IBM i Pthreads versus the POSIX standard, the Single UNIX Specification, and other threads implementations

Although the Pthread interfaces described in this document are based on a subset of the APIs defined in the POSIX standard and the Single UNIX^® Specification, the implementation of these APIs is not compliant with these standards. This means that applications written in other versions of threads are not necessarily portable to IBM i. Below is a list of the differences between the Pthread APIs and other threads implementations.

• All thread definitions in pthread.h
• Unsupported preprocessor and feature test macros
• Unsupported Pthread APIs
• Unsupported constants
• Unsupported cancellation points
• Unsupported sysconf() configuration variables
• Thread priority and scheduling
• Thread ID vs. Pthread Handle (pthread_t)
• Thread ID value and size
• Mutexes return EDEADLK when re-locked by owner
• Return values from thread start routines are not integers
• Threads do not necessarily start before pthread_create() returns
• Initial thread is special, cannot pthread_exit()
• Pthread APIs cause asynchronous signals initialization
• Not all jobs can create threads; pthread_create() fails with EBUSY
• Read/write locks are recursive
• Shared read/write locks are released at thread termination
• Read/write locks can be upgraded / downgraded
• Read/write locks do not favor writers
• Spawn API provides more POSIX-like process model
• C++ destructors and Pthread termination
• Unhandled exceptions terminate the thread (not the process)
• Exceptions vs. Asynchronous signals vs. ANSI C signals
• Mutexes can be named to aid in application debug

---

[ Back to top | Pthread APIs | APIs by category ]