yuck is a command line option parser for C that works on a minimal set of dependencies - only a C compiler and the m4 macro processor are required. It supports all the standard use cases: GNU-style long options (--option), condensable short options (-xab for -x -a -b), and optional arguments to long and short options (--foo[=BAR]), multiple occurrence of options (-vvv). Most importantly, it does not depend on libc's getopt() nor getopt_long().
MLPACK is a C++ machine learning library with an emphasis on scalability, speed, and ease-of-use. Its aim is to make machine learning possible for novice users by means of a simple, consistent API, while simultaneously exploiting C++ language features to provide maximum performance and maximum flexibility for expert users. It contains algorithms such as k-means, Gaussian mixture models, hidden Markov models, density estimation trees, kernel PCA, locality-sensitive hashing, sparse coding, linear regression and least-angle regression.
ENet for C# wraps the C-language ENet networking library. For games and other realtime applications, if you use TCP, old data can hold up newer data, even if it is no longer relevant (old positions, etc.). Your players will perceive any lost packet as a "lag burst". ENet supports multiple in-order streams of data, and allows you to decide on reliability on a per-packet basis. The library is useful for both client-server and peer-to-peer architectures.
Obix is an object-oriented programming language designed to make it easy to quickly write reliable code. More reliability is achieved through language features which consistently support the "Fail fast!" principle (every coding error should be detected as early as possible, preferably at compile-time, or else as early as possible at run-time). The Obix compiler generates Java binaries which can be executed on any Java virtual machine.
exceptions4c is a tiny, portable framework that brings the power of exceptions to your C applications. It provides a simple set of keywords (macros, actually) which map the semantics of exception handling you're probably already used to: try, catch, finally, throw. You can write try/catch/finally blocks just as if you were coding in Java. This way you will never have to deal again with boring error codes, or check return values every time you call a function.
SHTns is a high-performance Spherical Harmonic Transform library. It was designed for numerical simulation (fluid flows, mhd, etc.) in spherical geometries, but can be used for any kind of problem involving scalar or vector spherical harmonics. It is very fast, thanks to careful vectorization and runtime tuning. It supports multi-threaded transforms via OpenMP. It features scalar and vector transforms, synthesis and analysis, and flexible truncation and normalization. A Python interface is included.