[trestoncuzzort]

Fixes #63.

negative_binomial_distribution(0, 0.5) and (10, 1) satisfy the documented preconditions but abort in the gamma constructor. (0, 0) divides by zero before reaching that constructor.

Return zero directly when the target count is zero or the success probability is one. These cases need no random draws. For a positive target, clarify that the success probability must be strictly positive, and assert that generation precondition before dividing. The interior gamma-Poisson path is unchanged.

Extend the existing tests across signed/unsigned integer and floating representations, direct/parameter construction, parameter setters, explicit-parameter generation and stream round trips. Check both the zero results and unchanged engine state.

Validation on Linux:

• GCC 16, C++17: the complete B2 test target passes all 123 executables before and after, including its Release multiprecision-float target. Every runtime status/output was checked.
• GCC C++11 and Clang 23 C++17 ASan/UBSan with float-division/cast checks and leak detection: the affected executable passes all nine cases and 1,607 assertions. Each new boundary case fails against the original header.
• 387,072 interior samples and 1,512 complete engine states match the frozen original implementation byte for byte under UBSan. This checks sequence preservation, not an independent statistical proof.
• A separate assertion-handler probe confirms that a positive target with zero probability is caught before division.
• git diff --check passes. Native Windows/32-bit builds and generated documentation were not tested.

OpenAI Codex assisted with investigation, implementation and tests. dawnr checked the boundary guard's gamma-parameter conditions using an exact rational abstraction and seven provers; that does not prove floating-point sampling or the generic/compiled C++ implementation.

[mborland]

Looks good to me. Once the CI runs green I will merge. Thank you!

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.21%. Comparing base (15a0968) to head (117e344).

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #169      +/-   ##
===========================================
+ Coverage    96.18%   96.21%   +0.02%     
===========================================
  Files          129      130       +1     
  Lines         8446     8504      +58     
===========================================
+ Hits          8124     8182      +58     
  Misses         322      322              

Files with missing lines	Coverage Δ
...de/boost/random/negative_binomial_distribution.hpp	100.00% <100.00%> (ø)
test/test_negative_binomial_distribution.cpp	100.00% <100.00%> (ø)

---

Continue to review full report in Codecov by Harness.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 15a0968...117e344. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.