On Sun, 8 Nov 2009, Samuel Thibault wrote: > Joseph S. Myers, le Sun 08 Nov 2009 00:42:29 +0000, a écrit : > > On Fri, 6 Nov 2009, Samuel Thibault wrote: > > > There is a small error in the pthread_mutex_unlock manpage: as required > > > by the POSIX norm, recursive mutexes do check that its caller is the > > > owner of the lock. > > > > What system are you using that supports Linuxthreads with current EGLIBC > > (so supports TLS) but not NPTL? > > I'm not. But Debian uses the linuxthreads manpages for thread stuff.
I think doing so, and taking them from an almost-obsolete legacy part of EGLIBC, is a mistake. The primary documentation of EGLIBC is the Texinfo manual, which unfortunately lacks pthreads documentation. Neither EGLIBC nor FSF GLIBC generally maintains manpages for the contained functions; instead, there is a separate project that does so, the Linux Man-pages Project. It really would be better for these pages to be adopted by that project and maintained there. I'll apply the patch as it appears to be correct, but I strongly advise finding another home to maintain manpages that cover both Linuxthreads and NPTL, and suggest the Linux Man-pages Project as that home. (The timezone data is a successful case of something being imported into GLIBC without being maintained there. It would be good if locale data were maintained externally on a similar basis, and it's not clear all the miscellaneous utilities installed by (E)GLIBC need to come with the library either (for example, because of being closely tied to private datastructures or being needed to run the testsuite); (E)GLIBC should keep its focus on the actual libraries and things closely tied up with them.) -- Joseph S. Myers jos...@codesourcery.com