From: <tur...@ea...> - 2006-04-27 20:09:04
|
On Thursday 27 April 2006 18:19, Hartwig Deneke wrote: > > Well, I believe getting rid of code obfuscation is always worth it. Also, > code does bit-rot if it is unmaintained. > Examples of this are already present, as ithread.h currently contains > broken defines (ithread_mutexattr_setkind_np vs. pthread_mutexattr_settype > available on my SUSE Linux 9.3), and does miss functionality available in > my current pthreads package (function pthread_testcancel, macro > PTHREAD_MUTEX_INITIALIZER). I must admit that the ithread is a bit broken. It is incomplete (missing above initializers) which cause using programmes to be more complex than necessary. Also, it exports some non-portable definitions (*_np) which are not available everywhere. So I had trouble when porting the library to FreeBSD : some of the _np definitions are not available exactly in the same way, but are used in the "sample" programs. So should I make a compatible implementation of these definitions on FreeBSD ? It would have been better not to export them in the first place ... > > > a user-visible API change that is not good for compatibility. > > The problem of a user-visible API change can be aleviatedby carrying around > the ithread.h file for a while, possibly adding a big warning that it is > going to be removed in the future. > I don't know. What do other people think ? |