(scm_lreadr): Bugfix: include the last bit in the bit vector.

[bpt/guile.git] / libguile / ChangeLog

2004-10-29  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* unif.h, unif.c (scm_array_creator): New.
	(scm_i_get_old_prototype): New.
	(scm_array_prototype): use it to return old-style prototype, never
	return creators.
	(scm_make_uve): Use scm_call_1 instead of scm_call_2 with a second
	arg of SCM_UNDEFINED.  The latter is wrong.
	
	* unif.h, unif.c (scm_make_u1vector): New, but only temporary.
	(make_uve): Removed.
	(scm_i_proc_make_vector, scm_i_proc_make_string,
	scm_i_proc_make_u1vector): New.
	(scm_init_unif): Initialize them.
	(scm_i_convert_old_prototype): New.
	(scm_make_uve): Use it to get the creator procedure.  Removed all
	old code that created old-style uniform vectors.
	(scm_array_p): Handle generic vectors.
	(scm_dimensions_to_uniform_array): Do not fill new array with
	prototype when that is a procedure.
	(scm_list_to_uniform_array): Also accept a list of lower bounds as
	the NDIM argument.
	(scm_i_print_array): Print rank for shared or non-zero-origin
	vectors.
	(tag_proto_table, scm_i_tag_to_prototype, scm_i_read_array): New.
	(scm_raprin1): Do not call scm_i_array_print for enclosed arrays,
	which I do not understand yet.
	(scm_array_prototype): Explicitely handle generic vectors.

	* numbers.c, number.h (scm_i_print_complex, icmplx2str): New.
	(iflo2str): Use icmplx2str for complex numbers.

	* srfi-4.c, srfi-4.h (scm_i_read_homogenous_vector,
	scm_i_uniform_vector_prototype): Removed.
	(scm_i_uniform_vector_creator): New.
	(SCM_UVEC_C32, scm_c32vector, scm_make_c32vector, etc,
	SCM_UVEC_C64, scm_c64vector, scm_make_c64vector, etc): New.
	Updated all tables and generic functions to support them.
	(scm_i_proc_make_u8vector, scm_i_proc_make_s8vector, etc): New.
	(scm_init_srfi_4): Initialize them.
	
	* srfi-4.i.c (scm_take_u8vector, etc): use uvec_sizes instead of
	sizeof(CTYPE) as explained in the comment.

	* read.c (scm_lreadr): Call scm_i_read_array for all characters
	followinf '#' that can start an array.  Explicitely disambiguate
	'i' and 'e' between introducing numbers and uniform vectors.  Do
	not call scm_i_read_homogenous_vector, since that is also handled
	by scm_i_read_array now.

2004-10-27  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	First cut at integrating uniform vectors from srfi-4 with the rest
	of Guile.  This change replaces scm_tc7_byvect with a s8 uniform
	vector.  The plan is to gradually replace one type after the other
	until none is left and then to consider general cleanups and
	optimizations.
	
	* srfi-4.h, srfi-4.i.c (scm_u8vector_elements, etc): New.

	* srfi-4.h, srfi-4.c (scm_uniform_vector_p,
	scm_uniform_vector_ref, scm_uniform_vector_set_x,
	scm_uniform_vector_to_list, scm_is_uniform_vector,
	scm_c_uniform_vector_lengths, scm_c_uniform_vector_size,
	scm_uniform_vector_elements, scm_uniform_vector_element_size,
	scm_uniform_vector_release): New.
	(scm_i_uniform_vector_prototype, scm_i_uniform_vector_tag): New.
	(scm_uniform_element_size, scm_uniform_vector_length): Moved here
	from unif.h, unif.c and extended to handle both the old and new
	uniform vectors.

	* tags.h (scm_tc7_byvect): Commented out.

	* unif.h, unif.c (scm_uniform_vector_ref, scm_array_ref): Renamed
	the former to the latter.
	(scm_uniform_vector_length, scm_uniform_element_size): Moved to
	srfi-4.h, srfi-4.c.
	(scm_make_uve): Call scm_make_s8vector for #\nul prototype.
	(scm_array_p, scm_array_rank, scm_array_dimensions,
	scm_transpose_array, scm_enclose_array, scm_array_ref, scm_cvref,
	scm_array_set_x, scm_array_contents, scm_uniform_array_read_x,
	scm_array_to_list, scm_array_prototype): Handle srfi-4 uniform
	vectors.  Removed code for scm_tc7_byvect.
	(scm_dimensions_to_uniform_array): Fill array with 0 when
	prototype is #\nul.
	(scm_i_print_array_dimension, scm_i_legacy_tag,
	scm_i_print_array): New.
	(scm_raprin1): Call scm_i_print_array for arrays.  Removed code
	for scm_tc7_byvect.

	* ramap.c (scm_ra_matchp, scm_array_fill_int, racp,
	scm_array_index_map_x, raeql_1, scm_array_equal_p): Handle srfi-4
	uniform vectors.  Removed code for scm_tc7_byvect

	* print.c (iprin1): Removed code for scm_tc7_byvect.
	* objects.c (scm_class_of): Likewise.
	* gc-mark.c (scm_gc_mark_dependencies): Likewise.
	* gc-card.c (scm_i_sweep_card): Likewise.
	* evalext.c (scm_self_evaluating_p): Likewise.
	* eq.c (scm_equal_p): Likewise.

	* gh_data.c (gh_chars2byvect): Reimplemented with
	scm_make_s8vector.
	(gh_scm2chars): Handle s8vectors, removed code for scm_tc7_byvect.

	* srfi-4.c (take_uvec): New.
	(alloc_uvec): Use it.
	* srfi-4.h, srfi-4.i.c (scm_take_u8vector, etc): New.

	* random.c (vector_scale, vector_scale_x): Renamed former to the
	latter, since it modifies its argument.
	(vector_scale_x, vector_sum_squares, scm_random_normal_vector_x):
	Do not use scm_universal_vector_length for non-uniform vectors.
	Use scm_f64vector_elements to access innards of uniform vectors.

	* convert.i.c: Convert srfi-4 style uniform vectors when
	requested.
	* convert.c (scm_c_scm2chars, scm_c_chars2scm,
	scm_c_chars2byvect): Use a s8vector instead of a scm_tc7_byvect.
	
2004-10-26  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* numbers.h, numbers.c (scm_i_print_double): New.

	* srfi-4.c, srfi-4.h, srfi-4.i.c: New files, initially from
	../srfi/ but heavily modified.
	* Makefile.am: Add them in all the right places.
	* init.c (scm_init_guile_1): Call scm_init_srfi_4.
	* read.c (scm_lreadr): Call scm_i_read_homogenous_vector for '#f',
	'#u', and '#s'.

	* read.h, read.c (scm_i_input_error): Renamed from scm_input_error
	and made non-static.  Changed all uses.

2004-10-22  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* variable.c, threads.c, struct.c, stackchk.c, smob.c, root.c,
	print.c, ports.c, mallocs.c, hooks.c, hashtab.c, fports.c,
	guardians.c, filesys.c, coop-pthreads.c, continuations.c: Use
	scm_uintprint to print unsigned integers, raw heap words, and
	adresses, using a cast to scm_t_bits to turn pointers into
	integers.

	* unif.c: Include "libguile/print.h".

	* numbers.h, numbers.c (SCM_T_INTBUFLEN): Increased to cover
	scm_t_intmax values.
	(scm_uint2str): New, for scm_t_uintmax.
	(scm_iint2str): Argument type changed to scm_t_intmax,
	reimplemented in terms of scm_uint2str.
	
	* print.c, print.h (scm_uintprint): New, for printing
	scm_t_uintmax values.
	(scm_intprint): Argument type changed to scm_t_intmax.
	
	* sort.c (quicksort, scm_merge, scm_merge_list_x,
	scm_merge_list_step, scm_merge_vector_step): Inserted SCM_TICKs at
	strategic places so that the loops can be interrupted.

	* Makefile.am (INCLUDES): Use @LTDLINC@ instead of
	"-I$(top_srcdir)/libguile-ltdl".
	(libguile_la_LIBADD): Use @LIBLTDL@ instead of
	"../libguile-ltdl/libguile-ltdl.a".
	
	* guile.c, dynl.c: Switched to using libltdl directly.  Replaced
	all references to scm_lt_* with just lt_*.  Include <ltdl.h>
	instead of <libguile-ltdl.h>.

2004-10-20  Marius Vollmer  <mvo@zagadka.de>

	* sort.c (quicksort): Copy pivot out of the array while
	constructing the partitions; it could get overwritten otherwise.
	Because of the ultimate insertion sort, this bug did not cause
	quicksort to fail, it just put all the burdon on the insertion
	sort and was thus very slow.  Thanks to Rolan Orre for reporting
	the slowness!
	
2004-10-19  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* numbers.c (scm_i_range_error): New.
	* conv-integer.i.c, conv-uinteger.i.c: Use it instead of
	scm_out_of_range.	

	* sort.c (scm_restricted_vector_sort_x): Validate startpos <=
	endpos.  State inclusiveness/exclusiveness of bounds in docstring.

	* unif.c (scm_array_p): When no prototype is given, explicitely
	test for allowable types, do not simply return true.  Thanks to
	Roland Orre for reporting this!

	* private-gc.h (SCM_DEFAULT_MAX_SEGMENT_SIZE): Increase to 20 Mib.

	* gc-segment.c (scm_i_get_new_heap_segment): Limit size of new
	segment to scm_max_segment_size.

2004-10-08  Han-Wen Nienhuys  <hanwen@xs4all.nl>

	* inline.h (scm_double_cell): abort if GC running.
	(scm_cell): idem.

2004-10-04  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* error.c (scm_wrong_type_arg): Do not talk about "argument" for
	pos == 0.

	Keywords no longer store a 'dash symbol'.  Instead, they store a
	symbol with their real name.
	
	* keywords.h, keywords.c, deprecated.h, deprecated.c
	(SCM_KEYWORDP, SCM_KEYWORDSYM): Deprecated and implemented in
	terms of scm_is_keyword and scm_keyword_dash_symbol.
	
	* keywords.h, keywords.c, discouraged.h, discouraged.c
	(scm_make_keyword_from_dash_symbol, scm_keyword_dash_symbol,
	scm_c_make_keyword): Discouraged.

	* keywords.h, keywords.c (scm_symbol_to_keyword,
	scm_keyword_to_symbol): Implemented in C.
	(scm_is_keyword, scm_from_locale_keyword,
	scm_from_locale_keywordn): New.

	* goops.c: Replaced SCM_KEYWORDP with scm_is_keyword.	

	* snarf.h (SCM_KEYWORD, SCM_GLOBAL_KEYWORD): Use
	scm_from_locale_keyword instead of scm_c_make_keyword.

	* keywords.c (scm_symbol_to_keyword): Use SCM_ASSERT_TYPE for a
	better error message.

	* deprecated.c: Include discouraged.h and keywords.h.

	* read.c (scm_lreadr): Simply do (symbol->keyword (read)) after
	reading '#:' or ':'.  See NEWS for consequences.

2004-09-29  Marius Vollmer  <marius.vollmer@uni-dortmund.de>

	* read.c (scm_lreadr): Revert change from 2004-09-22: string
	literals are now read-write again (until SCM_STRING_CHARS is
	removed).

	* strings.c (SCM_STRING_CHARS): Explicitely reject read-only
	strings with an error message that blames SCM_STRING_CHARS.

	* options.c (change_option_setting): Use scm_car instead of
	explicit type check plus SCM_CAR.

	* print.h, print.c (SCM_PRINT_HIGHLIGHT_PREFIX,
	SCM_PRINT_HIGHLIGHT_SUFFIX): New printer options.
	(scm_iprin1): Use them instead of the previoulsy hardcoded
	strings.
	(scm_init_print): Initialize them.

	* backtrace.c (display_frame_expr): Do not remove control
	characters from the final string.  Print it directly using
	scm_display.

	* ramap.c (scm_array_equal_p): Include scm_tc7_svect in switch.
	Thanks to Roland Orre!

2004-09-29  Kevin Ryde  <user42@zip.com.au>

	* regex-posix.c (scm_regexp_exec): Correction to last change, should
	be whole original string in match struct, not offsetted substring.

2004-09-24  Han-Wen Nienhuys   <hanwen@xs4all.nl>

	* gc.c (scm_gc_unprotect_object): abort if called during GC.

2004-09-24  Marius Vollmer  <mvo@zagadka.de>

	* Makefile.am (EXTRA_DIST): Added gettext.h.

	* smob.c, smob.h (scm_assert_smob_type): New.

	* Makefile.am (guile_CFLAGS, guile_LDFLAGS, libguile_la_CFLAGS):
	Include GUILE_CFLAGS.
	(libguile_la_LIBADD): Removed THREAD_LIBS_LOCAL, which is unused
	now.
	(libpath.h): Put GUILE_CFLAGS in the build-info.
	
2004-09-23  Marius Vollmer  <mvo@zagadka.de>

        * print.h (scm_print_state): Added highlight_objects.
	* print.c (make_print_state, scm_free_print_state): Initialize it
	to SCM_EOL.
	(scm_iprin1): Wrap output in '{...}' when object is contained in
	highlight_objects.

	* backtrace.h, backtrace.c (scm_display_backtrace_with_highlights,
	scm_backtrace_with_highlights): New.  Set highlight_objects of
	printstate.
	
	* error.c (scm_error_scm): Document new meaning of data/rest
	argument for out-of-range and wrong-type-arg errors.
	(scm_out_of_range, scm_out_of_range_pos, scm_wrong_type_arg,
	scm_wrong_type_arg_msg): Pass bad_value in rest argument of
	exception so that it gets highlighted in the backtrace.
	Don't talk about "argument" when not giving a position.
	
	* throw.c (handler_message): The rest argument is the fourth
	argument, not everything after the third.  Call
	scm_display_backtrace_with_highlights, passing the rest argument
	when appropriate.

2004-09-22  Marius Vollmer  <mvo@zagadka.de>

	From Jan Nieuwenhuizen <janneke@gnu.org> and Bruno Haible
	<bruno@clisp.org>:

	* i18n.c: Handle --disable-nls (thanks Bruno).

	* posix.c (scm_init_posix): Add LC_PAPER, LC_NAME, LC_ADDRESS,
	LC_TELEPHONE, LC_MEASUREMENT, LC_IDENTIFICATION.

	* i18n.c (scm_i_to_lc_category): New name and export.  Support all
	LC categories.
	* posix.c (scm_setlocale): Use it.

	* i18n.h, i18n.c (scm_textdomain, scm_bindtextdomain,
	scm_bind_textdomain_codeset): Make wrappers similar to C function
	they wrap.

	* i18n.h: New file.
	* i18n.c: New file.
	* gettext.h: New file, taken from GNU gettext.
	* init.c: Include libguile/i18n.h.
	(scm_init_guile_1): Add call to scm_init_i18n().
	* Makefile.am (libguile_la_SOURCES): Add i18n.c.
	(DOT_X_FILES): Add i18n.x.
	(DOT_DOC_FILES): Add i18n.doc.
	(libguile_la_LDFLAGS): Add @LTLIBINTL@.
	(modinclude_HEADERS): Add i18n.h.

2004-09-22  Marius Vollmer  <mvo@zagadka.de>

	* gh_list.c: Replaced SCM_CAR, etc with scm_car, etc.

	* discouraged.h, tags.h (SCM_CONSP, SCM_NCONSP): Moved to
	discouraged.h.  Replaced all uses with scm_is_pair.
	(SCM_I_CONSP): New name for SCM_CONSP.

	* pairs.h, pairs.c (scm_is_pair, scm_is_null, scm_car, scm_cdr,
	scm_i_chase_pairs, SCM_I_A_PAT, SCM_I_D_PAT, etc, scm_caar,
	scm_cadr, etc): New.
	(SCM_NULLP, SCM_NNULLP): Moved to discouraged.h.  Replaced all
	uses with scm_is_null.

	* eval.c (scm_eval, scm_apply, call_cxr_1): Use scm_i_chase_pairs
	instead of explicit code.

2004-09-22  Marius Vollmer  <mvo@zagadka.de>

	* srfi-13.c (scm_string_contains, scm_string_contains_ci):
	Reworded logic a bit so that #f is returned immediately when s1 is
	too short to contain s2.

	* regex-posix.c (scm_regexp_exec): Convert string to
	zero-terminated locale string before matching against it.

	* strings.h, strings.c (scm_substring_read_only,
	scm_c_substring_read_only, scm_i_substring_read_only): New.
	(RO_STRING_TAG, IS_RO_STRING): New.
	(scm_i_string_writable_chars): Bail on read-only strings. 
	
	* read.c (scm_lreadr): use scm_c_substring_read_only for string
	literals, thus making them read-only as specified by R5RS.

2004-09-22  Marius Vollmer  <mvo@zagadka.de>

	* eq.c (scm_equal_p): Allow smobs with different flags to be equal
	by testing for smobs before insisting on equal SCM_CELL_TYPES.

2004-09-21  Marius Vollmer  <mvo@zagadka.de>

	* numbers.h, numbers.c: Include <gmp.h> in numbers.h, not in
	numbers.c.
	(scm_to_mpz, scm_from_mpz): New.