbers with the same algebraic polynomial), it is very important to set the N you're factoring, not only consistently in msieve.fb, worktodo.ini and the top of msieve.dat, but also to the number you're trying to factor rather than to the other one.

Otherwise msieve finishes the linear algebra and then starts complaining the dependencies aren't products of squares.

Thank you! In fact I had seen this problem before, over a year ago, and had no idea what was going on. The code checks that the input value of N with small factors removed matches the value of N in the factor base file, but it does not verify that plugging the rational poly coefficients into the algebraic polynomial actually produces a multiple of N.

I'll add code to do that and a few other checks too.