Change bitset word type from unsigned int to unsigned long int,
authorPaul Eggert <eggert@cs.ucla.edu>
Tue, 6 Sep 2005 07:36:48 +0000 (07:36 +0000)
committerPaul Eggert <eggert@cs.ucla.edu>
Tue, 6 Sep 2005 07:36:48 +0000 (07:36 +0000)
commitc4f640f1277934d02bad810d887dec6033f1ad4c
treea2698c9d6e497576a27a4a39bf7b1fd22d10f156
parent0762e23a4187592bc8e8fa064dbe1a08c77d2409
Change bitset word type from unsigned int to unsigned long int,
as this has better performance on typical 64-bit hosts.
Port bitset code to hosts with unusual word sizes.
* lib/regcomp.c (build_equiv_class, build_charclass):
(build_range_exp, build_collating_symbol):
Prefer bitset to re_bitset_ptr_t in prototypes, when the actual
argument is a bitset.  This is merely a style issue, but it makes
it clearer that an entire array is expected.
(re_compile_fastmap_iter, init_dfa, init_word_char, optimize_subexps):
* lib/regcomp.c (lower_subexp, parse_bracket_exp):
(built_charclass_op):
Port to the case where bitset_word is not the same as unsigned int.
* lib/regex_internal.h (bitset_set, bitset_clear, bitset_contain):
(bitset_not, bitset_merge, bitset_set_all, bitset_mask):
Likewise.
* lib/regexec.c (check_dst_limits_calc_pos_1):
(check_subexp_matching_top):
(build_trtable, group_nodes_into_DFAstates):
Likewise.
* lib/regcomp.c (re_compile_fastmap_iter, utf8_sb_map):
(optimize_utf8):
Don't assume that SBC_MAX is a multiple of BITSET_WORD_BITS.
* lib/regex_internal.h (bitset_set_all, bitset_not): Likewise.
* lib/regexec.c (group_nodes_into_DFAstates): Likewise.
* lib/regcomp.c (utf8_sb_map): Don't assume UINT_MAX == 0xffffffff.
* lib/regcomp.c (optimize_subexps, lower_subexp):
Work even if bitset_word has holes in its bitwise representation.
* lib/regex_internal.h (BITSET_WORD_BITS): Likewise.
* lib/regexec.c (check_dst_limits_calc_pos_1):
(heck_subexp_matching_top): Likewise.
* lib/regex_internal.c (re_string_reconstruct):
Don't assume UCHAR_MAX == 255.
* lib/regex_internal.h (bitset_set_all): Likewise.
* lib/regex_internal.h (BITSET_WORD_BITS): Renamed from UINT_BITS.
All uses changed.
(BITSET_WORDS): Renamed from BITSET_UINTS.  All uses changed.
(bitset_word): New type, replacing 'unsigned int' for bitset uses.
All uses changed.
(BITSET_WORD_MAX): New macro.
(bitset_set, bitset_clear, bitset_contain, bitset_empty):
(bitset_set_all, bitset_copy):  Now inline functions, not macros.
(bitset_empty, bitset_copy):
Prefer sizeof (bitset) to multiplying it out ourselves.
(bitset_not_merge): Remove; unused.
(bitset_contain): Return bool, not unsigned int with one bit on.
All callers changed.
* lib/regexec.c (build_trtable): Don't assume bitset has no stricter
alignment than re_node_set; do this by defining a new internal
type struct dests_alloc and using it to allocate memory.
* config/srclist.txt: Add glibc bug 1302.
config/ChangeLog
config/srclist.txt
lib/ChangeLog
lib/regcomp.c
lib/regex_internal.c
lib/regex_internal.h
lib/regexec.c