Re: Remove __FAVOR_BSD

["Michael Kerrisk (man-pages)" <mtk dot manpages at gmail dot com>]

On 01/07/14 03:08, Joseph S. Myers wrote:
> On Mon, 6 Jan 2014, Michael Kerrisk wrote:
> 
>> First of all, would this revised test for the feature_test_macros(7)
>> page be correct:
>>
>>        _BSD_SOURCE
>>               Defining this macro with any value causes  header  files
>>               to expose BSD-derived definitions.  In glibc versions up
>>               to and including 2.18, defining this macro  also  causes
>>               BSD definitions to be preferred in some situations where
>>               standards conflict, unless one or more of  _SVID_SOURCE,
>>               _POSIX_SOURCE,      _POSIX_C_SOURCE,      _XOPEN_SOURCE,
>>               _XOPEN_SOURCE_EXTENDED, or _GNU_SOURCE  is  defined,  in
>>               which  case BSD definitions are disfavored.  Since glibc
>>               2.19, _BSD_SOURCE no longer causes BSD definitions to be
>>               preferred in case of conflicts.
> 
> Yes.
> 
>> What do (admittedly, strange) programs that employ the BSD getpgrp()
>> and setpgrp() interfaces now do? It looks as though they'll require
>> source code changes to call a different API, right?
> 
> Yes.  Programs using BSD getpgrp, setpgrp, setjmp and sigpause will need 
> source code changes.

So, do the following new man-page texts look okay to you?

setpgrp()/getpgrp() [in setpgid(3)]

       The BSD-specific setpgrp() call, which takes arguments pid  and
       pgid, is is a wrapper function that calls

           setpgid(pid, pgid)

       Since  glibc  2.19,  the  BSD-specific setpgrp() function is no
       longer exposed by <unistd.h>; calls should be replaced with the
       setpgid() call shown above.

       The BSD-specific getpgrp() call, which takes a single pid argu-
       ment, is a wrapper function that calls

           getpgid(pid)

       Since glibc 2.19, the BSD-specific  getpgrp()  function  is  no
       longer  exposed  by  <unistd.h>;  calls should be replaced with
       calls to the POSIX.1 getpgrp() which takes no arguments (if the
       intent  is  to obtain the caller's PGID), or with the getpgid()
       call shown above.

sigpause(3):

       Glibc uses the BSD version  if  the  _BSD_SOURCE  feature  test
       macro  is  defined  and none of _POSIX_SOURCE, _POSIX_C_SOURCE,
       _XOPEN_SOURCE, _GNU_SOURCE, or _SVID_SOURCE is defined.  Other-
       wise,  the  System V version is used (and _XOPEN_SOURCE must be
       defined to obtain the declaration).  Since glibc 2.19, only the
       System  V  version  is exposed by <signal.h>; applications that
       formerly used the BSD sigpause() should be amended to use  sig-
       suspend(2).

setjmp(3):
       POSIX  does  not  specify whether setjmp() will save the signal
       mask (to be later restored during longjmp(3)).  In System V  it
       will  not.   In 4.3BSD it will, and there is a function _setjmp
       that will not.  On  Linux  with  glibc  versions  before  2.19,
       setjmp()  follows the System V behavior by default, but the BSD
       behavior is provided if the _BSD_SOURCE feature test  macro  is
       defined    and    none   of   _POSIX_SOURCE,   _POSIX_C_SOURCE,
       _XOPEN_SOURCE,    _XOPEN_SOURCE_EXTENDED,    _GNU_SOURCE,    or
       _SVID_SOURCE  is defined.  Since glibc 2.19, <setjmp.h> exposes
       only the System V version of setjmp().  Programs that need  the
       BSD  semantics  should  replace calls to setjmp() with calls to
       sigsetjmp() with a nonzero savesigs argument.

Thanks,

Michael

-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/