Wesley Morgan wrote: > On Wed, 9 Oct 2002, Alex Zepeda wrote: > > I'm quite suprised how this managed to avoid detection for so long > > (webcvs.kde.org indicated that the last commit to this branch was > 12 > > months ago). It's that whole feeling of fixing the same problems over and > > over again. > > Most likely this is a problem that was uncovered by some change in our > headers. I've seen a lot of commits come down the pipes removing > "unneeded" includes from one header or another... Which of course is the > correct thing to do and it encourages better coding practices to make > people put headers in the right order.
There are two issues. The first is that in order to be POSIX 1003.1-2001 compliant, we have to get rid of some promiscuous includes that are remaining. The second is that in order to be POSIX 1003.1-2001 compliant, some headers are required to define their own types for use in structure declarations. This is like the move from Draft 4 pthreads conformance to standard pthreads conformance, which defined PTHREAD_MTEX_INITIALIZER, but didn't make the rest of the required conformance changes at the same time: many programs (e.g. OpenLDAP) used PTHREAD_MUTEX_INITIALIZER as one of the few manifest constants that was available to distinguish between these two implementations (other code included the Moscow center for supercomputing activities Standard Template Library, and by extension, the ACAP reference implementation). The bottom line is that it is better to be fully conformant with an old standard, than partially conformant with a new standard. It's unfortunate that these type of changes that move FreeBSD away from that maxim are going in this close to the 5.0 release date; they are likely to break a *lot* of code (IMO). -- Terry To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
