configure and HP-UX

Michael Piotrowski mxp at dynalabs.de
Fri Oct 19 10:38:26 EDT 2001


I'm currently evaluating Python for a project, and the first thing to
do was to build Python 2.2a4 on HP-UX 11.  Judging from the archives,
this is notoriously difficult, and it *did* take a lot of
experimentation to get it right.  However, it turned out that almost
all problems are related to configure.

I'm now putting together a patch for configure.in to fix these
problems, but I have a question concerning the detection of thread
support in configure.in (so this is not really a Python question, but
since it is about *building* Python, I thought this is the best place
to ask).

configure.in uses the following approach (roughly):

if pthread_create() in -lpthread # (1)
   # Darwin-specific action
   LIBS="-lpthread $LIBS"
else
   if pthread_detach() available without a special library # (2)
      # Darwin-specific action; LIBS is *not* changed
   else
      [...]
      if __pthread_create_system() in -lpthread # (3)
         LIBS="-lpthread $LIBS"
      [...]

Now, on HP-UX, (3) would be the correct result.  However, since (2)
succeeds, it is never tested for (3).  But since in (2) -lpthreads is
not added to LIBS, -lpthread is not used, and trying to use threads
will fail.

My question is, is there a special reason for having (2)?  I can't
really see what its purpose is.

Thanks and greetings

-- 
Michael Piotrowski, M.A.                                  <mxp at dynalabs.de>



More information about the Python-list mailing list