From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; ja-JP; rv:1.7.8) Gecko/20050511 Firefox/1.0.4 Description of problem: Info document in glibc (libc.info) describes LinuxThreads. But when running or compiling a program, NPTL is used instead by default. There are perhaps some inappropriate explanations against NPTL. That's too misleading for general users. A few things I noticed were: . Cancellation points are too limited and eg. select is not a cancellation point. (on NPTL, cancellation points are conformed to POSIX standard and almost all syscalls are cancellation points). . pthread_cond_timedwait may return EINTR (one in LinuxThread also didn't return EINTR since glibc-2.1.3 though). Version-Release number of selected component (if applicable): glibc-2.3.4-2.9 How reproducible: Always Steps to Reproduce: 1. Invoke info libc. 2. Search Cancellation or pthread_cond_timedwait. Additional info:
For NPTL, the documentation is at http://www.opengroup.org/onlinepubs/009695399/nfindex.html or in the 3p section of manual pages. info libc thread section simply documents LinuxThreads, so you shouldn't be using it when writing NPTL (or portable POSIX threads) programs.
I understand what you are saying. But is there any documentation a general user can see at a glance that he or she should't consult info.libc about threads?
Use the POSIX man pages we ship (as Jakub pointed out) or buy a book on POSIX threads. Nobody has volunteered to write decent documentation for the upstream glibc so there is nothing for us to import. If you have a RHEL license and insist of RH providing such documentation then talk to your Red Hat representative.