Update. cvs/libc-ud-970326 cvs/libc-970329 cvs/libc-970328 cvs/libc-970327

1997-03-27 02:28  Ulrich Drepper  <drepper@cygnus.com>

	* gmon/gmon.c (monstartup): Mark all messages.
	(write_call_graph): Rewrite to use larger I/O vector for writev
	call to reduce syscall overhead.
	(write_bb_counts): Simplify writev handling.

	* inet/rexec.c: Make string parameters `const'.
	* resolv/netdb.h: Add prototypes for rcmd, rexec, ruserok, and
	rresvport.

	* math/Makefile: Don't define CFLAGS-* macros to prevent inlining
	in libm-test.
	* math/libm-test.c (this_does_nothing): Remove functions.  It's
	notuseful on any platform but ix86.
	(inverse_func_pair_test): Don't use this_does_nothing.  Use
	memory reference.
	(identities1_test): Likewise.
	(identities2_test): Likewise.
	(identities3_test): Likewise.
	(basic_test): Likewise.
	Patch by Andreas Schwab.
	(BUILD_COMPLEX): New macro.  Create complex number from real and
	imaginary parts.  This works around bugs/inefficiencies in current
	gcc.
	(cexp_test): Use BUILD_COMPLEX.  Add more tests.

	* nss/nsswitch.c: Fix typo.

	* posix/glob.h: Add declaration for glob_pattern_p.
	* posix/glob.c: Rename glob_pattern_p to __glob_pattern_p and
	make glob_pattern_p a weak alias.  This function is used in other
	packages (e.g. bash).

	* signal/Makefile (routines): Add sigisempty, sigandset, and
	sigorset.
	* signal/signal.h: Add prototypes for sigisempty, sigandset, and
	sigorset.
	* signal/sigisempty.c: New file.
	* signal/sigandset.c: New file.
	* signal/sigorset.c: New file.
	* sysdeps/generic/sigset.h: Define __sigisemptyset, __sigandset,
	and __sigorset.
	* sysdeps/unix/sysv/linux/sigset.h: Likewise.

	* stdlib/strtod.c: Handle `n-char-sequence' in NaN parsing.  It
	determines the bits in the mantissa part of the NaN.
	* stdlib/strtof.c: Define SET_MANTISSA for float type.
	* wcsmbs/wcstof.c: Define SET_MANTISSA for float type.
	* stdlib/strtold.c: Define SET_MANTISSA for long double type.
	* wcsmbs/wcstold.c: Define SET_MANTISSA for long double type.

	* sysdeps/libm-ieee754/s_cexp.c: Use explicit assignment to
	complex number components.  Some more corrects for special cases.
	* sysdeps/libm-ieee754/s_cexpf.c: Likewise.
	* sysdeps/libm-ieee754/s_cexpl.c: Likewise.

	* sysdeps/sparc/elf/start.S: Remove as per request of Miguel de Icaza.

	* sysdeps/unix/sysv/linux/netinet/icmp.h: Remove since we have
	ip_icmp.h.  Reported by HJ Lu.

1997-03-25 03:50  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/m68k/Makefile (CFLAGS-rtld.c): Add -Wno-unused.

	* sysdeps/m68k/dl-machine.h (elf_machine_rela): Rewritten as for
	i386.
	(elf_machine_lookup_noexec_p, elf_machine_lookup_noplt_p,
	ELF_MACHINE_RELOC_NOPLT): Define.

1997-03-25 03:48  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* grp/grp.h: Include <stddef.h> only once.

1997-03-25 09:38  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/unix/sysv/linux/termbits.h (OXTABS): Don't define.
	* sysdeps/unix/sysv/linux/alpha/termbits.h (OXTABS): Likewise.

	* termios/sys/ttydefaults.h (TTYDEF_OFLAG): Use either OXTABS or
	TAB3, if one of them is defined.

1997-03-26 04:53  Ulrich Drepper  <drepper@cygnus.com>

	* posix/glob.c (next_brace_sub): Decrement depth counter when '}'
	is found.
	Patch by Dennis Henriksen <opus@flamingo.osrl.dk>.

1997-03-25  16:25  Thorsten Kukuk  <kukuk@vt.uni-paderborn.de>

	* sunrpc/Makefile: Comment gccwarn out.
	* sunrpc/auth_none.c: Fix prototypes and parameters for compiling
	with enabled warnings.
	* sunrpc/auth_unix.c: Likewise.
	* sunrpc/authuxprot.c: Likewise.
	* sunrpc/bindrsvprt.c: Likewise.
	* sunrpc/clnt_gen.c: Likewise.
	* sunrpc/clnt_perr.c: Likewise.
	* sunrpc/clnt_raw.c: Likewise.
	* sunrpc/clnt_simp.c: Likewise.
	* sunrpc/clnt_tcp.c: Likewise.
	* sunrpc/clnt_udp.c: Likewise.
	* sunrpc/get_myaddr.c: Likewise.
	* sunrpc/getrpcport.c: Likewise.
	* sunrpc/pm_getmaps.c: Likewise.
	* sunrpc/pm_getport.c: Likewise.
	* sunrpc/pmap_clnt.c: Likewise.
	* sunrpc/pmap_prot.c: Likewise.
	* sunrpc/pmap_prot2.c: Likewise.
	* sunrpc/pmap_rmt.c: Likewise.
	* sunrpc/rpc/auth.h: Likewise.
	* sunrpc/rpc/clnt.h: Likewise.
	* sunrpc/rpc/pmap_clnt.h: Likewise.
	* sunrpc/rpc/svc.h: Likewise.
	* sunrpc/rpc/svc_auth.h: Likewise.
	* sunrpc/rpc/types.h: Likewise.
	* sunrpc/rpc/xdr.h: Likewise.
	* sunrpc/rpc_clntout.c: Likewise.
	* sunrpc/rpc_cmsg.c: Likewise.
	* sunrpc/rpc_dtable.c: Likewise.
	* sunrpc/rpc_prot.c: Likewise.
	* sunrpc/svc.c: Likewise.
	* sunrpc/svc_auth.c: Likewise.
	* sunrpc/svc_authux.c: Likewise.
	* sunrpc/svc_raw.c: Likewise.
	* sunrpc/svc_run.c: Likewise.
	* sunrpc/svc_simple.c: Likewise.
	* sunrpc/svc_tcp.c: Likewise.
	* sunrpc/svc_udp.c: Likewise.
	* sunrpc/xdr.c: Likewise.
	* sunrpc/xdr_array.c: Likewise.
	* sunrpc/xdr_mem.c: Likewise.
	* sunrpc/xdr_rec.c: Likewise.
	* sunrpc/xdr_ref.c: Likewise.
	* sunrpc/xdr_stdio.c: Likewise.

1997-03-25 13:39  Ulrich Drepper  <drepper@cygnus.com>

	* math/libm-test.c (log2_test): Compile this function and call it.
	(exp2_test): Likewise, but check whether function really exists
	before testing.

	* math/Makefile (libm-calls): Add s_log2 and s_exp2.

1997-03-25 04:50  Ulrich Drepper  <drepper@cygnus.com>

	Implement exp2 function.
	* sysdeps/libm-i387/s_exp2.S: New file.
	* sysdeps/libm-i387/s_exp2f.S: New file.
	* sysdeps/libm-i387/s_exp2l.S: New file.

	Implement log2 function.
	* sysdeps/libm-i387/s_log2.S: New file.
	* sysdeps/libm-i387/s_log2f.S: New file.
	* sysdeps/libm-i387/s_log2l.S: New file.
	* sysdeps/libm-ieee754/s_log2.c: New file.
	* sysdeps/libm-ieee754/s_log2f.c: New file.
	* sysdeps/stub/s_log2.c: New file.  Stub version.

2 files changed, 100 insertions, 52 deletions

diff --git a/math/Makefile b/math/Makefile
index d0c05835f5..2578dbf839 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -48,7 +48,7 @@ libm-calls = e_acos e_acosh e_asin e_atan2 e_atanh e_cosh e_exp e_fmod	\
 	     w_gamma_r w_hypot w_j0 w_j1 w_jn w_lgamma w_lgamma_r	\
 	     w_log w_log10 w_pow w_remainder w_scalb w_sinh w_sqrt	\
 	     s_signbit s_fpclassify s_fmax s_fmin s_fdim s_nan s_trunc	\
-	     s_remquo							\
+	     s_remquo s_log2 s_exp2					\
 	     conj cimag creal cabs s_cexp
 libm-routines = $(libm-support) $(libm-calls) \
 		$(patsubst %_rf,%f_r,$(libm-calls:=f))	\
@@ -73,9 +73,6 @@ tests = test-float test-double $(test-longdouble-$(long-double-fcts))
 # XXX This test is disabled for now since the functions are too buggy.
 #test-longdouble-yes = test-longdouble
 
-CFLAGS-test-float.c = -fno-inline
-CFLAGS-test-double.c = -fno-inline
-CFLAGS-test-longdouble.c = -fno-inline
 LDLIBS-test-float = libm
 LDLIBS-test-double = libm
 LDLIBS-test-longdouble = libm
diff --git a/math/libm-test.c b/math/libm-test.c
index 05acca2c53..a575cc66a8 100644
--- a/math/libm-test.c
+++ b/math/libm-test.c
@@ -63,10 +63,6 @@
    -s,--silent outputs only the error count (equals --verbose=0)
  */
 
-/* Define if the following ISO C 9X functions are implemented: exp2,
-   log2.  */
-#undef ISO_9X_IMPLEMENTED
-
 #ifndef _GNU_SOURCE
 # define _GNU_SOURCE
 #endif
@@ -103,6 +99,12 @@ static MATHTYPE plus_infty, minus_infty, nan_value;
 
 typedef MATHTYPE (*mathfunc) (MATHTYPE);
 
+#define BUILD_COMPLEX(real, imag) \
+  ({ __complex__ MATHTYPE __retval;					      \
+     __real__ __retval = (real);					      \
+     __imag__ __retval = (imag);					      \
+     __retval; })
+
 
 #define ISINF(x) \
 (sizeof (x) == sizeof (float) ?						      \
@@ -136,18 +138,6 @@ asm ("fnstsw":"=a" (sw));
 
 
 /*
-   Call to an external function so that floating point registers
-   get moved to memory
- */
-static void
-this_does_nothing (void)
-{
-  clock_t dummy;
-
-  dummy = clock ();
-}
-
-/*
    Get a random value x with min_value < x < max_value
    and min_value, max_value finite,
    max_value and min_value shouldn't be too close together
@@ -644,10 +634,15 @@ exp_test (void)
 }
 
 
-#ifdef ISO_9X_IMPLEMENTED
 static void
 exp2_test (void)
 {
+  errno = 0;
+  exp2(0);
+  if (errno == ENOSYS)
+    /* Function not implemented.  */
+    return;
+
   check ("exp2 (+0) == 1", FUNC(exp2) (0), 1);
   check ("exp2 (-0) == 1", FUNC(exp2) (minus_zero), 1);
 
@@ -655,7 +650,6 @@ exp2_test (void)
   check ("exp2 (-inf) == 0", FUNC(exp2) (minus_infty), 0);
   check ("exp2 (10) == 1024", FUNC(exp2) (10), 1024);
 }
-#endif
 
 
 static void
@@ -892,7 +886,6 @@ log1p_test (void)
 }
 
 
-#ifdef ISO_9X_IMPLEMENTED
 static void
 log2_test (void)
 {
@@ -914,7 +907,6 @@ log2_test (void)
   check ("log2 (256) == 8", FUNC(log2) (256.0), 8);
 
 }
-#endif
 
 
 static void
@@ -1540,32 +1532,92 @@ cexp_test (void)
 {
   __complex__ MATHTYPE result;
 
-  result = FUNC(cexp) (plus_zero + 1.0i * plus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_zero, plus_zero));
   check ("real(cexp(0 + 0i)) = 1", __real__ result, 1);
   check ("imag(cexp(0 + 0i)) = 0", __imag__ result, 0);
-  result = FUNC(cexp) (minus_zero + 1.0i * plus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_zero, plus_zero));
   check ("real(cexp(-0 + 0i)) = 1", __real__ result, 1);
   check ("imag(cexp(-0 + 0i)) = 0", __imag__ result, 0);
-  result = FUNC(cexp) (plus_zero + 1.0i * minus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_zero, minus_zero));
   check ("real(cexp(0 - 0i)) = 1", __real__ result, 1);
-  check ("imag(cexp(0 - 0i)) = 0", __imag__ result, 0);
-  result = FUNC(cexp) (minus_zero + 1.0i * minus_zero);
+  check ("imag(cexp(0 - 0i)) = -0", __imag__ result, minus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_zero, minus_zero));
   check ("real(cexp(-0 - 0i)) = 1", __real__ result, 1);
-  check ("imag(cexp(-0 - 0i)) = 0", __imag__ result, 0);
+  check ("imag(cexp(-0 - 0i)) = -0", __imag__ result, minus_zero);
 
-  result = FUNC(cexp) (plus_infty + 1.0i * plus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_infty, plus_zero));
   check_isinfp ("real(cexp(+inf + 0i)) = +inf", __real__ result);
   check ("imag(cexp(+inf + 0i)) = 0", __imag__ result, 0);
-  result = FUNC(cexp) (plus_infty + 1.0i * minus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_infty, minus_zero));
   check_isinfp ("real(cexp(+inf - 0i)) = +inf", __real__ result);
   check ("imag(cexp(+inf - 0i)) = 0", __imag__ result, 0);
 
-  result = FUNC(cexp) (minus_infty + 1.0i * plus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_infty, plus_zero));
   check ("real(cexp(-inf + 0i)) = 0", __real__ result, 0);
   check ("imag(cexp(-inf + 0i)) = 0", __imag__ result, 0);
-  result = FUNC(cexp) (minus_infty + 1.0i * minus_zero);
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_infty, minus_zero));
   check ("real(cexp(-inf - 0i)) = 0", __real__ result, 0);
-  check ("imag(cexp(-inf - 0i)) = 0", __imag__ result, 0);
+  check ("imag(cexp(-inf - 0i)) = -0", __imag__ result, minus_zero);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (100.0, plus_infty));
+  check_isnan ("real(cexp(x + i inf)) = NaN", __real__ result);
+  check_isnan ("imag(cexp(x + i inf)) = NaN", __imag__ result);
+  result = FUNC(cexp) (BUILD_COMPLEX (100.0, minus_infty));
+  check_isnan ("real(cexp(x - i inf)) = NaN", __real__ result);
+  check_isnan ("imag(cexp(x - i inf)) = NaN", __imag__ result);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_infty, 2.0));
+  check ("real(cexp(-inf + 2.0i)) = -0", __real__ result, minus_zero);
+  check ("imag(cexp(-inf + 2.0i)) = 0", __imag__ result, 0);
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_infty, 4.0));
+  check ("real(cexp(-inf + 4.0i)) = -0", __real__ result, minus_zero);
+  check ("imag(cexp(-inf + 4.0i)) = -0", __imag__ result, minus_zero);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_infty, 2.0));
+  check_isinfn ("real(cexp(+inf + 2.0i)) = -0", __real__ result);
+  check_isinfp ("imag(cexp(+inf + 2.0i)) = 0", __imag__ result);
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_infty, 4.0));
+  check_isinfn ("real(cexp(+inf + 4.0i)) = -0", __real__ result);
+  check_isinfn ("imag(cexp(+inf + 4.0i)) = -0", __imag__ result);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_infty, plus_infty));
+  check_isinfp ("real(cexp(+inf + i inf)) = +inf", __real__ result);
+  check_isnan ("imag(cexp(+inf + i inf)) = NaN", __imag__ result);
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_infty, minus_infty));
+  check_isinfp ("real(cexp(+inf - i inf)) = +inf", __real__ result);
+  check_isnan ("imag(cexp(+inf - i inf)) = NaN", __imag__ result);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_infty, plus_infty));
+  check ("real(cexp(-inf + i inf)) = 0", __real__ result, 0);
+  check ("imag(cexp(-inf + i inf)) = 0", __imag__ result, 0);
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_infty, minus_infty));
+  check ("real(cexp(-inf - i inf)) = 0", __real__ result, 0);
+  check ("imag(cexp(-inf - i inf)) = 0", __imag__ result, 0);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (minus_infty, nan_value));
+  check ("real(cexp(-inf + i NaN)) = 0", __real__ result, 0);
+  check ("imag(cexp(-inf + i NaN)) = 0", fabs (__imag__ result), 0);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (plus_infty, nan_value));
+  check_isinfp ("real(cexp(+inf + i NaN)) = +inf", __real__ result);
+  check_isnan ("imag(cexp(+inf + i NaN)) = NaN", __imag__ result);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (nan_value, 1.0));
+  check_isnan ("real(cexp(NaN + 1i)) = NaN", __real__ result);
+  check_isnan ("imag(cexp(NaN + 1i)) = NaN", __imag__ result);
+  result = FUNC(cexp) (BUILD_COMPLEX (nan_value, 1.0));
+  check_isnan ("real(cexp(NaN + i inf)) = NaN", __real__ result);
+  check_isnan ("imag(cexp(NaN + i inf)) = NaN", __imag__ result);
+  result = FUNC(cexp) (BUILD_COMPLEX (nan_value, 1.0));
+  check_isnan ("real(cexp(NaN + i NaN)) = NaN", __real__ result);
+  check_isnan ("imag(cexp(NaN + i NaN)) = NaN", __imag__ result);
+
+  result = FUNC(cexp) (BUILD_COMPLEX (0, nan_value));
+  check_isnan ("real(cexp(0 + i NaN)) = NaN", __real__ result);
+  check_isnan ("imag(cexp(0 + i NaN)) = NaN", __imag__ result);
+  result = FUNC(cexp) (BUILD_COMPLEX (1, nan_value));
+  check_isnan ("real(cexp(1 + i NaN)) = NaN", __real__ result);
+  check_isnan ("imag(cexp(1 + i NaN)) = NaN", __imag__ result);
 }
 
 
@@ -1578,9 +1630,9 @@ inverse_func_pair_test (const char *test_name,
   int result;
 
   a = f1 (x);
-  this_does_nothing ();
+  (void) &a;
   b = inverse (a);
-  this_does_nothing ();
+  (void) &b;
 
   result = check_equal (b, x, epsilon, &difference);
   output_result (test_name, result,
@@ -1630,11 +1682,11 @@ identities1_test (MATHTYPE x, MATHTYPE epsilon)
   int result;
 
   res1 = FUNC(sin) (x);
-  this_does_nothing ();
+  (void) &res1;
   res2 = FUNC(cos) (x);
-  this_does_nothing ();
+  (void) &res2;
   res3 = res1 * res1 + res2 * res2;
-  this_does_nothing ();
+  (void) &res3;
 
   result = check_equal (res3, 1.0, epsilon, &diff);
   output_result_ext ("sin^2 + cos^2 == 1", result,
@@ -1650,13 +1702,13 @@ identities2_test (MATHTYPE x, MATHTYPE epsilon)
   int result;
 
   res1 = FUNC(sin) (x);
-  this_does_nothing ();
+  (void) &res1;
   res2 = FUNC(cos) (x);
-  this_does_nothing ();
+  (void) &res2;
   res3 = FUNC(tan) (x);
-  this_does_nothing ();
+  (void) &res3;
   res4 = res1 / res2;
-  this_does_nothing ();
+  (void) &res4;
 
   result = check_equal (res4, res3, epsilon, &diff);
   output_result_ext ("sin/cos == tan", result,
@@ -1672,11 +1724,11 @@ identities3_test (MATHTYPE x, MATHTYPE epsilon)
   int result;
 
   res1 = FUNC(sinh) (x);
-  this_does_nothing ();
+  (void) &res1;
   res2 = FUNC(cosh) (x);
-  this_does_nothing ();
+  (void) &res2;
   res3 = res2 * res2 - res1 * res1;
-  this_does_nothing ();
+  (void) &res3;
 
   result = check_equal (res3, 1.0, epsilon, &diff);
   output_result_ext ("cosh^2 - sinh^2 == 1", result,
@@ -1721,7 +1773,10 @@ basic_tests (void)
   NaN_var = nan_value;
   Inf_var = one_var / zero_var;
 
-  this_does_nothing ();
+  (void) &zero_var;
+  (void) &one_var;
+  (void) &NaN_var;
+  (void) &Inf_var;
 
   check_isinfp ("isinf (inf) == +1", Inf_var);
   check_isinfn ("isinf (-inf) == -1", -Inf_var);
@@ -1854,9 +1909,7 @@ main (int argc, char *argv[])
   cos_test ();
   cosh_test ();
   exp_test ();
-#ifdef ISO_9X_IMPLEMENTED
   exp2_test ();
-#endif
   expm1_test ();
   frexp_test ();
   ilogb_test ();
@@ -1864,9 +1917,7 @@ main (int argc, char *argv[])
   log_test ();
   log10_test ();
   log1p_test ();
-#ifdef ISO_9X_IMPLEMENTED
   log2_test ();
-#endif
   logb_test ();
   modf_test ();
   scalb_test ();