I did a test for the negative binomial distribution with parameters k = 1 and p = {-0.5, 0.0, 0.5, 1.0, 1.5}.
Constructor:
fail: Constructor calls abort
pass: Constructor passes without any problems but mathemtically this distribution is not defined
Ok: Constructor passes and parameters are valid

Sampling:
fail: Sampling from a constructed negative binomial distribution object fails (mostly due to parameter problems while constructing the poisson (in std case) or gamma (in boost case) distribution
wrong: Sampling creates a value which is wrong. Either always zero or -2^31
Ok: Everything is fine.


STD random (-D_GLIBCXX_DEBUG)
p =         | -0.5 | 0.0  | 0.5 | 1.0  | 1.5  |
------------+------+------+-----+------+------+
Constructor | fail | fail | Ok  | pass | fail |
Sampling    |  --  |  --  | Ok  | fail |  --  |
Value       |  --  |  --  | Ok  |  --  |  --  |

BOOST random
p =         | -0.5 |  0.0  | 0.5 | 1.0  | 1.5  |
------------+------+-------+-----+------+------+
Constructor | pass | pass  | Ok  | pass | pass |
Sampling    | fail | wrong | Ok  | fail | fail |
Value       |  --  | -2^31 | Ok  |  --  |  --  |

STD random
p =         | -0.5  | 0.0   | 0.5 | 1.0   | 1.5   |
------------+-------+-------+-----+-------+-------+
Constructor | pass  | pass  | Ok  | pass  | pass  |
Sampling    | wrong | wrong | Ok  | wrong | wrong |
Value       |  0    | -2^31 | Ok  |  0    |  0    |