(scm_string_copy): Use scm_c_substring to get a copy-on-write string.

[bpt/guile.git] / libguile / ChangeLog

2004-08-19  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

        New string implementation, with copy-on-write strings and
	mutation-sharing substrings, and a new internal string API.
	Symbols can now share memory with strings.
	
	* tags.h (scm_tc7_stringbuf): New tag.

	* strings.h, strings.c: (scm_i_string_chars, scm_i_string_length,
	scm_i_string_writable_chars, scm_i_string_stop_writing): New, to
	replace SCM_I_STRING_CHARS and SCM_I_STRING_LENGTH.  Updated all
	uses.
	(scm_i_make_string, scm_c_make_string): New, to replace
	scm_allocate_string.  Updated all uses.
	(SCM_STRINGP, SCM_STRING_CHARS, SCM_STRING_UCHARS,
	SCM_STRING_LENGTH): Deprecated.
	(scm_allocate_string, scm_take_str, scm_take0str, scm_mem2string,
	scm_str2string, scm_makfrom0str, scm_makfrom0str_opt):
	Discouraged.  Replaced all uses with scm_from_locale_string or
	similar, as appropriate.
	(scm_c_string_length, scm_c_string_ref, scm_c_string_set_x,
	scm_c_substring, scm_c_substring_shared, scm_c_substring_copy,
	scm_substring_shared, scm_substring_copy): New.

	* symbols.c, symbols.h (SCM_SYMBOLP, SCM_SYMBOL_FUNC,
	SCM_SET_SYMBOL_FUNC, SCM_SYMBOL_PROPS, SCM_SET_SYMBOL_PROPS,
	SCM_SYMBOL_HASH, SCM_SYMBOL_INTERNED_P, scm_mem2symbol,
	scm_str2symbol, scm_mem2uninterned_symbol): Discouraged.
	(SCM_SYMBOL_LENGTH, SCM_SYMBOL_CHARS, scm_c_symbol2str):
	Deprecated.
	(SCM_MAKE_SYMBOL_TAG, SCM_SET_SYMBOL_LENGTH, SCM_SET_SYMBOL_CHARS,
	SCM_PROP_SLOTS, SCM_SET_PROP_SLOTS): Removed.
	(scm_is_symbol, scm_from_locale_symbol, scm_from_locale_symboln):
	New, to replace scm_str2symbol and scm_mem2symbol, respectively.
	Updated all uses.
	(scm_gensym): Generate only the number suffix in the buffer, just
	string-append the prefix.

	* error.c (scm_memory_error): Do not try to throw, just abort.
	Throwing will not work anyway.

	* gh.h, gh-data.c (gh_set_substr): Made src const.

	* ports.c (scm_i_mode_bits_n): New, for counted strings.
	(scm_mode_bits): Use it.
	(scm_c_port_for_each): Blocking GC does not seem to work, allocate
	a vector normally and fill that instead of consing a list with a
	blocked GC.

	* read.c (scm_i_casei_streq): New, for counted strings.

	* threads.c (gc_section_count): Removed, thread-sleeping can not
	be nested.
	(scm_i_thread_put_to_sleep): Call scm_i_leave_guile before locking
	admin mutex so that we can be put to sleep by other threads while
	blocking on that mutex.  Lock all the heap mutex of all threads,
	including ourselves.
	(scm_i_thread_wake_up): Unlock all threads, including ourselves,
	call scm_i_enter_guile.
	(scm_thread_mark_stacks): Expect all threads to be suspended.
	
	* gc.h, gc.c (scm_i_gc_admin_mutex): New, to protect
	scm_gc_mallocated, for now.
	(scm_init_storage): Initialize it.
	* gc-malloc.c (descrease_mtrigger, increase_mtrigger): Use it.
	
	* gc-mark.c (scm_gc_mark_dependencies): Call scm_i_string_mark,
	scm_i_stringbuf_mark and scm_i_symbol_mark, as appropriate.
	* gc-card.c (scm_i_sweep_card):  Call scm_i_string_free,
	scm_i_stringbuf_free and scm_i_symbol_free, as appropriate.

2004-08-18  Kevin Ryde  <user42@zip.com.au>

	* arbiters.c (FETCH_STORE): New macro.
	(SCM_LOCK_VAL, SCM_UNLOCK_VAL): New constants.
	(SCM_LOCK_ARB, SCM_UNLOCK_ARB): Remove, effectively absorbed into
	scm_try_arbiter and scm_release_arbiter.
	(scm_try_arbiter, scm_release_arbiter): Use FETCH_STORE to get xchg
	for speed on i386, otherwise using mutex.

	* eq.c (scm_equal_p): Remove real==fraction and fraction==real, they
	must be #f according to R5RS.  (equal? follows eqv?, and for eqv? an
	exact and inexact is #f.)

	* fports.c (fport_print): Use scm_ttyname instead of ttyname directly,
	to get thread safety of scm_ttyname.

	* ports.c (ttyname): Remove prototype, unused.

	* socket.c (scm_init_socket): Add SOCK_SEQPACKET and SOCK_RDM.
	Reported by Michael Tuexen.

2004-08-13  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* load.c (scm_init_load_path): Do not pass NULL to
	scm_to_locale_string, which would happen when GUILE_LOAD_PATH is
	not set.  Thanks to Bill Schottstaedt.

2004-08-12  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* socket.c (scm_inet_aton, scm_inet_pton): Convert SCM strings to
	locale strings instead of accessing their internals.
	(scm_recv, scm_send, scm_recvfrom, scm_sendto): Use
	SCM_I_STRING_CHARS and SCM_I_STRING_LENGTH instead of
	SCM_STRING_CHARS and SCM_STRING_LENGTH.
	
	* simpos.c (scm_system): Convert SCM strings to locale strings
	instead of accessing their internals.

	* script.c (scm_compile_shell_switches): Convert version to locale
	string before printing it.

	* rdelim.c (scm_read_delimited_x): Avoid
	SCM_VALIDATE_SUBSTRING_SPEC_COPY and use scm_from_size_t instead
	of scm_from_long for the returned number of read characters.

	* ioext.c (scm_fdopen): Use scm_i_fdes_to_port together with
	scm_i_mode_bits to avoid accessing internals of SCM string from C.

	* filesys.c (STRING_SYSCALL, STRING2_SYSCALL): New helper macros.
	Use them instead of SCM_SYSCALL when one or two strings need to be
	converted into locale strings.
	(my_rename): New, gathers platform dependent code for renaming.
	(scm_rename): Use it.
	(scm_readlink, scm_copy_file): Convert SCM strings to locale
	strings instead of accessing their internals.
	(scm_basename, scm_dirname): Use SCM_I_STRING_CHARS and
	SCM_I_STRING_LENGTH instead of SCM_STRING_CHARS and
	SCM_STRING_LENGTH.
	
	* extensions.c (load_extension): Convert lib and init to locale
	strings instead of accessing the internals directly.
	(scm_c_load_extension): Use scm_from_locale_string instead of
	scm_makfrom0str.
	
	* fports.h, fports.c (scm_i_fdes_to_port): New, like
	scm_fdes_to_port, but take mode bits directly instead of as a C
	string.
	(scm_i_fdes_to_port): Implement using above.
	(scm_open_file): Use scm_i_fdes_to_port together with
	scm_i_mode_bits to avoid accessing internals of SCM string from C.
	* vports.c (scm_make_soft_port): Use scm_i_fdes_to_port together
	with scm_i_mode_bits to avoid accessing internals of SCM string
	from C.

	* ports.h (scm_i_mode_bits): New, same as scm_mode_bits but with a
	SCM string as argument.

	* ports.c (scm_i_void_port): New, like scm_void_port but take mode
	bits directly instead of C string.
	(scm_void_port): Implement using above.
	(scm_sys_make_void_port): Use scm_i_void_port together with
	scm_i_mode_bits to avoid accessing internals of SCM string.
	
	* strings.h, strings.c (scm_i_get_substring_spec): New.

	* socket.c, rw.c, deprecated.h, validate.h
	(SCM_VALIDATE_STRING_COPY): Deprecated.  Replaced all uses with
	SCM_VALIDATE_STRING plus SCM_I_STRING_CHARS or
	scm_to_locale_string, etc.
	(SCM_VALIDATE_SUBSTRING_SPEC_COPY): Deprecated.  Replaced as
	above, plus scm_i_get_substring_spec.
	
	* regex-posix.c, read.c, random.c, ramap.c, print.c, numbers.c,
	hash.c, gc.c, gc-card.c, convert.i.c, backtrace.c, strop.c,
	strorder.c, strports.c, struct.c, symbols.c, unif.c, ports.c: Use
	SCM_I_STRING_CHARS, SCM_I_STRING_UCHARS, and SCM_I_STRING_LENGTH
	instead of SCM_STRING_CHARS, SCM_STRING_UCHARS, and
	SCM_STRING_LENGTH, respectively.  Also, replaced scm_return_first
	with more explicit scm_remember_upto_here_1, etc, or introduced
	them in the first place.

	* posix.c (WITH_STRING): New helper macro.  Use it where one
	locale string is needed for a short piece of code.
	(STRING_SYSCALL): New helper macro.  Use it instead of SCM_SYSCALL
	when one locale string is needed.
	(scm_mkstemp): Convert tmpl to a locale string.
	(scm_putenv): Rewritten to use only C strings.
	(scm_setlocale, scm_crpt): Convert argument strings to locale
	strings.
	
2004-08-11  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* load.c (scm_primitive_load_path): Do not check for absolute
	filenames when scm_sys_search_load_path returns false, which will
	return absolute filenames unchanged.

2004-08-11  Marius Vollmer  <mvo@zagadka.de>

	* gc.c, procprop.c (scm_init_storage, scm_stand_in_procs,
	scm_stand_in_proc): Use a hastable for scm_stand_in_procs instead
	of an alist.  Thanks to Matthias Koeppe!

2004-08-10  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* strings.h, deprecated.h (SCM_STRING_COERCE_0TERMINATION_X):
	Moved from string.h to deprecated.h.

	* deprecated.c, deprecated.h (SCM_CHARS, SCM_LENGTH): Removed.

	* strings.h, strings.c (SCM_MAKE_STRING_TAG): Renamed to
	SCM_I_MAKE_STRING_TAG, changed all uses.
	(SCM_STRING_CHARS, SCM_STRING_UCHARS, SCM_STRING_LENGTH): Renamed
	to SCM_I_STRING_CHARS, SCM_I_STRING_UCHARS, and SCM_I_LENGTH
	respectively.  For a short time, the old names are still there as
	aliases.  Not all uses have been changed yet, but the ones in
	strings.c have.
	(SCM_STRING_MAX_LEN): Do not hardcode to 24 bits, compute from
	SCM_T_BITS_MAX.
	(scm_is_string, scm_from_locale_string, scm_from_locale_stringn,
	scm_take_locale_string, scm_take_locale_stringn,
	scm_to_locale_string, scm_to_locale_stringn,
	scm_to_locale_stringbuf): New.
	(scm_c_string2str, scm_c_substring2str): Deprecated by moving to
	deprecated.[hc]. Implemented in terms of the new functions above.
	(scm_take_str, scm_take0str, scm_mem2string, scm_str2string,
	scm_makfrom0str): Reimplemented in terms of the new functions from
	above.  They will be discouraged shortly.
	(scm_substring): Do not use scm_mem2string.
	(scm_i_allocate_string_pointers, scm_i_free_string_pointers): New,
	to replace similar code from posix.c, simpos.c, and dynl.c.
	(scm_string_append): Use memcpy instead of explicit loop.  Do not
	use register keyword.  Use plain 'char' instead of 'unsigned
	char'.
	
	* strports.c (scm_mkstrport): Use SCM_I_STRING_UCHARS instead of
	SCM_STRING_UCHARS.  Use SCM_I_STRINGP instead of SCM_STRINGP.

	* strop.c (scm_i_index): Replaced SCM_STRINGP, SCM_STRING_CHARS,
	and SCM_STRING_LENGTH with SCM_I_STRINGP, SCM_I_STRING_CHARS, and
	SCM_I_STRING_LENGTH, respectively.  Pass string object directly,
	not as a pointer.  Use scm_remember_upto_here_1 to protect it.

	* read.c (scm_input_error): Use a SCM value for 'fn', not a C
	string.  This avoids a conversion round-trip.

	* gh_data.c: Replaced SCM_STRINGP, SCM_STRING_CHARS, and
	SCM_STRING_LENGTH with SCM_I_STRINGP, SCM_I_STRING_CHARS, and
	SCM_I_STRING_LENGTH, respectively.
	(gh_scm2newstr): Implement in terms of scm_to_locale_string.
	
	* environments.c: Instead calling scm_puts on the SCM_STRING_CHARS
	of a string, call scm_display on the string itself.

	* dynwind.c, dynwind.h (scm_frame_free): New.

	* stime.c, socket.c, simpos.c, procs.c, posix.c, ports.c,
	net_db.c, fports.c, filesys.c, eval.c, deprecation.c, dynl.c:
	Replaced uses of SCM_STRING_CHARS with proper uses of
	scm_to_locale_string.  Replaced SCM_STRINGP with scm_is_string.
	Replaced scm_mem2string with scm_from_locale_string.

	* simpos.c, posix.c (allocate_string_pointers, environ_list_to_c):
	Removed, replaced all uses with scm_i_allocate_string_pointers.

	* load.h, load.c (scm_internal_parse_path): Removed.
	(scm_parse_path): Use scm_string_split to do the work.
	(scm_init_load_path): Use scm_parse_path instead of
	scm_internal_parse_path.
	(scm_search_path): Rewritten string handling part of the code in
	terms of scm_to_locale_stringbuf and so that it is thread safe.

	* error.c (scm_error_scm): Throw directly instead of calling
	scm_error, this avoids the back and forth conversion of SUBR and
	MESSAGE and also plugs a memory leak.
	(scm_error): Call scm_error_scm.
	
	* backtrace.c: Replaced SCM_STRINGP with scm_is_string.
	(display_header): Print FNAME when it is true, not
	merely when it is a string.

	* strings.h (SCM_SET_STRING_LENGTH, SCM_SET_STRING_CHARS): Removed
	unceremoniously.  They were unused by Guile itself, and external
	use should stop immediately.
	

2004-08-10  Marius Vollmer  <mvo@zagadka.de>

	* numbers.h, number.c, deprecated.h, deprecated.c (scm_round,
	scm_truncate): Renamed to scm_c_round and scm_c_truncate;
	deprecated versions installed in deprecated.h and deprecated.c.
	Changed all uses.
	
2004-08-06  Rob Browning  <rlb@defaultvalue.org>

	* net_db.c (scm_resolv_error): don't cause an exception while
	trying to throw an exception -- call scm_misc_error with correct
	arguments.  The previous arguments needed a format escape that
	wasn't in any of the format strings.

2004-08-06  Kevin Ryde  <user42@zip.com.au>

	* ramap.c (scm_array_fill_x): For byvect char fill, force signed char
	so as not to depend on signedness of plain char.  For byvect range
	check, throw out-of-range rather than wrong-type-arg.

	* unif.c (scm_uniform_vector_ref, scm_array_set_x): For byvect, force
	signed byte range checks by using scm_to_schar and scm_from_schar,
	don't want to depend on signedness of C char.

2004-08-05  Kevin Ryde  <user42@zip.com.au>

	* arbiters.c (scm_try_arbiter): Use scm_i_misc_mutex instead of
	SCM_DEFER_INTS.
	(scm_release_arbiter): Use scm_i_misc_mutex so return value can be
	guaranteed if multiple threads compete to unlock.
	Update docstrings per doc/ref/api-scheduling.texi.

	* filesys.c (scm_copy_file): Use fstat on the input fd rather than
	stat on the filename, to be certain a file rename can't mean we get
	info on one filesystem object but open another.  This fstat usage is
	similar to Emacs copy-file.
	
	* posix.c (scm_setgroups): Enhance docstring, per doc/ref/posix.texi.

	* simpos.c (scm_system_star): Change scm_from_long to scm_from_int on
	SIGINT and SIGQUIT, since those values are ints.

2004-08-03  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* num2integral.i.c, num2float.i.c: Removed.
	* Makefile.am (noinst_HEADERS): Updated.

	* numbers.h. numbers.c (scm_make_ratio): Renamed to
	scm_i_make_ratio and made static, replaced uses with scm_divide.
	(scm_complex_p): New, export as "complex?" to Scheme.
	(scm_number_p): Export as "number?" to Scheme.
	(scm_is_complex, scm_is_number): New.
	(scm_c_make_rectangular, scm_c_make_polar): New.
	(scm_make_rectangular, scm_make_polar): Use above.
	(scm_c_real_part, scm_c_imag_part, scm_c_magnitude, scm_c_angle):
	New.
	(scm_make_complex): Discouraged by moving to discouraged.h and
	discouraged.c.  Replaced all uses with scm_c_make_rectangular.
	
	* discouraged.h, discouraged.c, numbers.c, numbers.h
	(scm_is_rational): New.
	(scm_i_short2big, scm_i_int2big, scm_i_uint2big, scm_i_size2big,
	scm_i_ptrdiff2big, scm_i_long_long2big, scm_i_ulong_long2big):
	Removed prototypes.
	(scm_make_real, scm_num2dbl, scm_float2num, scm_double2num):
	Discouraged by moving to discouraged.h and discouraged.c.
	Replaced all uses with scm_from_double.
	(scm_num2float, scm_num2double): Discouraged by moving prototype
	to discouraged.h and rewriting in terms of scm_to_double.
	Replaced all uses with scm_to_double.
	(scm_to_double): Do not implement in terms of scm_num2dbl, use
	explicit code.
	(scm_from_double): Do not implement in terms of scm_make_real, use
	explicit code.
	
2004-08-02  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* init.c (scm_init_guile_1): Call scm_i_init_discouraged.

	* gen-scmconfig.h.in (SCM_I_GSC_ENABLE_DISCOURAGED): New.
	* gen-scmconfig.c (SCM_ENABLE_DISCOURAGED): Emit based on above.

	* eval.c (SCM_EVALIM, SCM_EVALIM2, SCM_XEVAL, SCM_XEVALCAR):
	Renamed to SCM_I_* in order to avoid collisions with the versions
	defined in deprecated.h.

	* discouraged.h, discouraged.c: New files.

	* deprecated.h (SCM_FALSEP, SCM_NFALSEP, SCM_BOOLP, SCM_EQ_P,
	SCM_NEGATE_BOOL, SCM_BOOL, SCM_BOOT_NOT): Promoted from being
	deprecated to being discouraged by moving to discouraged.h.

	* numbers.h, numbers.c, discouraged.h, discouraged.c
	(scm_short2num, scm_ushort2num, scm_int2num, scm_uint2num,
	scm_long2num, scm_ulong2num, scm_size2num, scm_ptrdiff2num,
	scm_num2short, scm_num2ushort, scm_num2int, scm_num2uint,
	scm_num2long, scm_num2ulong, scm_num2size, scm_num2ptrdiff,
	scm_long_long2num, scm_ulong_long2num, scm_num2long_long,
	scm_num2ulong_long): Discouraged by moving to discouraged.h and
	discouraged.c and reimplementing in terms of scm_from_* and
	scm_to_*.  Changed all uses to the new scm_from_* and scm_to_*
	functions.

	* numbers.h, numbers.c: Removed GUILE_DEBUG code.
	(scm_i_short2big, scm_i_ushort2big, scm_i_int2big, scm_i_uint2big,
	scm_i_size2big, scm_i_ptrdiff2big): Removed.
	(scm_i_long2big, scm_i_ulong2big): New, explicit definitions.  
	* conv-integer.i.c, conv-uinteger.i.c: Use them instead of
	explicit code.

2004-08-02  Kevin Ryde  <user42@zip.com.au>

	* __scm.h (SCM_DEFER_INTS, SCM_ALLOW_INTS): Add comments about past
	and current usage and migration.

2004-07-31  Kevin Ryde  <user42@zip.com.au>

	* error.c (scm_strerror): Use scm_i_misc_mutex around strerror since
	it's not thread safe.
	(scm_syserror): Use scm_strerror rather than SCM_I_STRERROR, to take
	advantage of this.
	* fports.c (scm_open_file): Use scm_strerror likewise.
	* filesys.c (scm_stat, scm_lstat): Ditto.

	* filesys.c (scm_copy_file): Avoid fd leak when destination file
	cannot be opened.

	* symbols.c (scm_gensym): Use scm_i_misc_mutex around gensym_counter
	update, for thread safety.
	(gensym_counter): Move into scm_gensym which is its only user.
	(scm_init_symbols): No need to explicitly initialize gensym_counter.

2004-07-29  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* numbers.h (scm_to_schar, scm_to_uchar, scm_to_char,
	scm_to_short, scm_to_ushort, scm_to_int, scm_to_uint, scm_to_long,
	scm_to_ulong, scm_to_long_long, scm_to_ulong_long, scm_to_intmax,
	scm_to_uintmax, scm_to_size_t, scm_to_ssize_t scm_from_schar,
	scm_from_uchar, scm_from_char, scm_from_short, scm_from_ushort,
	scm_from_int, scm_from_uint, scm_from_long, scm_from_ulong,
	scm_from_long_long, scm_from_ulong_long, scm_from_intmax,
	scm_from_uintmax, scm_from_size_t, scm_from_ssize_t): No longer
	defined in terms of scm_to_signed_integer, etc, but in terms of
	scm_to_int8, etc.

	* gen-scmconfig.c (SCM_SIZEOF_INTMAX, SCM_SIZEOF_SIZE_T): New.

	* gen-scmconfig.h.in: Removed SCM_I_GSC_*_LIMITS macros, they are
	no longer used.

	* __scm.h (SCM_I_UTYPE_MAX, SCM_I_TYPE_MAX, SCM_I_TYPE_MIN,
	SCM_I_SIZE_MAX, SCM_I_SSIZE_MIN, SCM_I_SSIZE_MAX): New.

	* __scm.h, gen-scmconfig.c (SCM_I_LLONG_MAX, SCM_I_LLONG_MIN,
	SCM_I_ULLONG_MAX, SCM_T_INT8_MIN, SCM_T_INT8_MAX, SCM_T_UINT8_MAX,
	SCM_T_INT16_MIN, SCM_T_INT16_MAX, SCM_T_UINT16_MAX,
	SCM_T_INT32_MIN, SCM_T_INT32_MAX, SCM_T_UINT32_MAX,
	SCM_T_INT64_MIN, SCM_T_INT64_MAX, SCM_T_UINT64_MAX,
	SCM_T_INTMAX_MIN, SCM_T_INTMAX_MAX, SCM_T_UINTMAX_MAX): Moved
	definition into __scm.h, using new SCM_I_TYPE_MIN, etc.
	
	* conv-integer.i.c, conv-uinteger.i.c: New files, used to generate
	the functions below.

	* Makefile.am (noinst_HEADERS): Added conv-integer.i.c and
	conv-uinteger.i.c.

	* numbers.c, numbers.h (scm_to_int8, scm_to_uint8, scm_to_int16,
	scm_to_uint16, scm_to_int32, scm_to_uint32, scm_to_int64,
	scm_to_uint64, scm_from_int8, scm_from_uint8, scm_from_int16,
	scm_from_uint16, scm_from_int32, scm_from_uint32, scm_from_int64,
	scm_from_uint64): Turned from macros into proper functions.
	(scm_to_signed_integer, scm_to_unsigned_integer,
	scm_from_signed_integer, scm_from_unsigned_integer): Generate via
	conv-integer.i.c and conv-uinteger.i.c, as well.
	
	* number.h (scm_to_ssize_t, scm_to_size_t): Use the new
	SCM_I_SSIZE_MIN, SCM_I_SSIZE_MAX, and SCM_I_SIZE_MAX macros for
	the limits.  Those are always defined.
	
2004-07-29  Kevin Ryde  <user42@zip.com.au>

	* posix.c (scm_ttyname): Use scm_i_misc_mutex for thread safety.

2004-07-28  Kevin Ryde  <user42@zip.com.au>

	* posix.c (scm_ctermid): Use an L_ctermid buf on the stack, for thread
	safety.

	* unif.c (scm_array_set_x): For svect, use scm_num2short for
	consistency with other vector types and to get arg and func name into
	error message.

2004-07-27  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* deprecated.h (SCM_FALSEP, SCM_NFALSEP, SCM_BOOLP, SCM_BOOL):
	Reimplement using scm_is_false, scm_is_true, scm_is_bool, and
	scm_from_bool, respectively.
	(SCM_NINUMP): Added.
	
	* tags.h, deprecated.h (SCM_EQ_P): Deprecated by moving it into
	deprecated.h.  Replaced all uses with scm_is_eq.

2004-07-24  Kevin Ryde  <user42@zip.com.au>

	* threads.c, threads.h (scm_i_misc_mutex): New SCM_GLOBAL_MUTEX.
	* posix.c (scm_crypt): Use it to protect static data in crypt().

2004-07-23  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* deprecated.h, deprecated.c, numbers.h (SCM_INUMP, SCM_NINUMP,
	SCM_INUM): Deprecated by renaming them to SCM_I_INUMP,
	SCM_I_NINUMP and SCM_I_INUM, respectively and adding deprecated
	versions to deprecated.h and deprecated.c.  Changed all uses to
	either use the SCM_I_ variants or scm_is_*, scm_to_*, or
	scm_from_*, as appropriate.

	* dynwind.c (scm_i_dowinds): Removed unused code that would call
	the unexisting scm_cross_dynwind_binding_scope for inums on the
	windlist.
	
2004-07-10  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* socket.c (ipv6_net_to_num, scm_from_ipv6): Renamed
	ipv6_net_to_num to scm_from_ipv6, for converting from an IPv6
	byte-wise address to a SCM integer.  Changed all uses.
	(ipv6_num_to_net, scm_to_ipv6): Renamed ipv6_num_to_net to
	scm_to_ipv6 and added type and range checking, for converting from
	an IPv& byte-wise address to a SCM integer.  Changed all uses.
	(bignum_in_ipv6_range_p, VALIDATE_INET6): Removed, their function
	is now done by scm_to_ipv6.
	
	* numbers.c (scm_to_signed_integer, scm_to_unsigned_integer): dot
	not accept inexact integers.

	* validate.h, deprecated.h (SCM_VALIDATE_INUM,
	SCM_VALIDATE_INUM_COPY, SCM_VALIDATE_BIGINT,
	SCM_VALIDATE_INUM_MIN, SCM_VALIDATE_INUM_MIN_COPY,
	SCM_VALIDATE_INUM_MIN_DEF_COPY,SCM_VALIDATE_INUM_DEF,
	SCM_VALIDATE_INUM_DEF_COPY, SCM_VALIDATE_INUM_RANGE,
	SCM_VALIDATE_INUM_RANGE_COPY): Deprecated because they make the
	fixnum/bignum distinction visible.  Changed all uses to
	scm_to_size_t or similar.

2004-07-09  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* cpp_cnvt.awk: Use scm_from_int instead of SCM_MAKINUM.

2004-07-10  Kevin Ryde  <user42@zip.com.au>

	* hash.c (scm_hashq, scm_hashv, scm_hash): Restrict to size>=1 rather
	than size>=0, since 0<=hash<size cannot be satisfied for size==0, and
	such a size causes divide-by-zeros in scm_hasher.

        * regex-posix.c (scm_make_regexp): Free rx on error, to avoid memory
	leak.

2004-07-08  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* numbers.c (scm_is_signed_integer, scm_is_unsigned_integer):
	Rewritten using the same logic as scm_to_signed_integer and
	scm_to_unsigned_integer, respectively, which is better(tm).  Also,
	use CHAR_BIT instead of hardcoding 8.
	(LLONG_MIN, LLONG_MAX, ULLONG_MAX): Removed and used
	SCM_I_LLONG_MIN etc. instead.
	
	* numbers.h (SCM_MAKINUM, SCM_I_MAKINUM): Renamed SCM_MAKINUM to
	SCM_I_MAKINUM and changed all uses.
	* deprecated.h, deprecated.c (SCM_MAKINUM): Newly deprecated.

	* gen-scmconfig.c (SCM_I_LLONG_MIN, SCM_I_LLONG_MAX,
	SCM_I_ULLONG_MAX): Instead of hard-coding the numbers, compute
	them by assuming twos-complement.

2004-07-07  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* gen-scmconfig.h.in: Added all the new SCM_I_GSC_*_LIMITS that
	configure now produces.
	* gen-scmconfig.c: Use them to output SCM_T_INT8_MIN, etc
	definitions, giving the limits of the integer types defined by
	Guile. Also, output a hard coded SCM_I_LLONG_MIN, etc since
	LLONG_MIN or LONG_LONG_MIN is hard to get at.

	* numbers.h (scm_to_short, scm_to_ushort): It's SHRT_MIN, etc, not
	SHORT_MIN.
	(scm_to_size_t): Use SIZE_MAX instead of cooking our own.
	(scm_to_long_long, scm_to_ulong_long, scm_to_int8, scm_to_uint8,
	scm_to_int16, scm_to_uint16, scm_to_int32, scm_to_uint32,
	scm_to_int64, scm_to_uint64, scm_to_intmax, scm_to_uintmax,
	scm_from_long_long, scm_from_ulong_long, scm_from_int8,
	scm_from_uint8, scm_from_int16, scm_from_uint16, scm_from_int32,
	scm_from_uint32, scm_from_int64, scm_from_uint64, scm_from_intmax,
	scm_from_uintmax): New.
	
2004-07-06  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* tags.h (scm_is_eq): New.

	* deprecated.h, boolean.h (SCM_FALSEP, SCM_NFALSEP, SCM_BOOL,
	SCM_NEGATE_BOOL, SCM_BOOLP): Deprecated by moving into
	"deprecated.h".  Replaced all uses with scm_is_false, scm_is_true,
	scm_from_bool, and scm_is_bool, respectively.

	* boolean.h (scm_is_bool): Fix bug in prototype.
	(scm_from_bool): The argument is "x" not "f", stupid.

	* boolean.c (scm_is_bool): Fix typo.

	* numbers.h, numbers.c (scm_is_integer, scm_is_signed_integer,
	scm_is_unsigned_integer, scm_to_signed_integer,
	scm_to_unsigned_integer, scm_to_schar, scm_to_uchar, scm_to_char,
	scm_to_short, scm_to_ushort, scm_to_long, scm_to_ulong,
	scm_to_size_t, scm_to_ssize_t, scm_from_schar, scm_from_uchar,
	scm_from_char, scm_from_short, scm_from_ushort, scm_from_int,
	scm_from_uint, scm_from_long, scm_from_ulong, scm_from_size_t,
	scm_from_ssize_t, scm_is_real, scm_to_double, scm_from_double):
	New.
	
2004-07-05  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* boolean.h, boolean.c (scm_is_true, scm_is_false, scm_from_bool,
	scm_to_bool): New.

2004-06-27  Dirk Herrmann  <dirk@dirk-herrmanns-seiten.de>

	* backtrace.c (display_expression, display_frame): Call
	scm_i_unmemoize_expr for unmemoizing a memoized object holding a
	single memoized expression.

	* debug.c (memoized_print): Don't try to unmemoize the memoized
	object, since we can't know whether it holds a single expression
	or a body.

	(scm_mem_to_proc): Removed check for lambda expression, since it
	was moot anyway.  Whoever uses these functions for debugging
	purposes should know what they do: Creating invalid memoized code
	will cause crashes, independent of whether this check is present
	or not.

	(scm_proc_to_mem): Take the closure's code as it is and don't
	append a SCM_IM_LAMBDA isym.  To allow easier debugging, the
	memoized code should not be modified.

	* debug.[ch] (scm_unmemoize, scm_i_unmemoize_expr): Removed
	scm_unmemoize from public use, but made scm_i_unmemoize_expr
	available as a guile internal function instead.  However,
	scm_i_unmemoize_expr will only work on memoized objects that hold
	a single memoized expression.  It won't work with bodies.

	* debug.c (scm_procedure_source), macros.c (macro_print), print.c
	(scm_iprin1): Call scm_i_unmemocopy_body for unmemoizing a body,
	i. e. a list of expressions.

	* eval.c (unmemoize_exprs): Drop internal body markers from the
	output during unmemoization.

	* eval.[ch] (scm_unmemocopy, scm_i_unmemocopy_expr,
	scm_i_unmemocopy_body): Removed scm_unmemocopy from public use,
	but made scm_i_unmemocopy_expr and scm_i_unmemocopy_body available
	as guile internal functions instead.  scm_i_unmemoize_expr will
	only work on a single memoized expression, while
	scm_i_unmemocopy_body will only work on bodies.

2004-06-21  Dirk Herrmann  <dirk@dirk-herrmanns-seiten.de>

	* eval.c (unmemoize_exprs): Handle semi-memoized code.

	(scm_cons_source, scm_primitive_eval): Prefer higher level
	predicate SCM_FALSEP over SCM_IMP.

2004-06-15  Rob Browning  <rlb@defaultvalue.org>

	* script.c (scm_shell_usage): minor phrasing change.

	* gc_os_dep.c: update ifdefery for macosx.
	(scm_get_stack_base): separate result initialization from
	declaration to slience warnings with macosx and hp-ux using gcc