Changing directories in bash can be tedious if you have long names or nested paths. Creating aliases or adding to the CDPATH can help, but can be improved on. Bashcd adds 6 new commands to make changing directories a bit easier. This commands use find, the locate database, the mdfind database, or other contextual information to make it easier to change to other directories.
OnPosix is a tiny library to abstract POSIX mechanisms to C++ developers. Most features offered by this library can be found either inside the Boost library or in a library compliant with the C++11 standard. Unfortunately, however, for some embedded Linux devices, these libraries cannot represent viable solutions, due to the lack of memory space (for the Boost libraries) and the lack of a new C++ compiler (e.g., on Xilinx MicroBlaze). On these platforms, the OnPosix library represents a good and cheap solution to have object-oriented POSIX mechanisms. The library offers support for threads, mutual exclusion, sockets, logging, timing, etc.
cQRCode is a class that generates QR code images in GIF, JPEG, and PNG formats. It takes a string of information to encode and can either serve the image with the current script output or save it to a file. The size of matrix points and the error correction level parameters are configurable.
ashttpd was originally a testbed for a kernel async-sendfile patch. Today, it's just another Web server for serving static content rapidly. It differs from its competitors by storing its webroot in a database, which provides a 50% performance increase over using the operating system's VFS. It also allows precalculation of strong cache-validators (ETags).
cola is a C implementation of the COLA structure described in the paper "Cache Oblivious Streaming B-Trees" by Bender, Farach-Colton, et al. This algorithm is a drop-in replacement for B-Tree databases/indexes which performs faster on spinning disks than B-Trees do on SSD. It's asymptotically optimal in terms of data transfers to and from disk. For example, random key inserts are turned into linear sequential writes while maintaining the property which queries always require, at most, precisely log N blocks read sequentially (but not contiguously) from disk (or half that, on average).