Update.cvs/libc-ud-970530

1997-05-31 02:33  Ulrich Drepper  <drepper@cygnus.com>

	* io/ftwtest-sh: More tests.

	* misc/tsearch.c: Rewrite tdestroy_recursive.

	* libio/libio.h: Define bits for libg++-2.8.

1997-05-30 22:21  Thorsten Kukuk  <kukuk@vt.uni-paderborn.de>

	* nis/nss_nis/nis-hosts.c: Same changes as in
	nss/nss_files/files-hosts.c (Always use inet_pton).
	* nis/nss_nisplus/nisplus-hosts.c: Likewise.

1997-05-30 09:50  Richard Henderson  <rth@tamu.edu>

	* manual/maint.texi: ECOFF hasn't been tested in ages and I don't
	plan on doing so ever again.  Don't say alpha-linuxecoff is known
	to work.

	* elf/dl-lookup.c: Include <alloca.h>.
	* misc/tsearch.c: Include <string.h>.
	* posix/execle.c: Include <alloca.h>.
	* posix/execlp.c: Likewise.
	* stdio-common/printf_fphex.c: Include <string.h>.
	* sunrpc/xdr.c: Include <string.h>.
	* sysdeps/generic/memccpy.c: Include only <string.h>.

	* sunrpc/clnt_udp.c (clntudp_call): Make fromlen a size_t to fix
	parameters to network functions.
	* sunrpc/pmap_rmt.c (clnt_broadcast): Likewise.
	* sunrpc/svc_tcp.c (svctcp_create): Likewise with `len'.
	(rendezvous_request): Likewise.
	* sunrpc/svc_udp.c (svcudp_bufcreate): Likewise.
	(svcudp_recv): Similar, but go through a local variable.  The old code
	used &xprt->xp_addrlen.  Someone should determine if the type of the
	structure member should be changed instead.
	* sunrpc/xdr_rec.c (xdrrec_create): Expunge a ptr->int cast warning
	and transform a nonsense for-loop to a closed form calculation.
	(fill_input_buf): Another ptr->int warning.

	* sysdeps/alpha/Makefile: Temporarily turn on -mieee globally.  This
	will last until I figure out how to build a parallel libm_ieee.

	* sysdeps/alpha/fpu/fclrexcpt.c,
	* sysdeps/alpha/fpu/fegetenv.c,
	* sysdeps/alpha/fpu/fegetround.c,
	* sysdeps/alpha/fpu/feholdexcpt.c,
	* sysdeps/alpha/fpu/fenvbits.h,
	* sysdeps/alpha/fpu/fesetenv.c,
	* sysdeps/alpha/fpu/fesetround.c,
	* sysdeps/alpha/fpu/feupdateenv.c,
	* sysdeps/alpha/fpu/fgetexcptflg.c,
	* sysdeps/alpha/fpu/fraiseexcpt.c,
	* sysdeps/alpha/fpu/fsetexcptflg.c,
	* sysdeps/alpha/fpu/ftestexcept.c: New files.  There is, btw, a small
	kernel patch that must be applied as of 2.0.31-pre1 and 2.1.41 for
	this to work properly.

	* sysdeps/alpha/fpu/fpu_control.h: Update copyright.  Change default
	FP mode to round to nearest, no exceptions as for Intel.
	* sysdeps/unix/sysv/linux/alpha/setfpucw.c: Add copyright.
	(rdfpcr): Use excb rather than trapb.  Be more efficient about
	getting at the fpcr.
	(wrfpcr): Likewise.
	(__setfpucw): Reformat.

	* sysdeps/ieee754/huge_val.h: Change GCC's HUGE_VAL definition to use
	a single DI-mode integer instead of bytes.  This produces better code
	on Alpha.  Similar for HUGE_VALF.
	* sysdeps/ieee754/nan.h: Similar for NAN.  Add NANF and NANL.

	* sysdeps/libm-ieee754/s_log2.c: Missing */ and weak_alias.

	* sysdeps/unix/alpha/sysdep.S: Tiny cleanups.
	* sysdeps/unix/sysv/linux/alpha/syscall.S: Likewise.
	* sysdeps/unix/sysv/linux/alpha/brk.S: Don't use jmp macro to make
	sure the stack frame is released the very instruction before the
	real jump out.
	* sysdeps/unix/sysv/linux/alpha/ieee_get_fp_control.S: Likewise.
	* sysdeps/unix/sysv/linux/alpha/ieee_set_fp_control.S: Likewise.

	* sysdeps/unix/sysv/linux/alpha/gnu/types.h: Undef some bits we may
	have gotten from <linux/posix_types.h>.

	* sysdeps/unix/sysv/linux/alpha/ioperm.c: Add definition for Miata.

1997-05-30 13:05  Ulrich Drepper  <drepper@cygnus.com>

	* misc/tst-tsearch.c (mangle_tree): Update casts to prevent
	warnings.

1997-05-29 14:02  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/unix/sysv/linux/sys/ttydefaults.h (CEOL, CSTATUS): Use
	correct value for _POSIX_VDISABLE.

1997-05-29 13:59  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* Makefile (manual/dir-add.texi): Force execution.

1997-05-29 13:57  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysexits.h: New file.

1997-05-29 13:55  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* stdlib/strfmon.c: Don't run past EOS after `='.

1997-02-27 10:34  Richard Henderson  <rth@tamu.edu>

1 files changed, 107 insertions, 0 deletions

diff --git a/sysdeps/alpha/fpu/fenvbits.h b/sysdeps/alpha/fpu/fenvbits.h
new file mode 100644
index 0000000000..02414e4c30
--- /dev/null
+++ b/sysdeps/alpha/fpu/fenvbits.h
@@ -0,0 +1,107 @@
+/* Copyright (C) 1997 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Richard Henderson <rth@tamu.edu>, 1997
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Library General Public License as
+   published by the Free Software Foundation; either version 2 of the
+   License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Library General Public License for more details.
+
+   You should have received a copy of the GNU Library General Public
+   License along with the GNU C Library; see the file COPYING.LIB.  If not,
+   write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+   Boston, MA 02111-1307, USA.  */
+
+/* This file should never be included directly.  */
+
+#ifndef _FENVBITS_H
+#define _FENVBITS_H	1
+
+/* Define the bits representing the exception.
+
+   Note that these are the bit positions as defined by the OSF/1
+   ieee_{get,set}_control_word interface and not by the hardware fpcr.
+
+   See the Alpha Architecture Handbook section 4.7.7.3 for details,
+   but in summary, trap shadows mean the hardware register can acquire
+   extra exception bits so for proper IEEE support the tracking has to
+   be done in software -- in this case with kernel support.
+
+   As to why the system call interface isn't in the same format as
+   the hardware register, only those crazy folks at DEC can tell you.  */
+
+enum
+  {
+    FE_INEXACT =	1UL << 21,
+#define FE_INEXACT	FE_INEXACT
+
+    FE_UNDERFLOW =	1UL << 20,
+#define FE_UNDERFLOW	FE_UNDERFLOW
+
+    FE_OVERFLOW =	1UL << 19,
+#define FE_OVERFLOW	FE_OVERFLOW
+
+    FE_DIVBYZERO =	1UL << 18,
+#define FE_DIVBYZERO	FE_DIVBYZERO
+
+    FE_INVALID =	1UL << 17,
+#define FE_INVALID	FE_INVALID
+    
+    FE_ALL_EXCEPT =
+	(FE_INEXACT | FE_DIVBYZERO | FE_UNDERFLOW | FE_OVERFLOW | FE_INVALID)
+#define FE_ALL_EXCEPT	FE_ALL_EXCEPT 
+  };
+
+
+/* Alpha chips support all four defined rouding modes. 
+
+   Note that code must be compiled to use dynamic rounding (/d) instructions
+   to see these changes.  For gcc this is -mfp-rounding-mode=d; for DEC cc
+   this is -fprm d.  The default for both is static rounding to nearest. 
+
+   These are shifted down 58 bits from the hardware fpcr because the 
+   functions are declared to take integers.  */
+
+enum
+  {
+    FE_TOWARDSZERO =	0,
+#define FE_TOWARDSZERO	FE_TOWARDSZERO
+
+    FE_DOWNWARD = 	1,
+#define FE_DOWNWARD	FE_DOWNWARD
+
+    FE_TONEAREST =	2,
+#define FE_TONEAREST	FE_TONEAREST
+
+    FE_UPWARD =		3,
+#define FE_UPWARD	FE_UPWARD
+  };
+
+
+/* Type representing exception flags.  */
+typedef unsigned long fexcept_t;
+
+/* Type representing floating-point environment.  */
+typedef unsigned long fenv_t;
+
+/* If the default argument is used we use this value.  Note that due to
+   architecture-specified page mappings, no user-space pointer will ever
+   have its two high bits set.  Co-opt one.  */
+#define FE_DFL_ENV	((fenv_t *) 0x8800000000000000UL)
+
+#ifdef __USE_GNU
+/* Floating-point environment where none of the exceptions are masked.  */
+# define FE_NOMASK_ENV	((fenv_t *) 0x880000000000003eUL)
+#endif
+
+/* The system calls to talk to the kernel's FP code.  */
+extern unsigned long __ieee_get_fp_control(void);
+extern void __ieee_set_fp_control(unsigned long);
+
+
+#endif /* fenvbits.h */