glibc/string/strxfrm.c
Ulrich Drepper 5a97622d5e update from main archive 970221
1997-02-22 00:17  Ulrich Drepper  <drepper@cygnus.com>

	* catgets/gencat.c: Change to use argp.
	* db/makedb: Likewise.
	* locale/programs/localedef.c: Likewise.

	* locale/programs/locale.c: Little adjustment for better usage of
	argp.

1997-02-20 20:07  Greg McGary  <gkm@eng.ascend.com>

	* Makeconfig: Add rules for libc with bounded pointers.
	* Makerules: Likewise.
	* config.make.in: Likewise.
	* configure.in: Likewise.

1997-02-21 10:41  Miles Bader  <miles@gnu.ai.mit.edu>

	* argp.h (OPTION_NO_USAGE): New macro.
	* argp-help.c (usage_long_opt, usage_argful_short_opt,
	add_argless_short_opt): Implement OPTION_NO_USAGE.

1997-02-20 16:41  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* malloc/obstack.h: Fix typo.

1997-02-20 15:56  Miles Bader  <miles@gnu.ai.mit.edu>

	* argp-fmtstream.c (__argp_fmtstream_update): Account for case
	where NEXTLINE points one past the end of the active buffer.

	* argp-help.c <stddef.h>: New include.
	(__argp_failure): Only exit if STATE says it's ok.
	(print_header, hol_entry_help): Use UPARAMS fields rather than
	constants.
	(_help): Call fill_in_uparams if necessary.
	(struct hol_help_state): New type.
	(struct pentry_state): Add hhstate field.  Remove prev_entry &
	sep_groups fields.
	(hol_entry_help): Add HHSTATE parameter.  Remove prev_entry &
	sep_groups parameters.
	Suppress duplicate arguments if requested, and note the fact.
	(print_header, comma): Use PEST->hhstate fields.
	(hol_help): Add HHSTATE variable & pass to hol_entry_help.
	Remove LAST_ENTRY & SEP_GROUPS variables.
	If any suplicate arguments were suppressed, print explanatory note.
	(filter_doc): Replace PEST parameter with STATE.
	(struct uparams): New type.
	(uparams): New variable.
	(struct uparam_name): New type.
	(uparam_names): New variable.
	(fill_in_uparams): New function.
	(__argp_failure, __argp_error, __argp_state_help): Make STATE
	parameter const.
	* argp.h (argp_state_help, __argp_state_help, argp_usage,
	__argp_usage, argp_error, __argp_error, argp_failure,
	__argp_failure): Make STATE parameter const.
	(ARGP_KEY_HELP_DUP_ARGS_NOTE): New macro.

	* argp.h (argp_program_bug_address): Make const.

1997-02-20 19:20  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/unix/mman/syscalls.list: Explain msync interface.

1997-02-19 01:37  Erik Troan  <ewt@redhat.com>

	* shadow/sgetspent_r.c: Accept empty third, fourth and fifth fields.

1997-02-20 14:44  Andreas Jaeger  <aj@arthur.pfalz.de>

	* stdio-common/test-fseek.c: Remove temporary file, add
	copyright.

1997-02-20 17:51  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/generic/netinet/in.h: Protect contents using
	__BEGIN/END_DECLS.  Reported by a sun <asun@zoology.washington.edu>.

	* inet/net/ethernet.h: Move to sysdeps/unix/sysv/linux/net.
	* inet/Makefile (headers): Remove net/ethernet.h.
	* sysdeps/unix/sysv/linux/Makefile: Install net/ethernet.h.
	* sysdeps/unix/sysv/linux/Dist: Distribute net/ethernet.h.

1997-02-20 15:23  Thorsten Kukuk  <kukuk@weber.uni-paderborn.de>

	* nss/nsswitch.c (__nss_configure_lookup): Use correct test when
	searching in sorted array.

1997-02-20 01:24  Philip Blundell  <pjb27@cam.ac.uk>

	* inet/getnameinfo.c: Change to use reentrant getXXbyYY functions
	and protect modification of global data.

1997-02-19 18:48  Miles Bader  <miles@gnu.ai.mit.edu>

	* argp-parse.c (argp_default_parser): Set STATE->name for OPT_PROGNAME.
	(parser_init): Use the basename for PARSER->state.name.
	* argp-help.c (__argp_error, __argp_failure, __argp_state_help):
	Use PROGRAM_INVOCATION_SHORT_NAME instead of PROGRAM_INVOCATION_NAME.

	* argp-parse.c (parser_init): Set PARSER->state.flags.
	Make check whether PARSER has the prog name in argv[0] at the
	proper place.

1997-02-19 23:34  Ulrich Drepper  <drepper@cygnus.com>

	* locale/programs/ld-time.c (time_finish): t_fmt_ampm is optional.
	Use default value instead of printing a warning.

	* nss/XXX-lookup.c: Add misssing explanation.

1997-02-19 19:14  Andreas Jaeger  <aj@arthur.pfalz.de>

	* inet/in6_addr.c: Add missing braces.

	* inet/getnameinfo.c: Include <arpa/inet.h>.

	* sysdeps/posix/getaddrinfo.c: Include <arpa/inet.h>.

1997-02-19 11:46  Ulrich Drepper  <drepper@cygnus.com>

	* string/strxfrm.c (STRCOLL): Correct handling of `position'
	levels with no non-IGNOREd element and handling of NUL byte.
	* string/strcoll.c (STRXFRM): Likewise.
	* locale/weight.h: Likewise.

	* shadow/sgetspent_r.c (LINE_PARSER): Add missing ')'.
1997-02-22 01:20:46 +00:00

270 lines
7.0 KiB
C

/* Copyright (C) 1995, 1996, 1997 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Written by Ulrich Drepper <drepper@gnu.ai.mit.edu>, 1995.
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. */
#include <stddef.h>
#include <stdlib.h>
#include <string.h>
#ifndef WIDE_VERSION
# define STRING_TYPE char
# define USTRING_TYPE unsigned char
# define L_(Ch) Ch
# ifdef USE_IN_EXTENDED_LOCALE_MODEL
# define STRXFRM __strxfrm_l
# else
# define STRXFRM strxfrm
# endif
# define STRLEN strlen
# define STPNCPY __stpncpy
#endif
/* Include the shared helper functions. `strxfrm'/`wcsxfrm' also use
these functions. */
#include "../locale/weight.h"
#ifndef WIDE_VERSION
/* Write 32 bit value UTF-8 encoded but only if enough space is left. */
static __inline size_t
print_val (u_int32_t value, char *dest, size_t max, size_t act)
{
char tmp[6];
int idx = 0;
if (value < 0x80)
tmp[idx++] = (char) value;
else
{
tmp[idx++] = '\x80' + (char) (value & 0x3f);
value >>= 6;
if (value < 0x20)
tmp[idx++] = '\xc0' + (char) value;
else
{
tmp[idx++] = '\x80' + (char) (value & 0x3f);
value >>= 6;
if (value < 0x10)
tmp[idx++] = '\xe0' + (char) value;
else
{
tmp[idx++] = '\x80' + (char) (value & 0x3f);
value >>= 6;
if (value < 0x08)
tmp[idx++] = '\xf0' + (char) value;
else
{
tmp[idx++] = '\x80' + (char) (value & 0x3f);
value >>= 6;
if (value < 0x04)
tmp[idx++] = '\xf8' + (char) value;
else
{
tmp[idx++] = '\x80' + (char) (value & 0x3f);
tmp[idx++] = '\xfc' + (char) (value >> 6);
}
}
}
}
}
while (idx-- > 0)
{
if (act < max)
dest[act] = tmp[idx];
++act;
}
return act;
}
#else
static __inline size_t
print_val (u_int32_t value, wchar_t *dest, size_t max, size_t act)
{
/* We cannot really assume wchar_t is 32 bits wide. But it is for
GCC and so we don't do much optimization for the other case. */
if (sizeof (wchar_t) == 4)
{
if (act < max)
dest[act] = (wchar_t) value;
++act;
}
else
{
wchar_t tmp[3];
size_t idx = 0;
if (value < 0x8000)
tmp[idx++] = (wchar_t) act;
else
{
tmp[idx++] = (wchar_t) (0x8000 + (value & 0x3fff));
value >>= 14;
if (value < 0x2000)
tmp[idx++] = (wchar_t) (0xc000 + value);
else
{
tmp[idx++] = (wchar_t) (0x8000 + (value & 0x3fff));
value >>= 14;
tmp[idx++] = (wchar_t) (0xe000 + value);
}
}
while (idx-- > 0)
{
if (act < max)
dest[act] = tmp[idx];
++act;
}
}
return act;
}
#endif
/* Transform SRC into a form such that the result of strcmp
on two strings that have been transformed by strxfrm is
the same as the result of strcoll on the two strings before
their transformation. The transformed string is put in at
most N characters of DEST and its length is returned. */
#ifndef USE_IN_EXTENDED_LOCALE_MODEL
size_t
STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n)
#else
size_t
STRXFRM (STRING_TYPE *dest, const STRING_TYPE *src, size_t n, __locale_t l)
#endif
{
#ifdef USE_IN_EXTENDED_LOCALE_MODEL
struct locale_data *current = l->__locales[LC_COLLATE];
# if BYTE_ORDER == BIG_ENDIAN
const u_int32_t *collate_table = (const u_int32_t *)
current->values[_NL_ITEM_INDEX (_NL_COLLATE_TABLE_EB)].string;
const u_int32_t *collate_extra = (const u_int32_t *)
current->values[_NL_ITEM_INDEX (_NL_COLLATE_EXTRA_EB)].string;
# elif BYTE_ORDER == LITTLE_ENDIAN
const u_int32_t *collate_table = (const u_int32_t *)
current->values[_NL_ITEM_INDEX (_NL_COLLATE_TABLE_EL)].string;
const u_int32_t *collate_extra = (const u_int32_t *)
current->values[_NL_ITEM_INDEX (_NL_COLLATE_EXTRA_EL)].string;
# else
# error bizarre byte order
# endif
#endif
weight_t *forw = NULL;
weight_t *backw = NULL;
size_t pass;
size_t written;
/* If the current locale does not specify locale data we use normal
8-bit string comparison. */
if (collate_nrules == 0)
{
if (n != 0)
STPNCPY (dest, src, n);
return STRLEN (src);
}
/* Get full information about the string. This means we get
information for all passes in a special data structure. */
get_string (src, forw, backw);
/* Now we have all the information. In at most the given number of
passes we can finally decide about the order. */
written = 0;
for (pass = 0; pass < collate_nrules; ++pass)
{
int forward = (collate_rules[pass] & sort_forward) != 0;
const weight_t *run = forward ? forw : backw;
int idx = forward ? 0 : run->data[pass].number - 1;
while (1)
{
int ignore = 0;
u_int32_t w = 0;
/* Here we have to check for IGNORE entries. If these are
found we count them and go on with he next value. */
while (run != NULL
&& ((w = run->data[pass].value[idx])
== (u_int32_t) IGNORE_CHAR))
{
++ignore;
if ((forward && ++idx >= run->data[pass].number)
|| (!forward && --idx < 0))
{
weight_t *nextp = forward ? run->next : run->prev;
if (nextp == NULL)
{
w = 0;
/* No more non-INGOREd elements means lowest
possible value. */
ignore = -1;
}
else
idx = forward ? 0 : nextp->data[pass].number - 1;
run = nextp;
}
}
/* Stop if all characters are processed. */
if (run == NULL)
break;
/* Now we have information of the number of ignored weights
and the value of the next weight. We have to add 2
because 0 means EOS and 1 is the intermediate string end. */
if ((collate_rules[pass] & sort_position) != 0)
written = print_val (ignore + 2, dest, n, written);
if (w != 0)
written = print_val (w, dest, n, written);
/* We have to increment the index counters. */
if ((forward && ++idx >= run->data[pass].number)
|| (!forward && --idx < 0))
if (forward)
{
run = run->next;
idx = 0;
}
else
{
run = run->prev;
if (run != NULL)
idx = run->data[pass].number - 1;
}
}
/* Write marker for end of word. */
if (pass + 1 < collate_nrules)
written = print_val (1, dest, n, written);
}
/* Terminate string. */
if (written < n)
dest[written] = L_('\0');
/* Return length without counting the terminating '\0'. */
return written;
}