[Bug c++/15415] _REENTRANT defined if string header included without -pthread

["cludwig at cdc dot informatik dot tu-darmstadt dot de" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From cludwig at cdc dot informatik dot tu-darmstadt dot de  2004-05-13 12:46 -------
Subject: Re:  _REENTRANT defined if string header included without -pthread

On Thu, May 13, 2004 at 10:53:29AM -0000, pinskia at gcc dot gnu dot org wrote:
> This is a dup of bug 11953, 

I don't know why I missed #11953. Sorry.

> this was an intended change.

Very well, but I doubt the consequence I complained about (not beeing
able to detect -pthread on x86 / Linux) was intended... Note that the
patch that introduced this change was supposed to solve a problem on
True64 Unix, http://gcc.gnu.org/ml/gcc-patches/2003-07/msg01607.html
does not talk about x86 / Linux  at all.

But more important: The solution to #11953 proposed by Benjamin Kosnik
(test for _POSIX_THREADS and check >=0) fails! If I amend my test and
compile it with gcc 3.4.0, I get the following:

  cludwig@lap200:~/C++/gcc3.4/tmp> cat REENTRANT.cc
  #if defined(INCLUDE_STRING)
  #  include <string>
  #  warning "string header included"
  #else
  #  warning "no header included"
  #endif
  
  #if defined(_REENTRANT)
  #  warning "_REENTRANT is defined"
  #else
  #  warning "_REENTRANT is not defined"
  #endif
  
  #if defined(_POSIX_THREADS) && _POSIX_THREADS >= 0
  #  warning "_POSIX_THREADS >= 0"
  #elif defined(_POSIX_THREADS)
  #  warning "_POSIX_THREADS defxined"
  #else
  #  warning "_POSIX_THREADS not defined"
  #endif
  
  #include <unistd.h>
  
  #warning "After inclusion of unistd.h:"
  
  #if defined(_REENTRANT)
  #  warning "_REENTRANT is defined"
  #else
  #  warning "_REENTRANT is not defined"
  #endif
  
  #if defined(_POSIX_THREADS) && _POSIX_THREADS >= 0
  #  warning "_POSIX_THREADS >= 0"
  #elif defined(_POSIX_THREADS)
  #  warning "_POSIX_THREADS defxined"
  #else
  #  warning "_POSIX_THREADS not defined"
  #endif
  
  cludwig@lap200:~/C++/gcc3.4/tmp> g++ -o /dev/null -c REENTRANT.cc
  REENTRANT.cc:5:4: Warnung: #warning "no header included"
  REENTRANT.cc:11:4: Warnung: #warning "_REENTRANT is not defined"
  REENTRANT.cc:19:4: Warnung: #warning "_POSIX_THREADS not defined"
  REENTRANT.cc:24:2: Warnung: #warning "After inclusion of unistd.h:"
  REENTRANT.cc:29:4: Warnung: #warning "_REENTRANT is not defined"
  REENTRANT.cc:33:4: Warnung: #warning "_POSIX_THREADS >= 0"
  
  cludwig@lap200:~/C++/gcc3.4/tmp> g++ -DINCLUDE_STRING -o /dev/null -c REENTRANT.cc
  REENTRANT.cc:3:4: Warnung: #warning "string header included"
  REENTRANT.cc:9:4: Warnung: #warning "_REENTRANT is defined"
  REENTRANT.cc:15:4: Warnung: #warning "_POSIX_THREADS >= 0"
  REENTRANT.cc:24:2: Warnung: #warning "After inclusion of unistd.h:"
  REENTRANT.cc:27:4: Warnung: #warning "_REENTRANT is defined"
  REENTRANT.cc:33:4: Warnung: #warning "_POSIX_THREADS >= 0"
  
  ludwig@lap200:~/C++/gcc3.4/tmp> g++ -pthread -o /dev/null -c REENTRANT.cc
  REENTRANT.cc:5:4: Warnung: #warning "no header included"
  REENTRANT.cc:9:4: Warnung: #warning "_REENTRANT is defined"
  REENTRANT.cc:19:4: Warnung: #warning "_POSIX_THREADS not defined"
  REENTRANT.cc:24:2: Warnung: #warning "After inclusion of unistd.h:"
  REENTRANT.cc:27:4: Warnung: #warning "_REENTRANT is defined"
  REENTRANT.cc:33:4: Warnung: #warning "_POSIX_THREADS >= 0"


_POSIX_THREADS is defined only if certain headers are included. It
does not matter at all if -pthread was on the compilers command line. 
It seems to me (without looking up any references) that _POSIX_THREADS
indicates whether the *system* offers pthread support. But it's
totally useless if you want to decide whether the current translation
unit is to be built with thread support. (After all, I don't want to
pay for any overhead in a single threaded application, even if both
the compiler and the system are capable of pthread support.)

I therefore don't consider this bug report as resolved yet. There is
no documented way to decide at compile time whether -pthread was on
the command line.

Regards

Christoph


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15415