On 29-Oct-12, at 5:10 PM, Paul Eggert wrote:
On 10/29/2012 12:39 PM, John David Anglin wrote:
hppa1.1-hp-hpux10.20 has DCE threads which predates the posix
thread interface.
On such a platform, coreutils should run single-threaded
and shouldn't be mucking with DCE threads at all.
Can you investigate what went wrong? Perhaps you can
send us the output of 'configure', and what's in config.h,
and the output of 'diff -u lib/pthread.in.h lib/pthread.h'.
The attached change results in a successful build of sort. It's not
optimal but I tried to avoid messing with the automake files. I also
note that sort only seems to support pthreads.
Sorry, it appears the configure output, etc, only went to Paul.
Dave
--
John David Anglin dave.ang...@bell.net
--- pthread.in.h.orig Tue Oct 23 10:19:34 2012
+++ pthread.in.h Sat Nov 3 18:37:43 2012
@@ -26,7 +26,9 @@
/* The include_next requires a split double-inclusion guard. */
#if @HAVE_PTHREAD_H@
-# @INCLUDE_NEXT@ @NEXT_PTHREAD_H@
+# ifdef USE_POSIX_THREADS
+# @INCLUDE_NEXT@ @NEXT_PTHREAD_H@
+# endif
#endif
#ifndef _@GUARD_PREFIX@_PTHREAD_H_
@@ -38,7 +40,7 @@
#include <sys/types.h>
#include <time.h>
-#if ! @HAVE_PTHREAD_T@
+#if ! @HAVE_PTHREAD_T@ || ! defined(USE_POSIX_THREADS)
# if !GNULIB_defined_pthread_types
typedef int pthread_t;
typedef int pthread_attr_t;
@@ -97,7 +99,7 @@
#define PTHREAD_SCOPE_PROCESS 1
#endif
-#if ! @HAVE_PTHREAD_T@
+#if ! @HAVE_PTHREAD_T@ || ! defined(USE_POSIX_THREADS)
# if !GNULIB_defined_pthread_functions
@@ -226,7 +228,7 @@
#endif
-#if ! @HAVE_PTHREAD_SPINLOCK_T@
+#if ! @HAVE_PTHREAD_SPINLOCK_T@ || ! defined(USE_POSIX_THREADS)
# if !GNULIB_defined_pthread_spinlock_t