diff options
| -rw-r--r-- | ChangeLog | 15 | ||||
| -rw-r--r-- | locale/programs/ld-collate.c | 11 | ||||
| -rw-r--r-- | posix/fnmatch.c | 1 | ||||
| -rw-r--r-- | posix/fnmatch_loop.c | 474 | ||||
| -rw-r--r-- | posix/regex.c | 2 | ||||
| -rw-r--r-- | posix/tst-fnmatch.input | 57 |
6 files changed, 495 insertions, 65 deletions
diff --git a/ChangeLog b/ChangeLog index 3d39869572..32d4633a34 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,18 @@ +2000-07-05 Ulrich Drepper <drepper@redhat.com> + + * locale/loadlocale.c (_nl_unload_locale): Add cast to avoid warning. + + * locale/programs/ld-collate.c (collate_output): Also write out the + collation sequence values and the wide character string for the + collation symbol table. + + * posix/fnmatch.c: Include "../locale/elem-hash.h". + * posix/fnmatch_loop.c: Implement collating symbol handling. + * posix/tst-fnmatch.input: Add more tests, especially for collating + symbol handling. + + * posix/regex.c: Fix comment. + 2000-07-05 Andreas Jaeger <aj@suse.de> * sysdeps/mips/fpu_control.h: Fix type of fpu_control_t. diff --git a/locale/programs/ld-collate.c b/locale/programs/ld-collate.c index 97059c2593..89621c82e1 100644 --- a/locale/programs/ld-collate.c +++ b/locale/programs/ld-collate.c @@ -2611,6 +2611,17 @@ collate_output (struct localedef_t *locale, struct charmap_t *charmap, (sizeof (int32_t) - ((1 + namelen + 1 + runp->nmbs) % sizeof (int32_t)))); + + /* Now some 32-bit values: multibyte collation sequence, + wide char string (including length), and wide char + collation sequence. */ + obstack_int_grow (&extrapool, runp->mbseqorder); + + obstack_int_grow (&extrapool, runp->nwcs); + obstack_grow (&extrapool, runp->wcs, + runp->nwcs * sizeof (uint32_t)); + + obstack_int_grow (&extrapool, runp->wcseqorder); } } diff --git a/posix/fnmatch.c b/posix/fnmatch.c index c4b11080fe..62cfa5fee5 100644 --- a/posix/fnmatch.c +++ b/posix/fnmatch.c @@ -53,6 +53,7 @@ we support a correct implementation only in glibc. */ #ifdef _LIBC # include "../locale/localeinfo.h" +# include "../locale/elem-hash.h" # define CONCAT(a,b) __CONCAT(a,b) #endif diff --git a/posix/fnmatch_loop.c b/posix/fnmatch_loop.c index 005dd99d7b..b7baa7be24 100644 --- a/posix/fnmatch_loop.c +++ b/posix/fnmatch_loop.c @@ -387,7 +387,10 @@ FCT (pattern, string, no_leading_period, flags) const UCHAR *np = (const UCHAR *) n; idx2 = findidx (&np); -# if !WIDE_CHAR_VERSION +# if WIDE_CHAR_VERSION + if (idx2 != 0 && weights[idx] == weights[idx2]) + goto matched; +# else if (idx2 != 0 && len == weights[idx2]) { int cnt = 0; @@ -400,9 +403,6 @@ FCT (pattern, string, no_leading_period, flags) if (cnt == len) goto matched; } -# else - if (idx2 != 0 && weights[idx] == weights[idx2]) - goto matched; # endif } } @@ -415,13 +415,187 @@ FCT (pattern, string, no_leading_period, flags) return FNM_NOMATCH; else { - c = FOLD (c); - normal_bracket: - if (c == fn) - goto matched; + int is_seqval = 0; + int is_range = 0; - cold = c; - c = *p++; +#ifdef _LIBC + if (c == L('[') && *p == L('.')) + { + uint32_t nrules = + _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); + const CHAR *startp = p; + size_t c1 = 0; + + while (1) + { + c = *++p; + if (c == L('.') && p[1] == L(']')) + { + p += 2; + break; + } + if (c == '\0') + return FNM_NOMATCH; + ++c1; + } + + /* We have to handling the symbols differently in + ranges since then the collation sequence is + important. */ + is_range = *p == L('-') && p[1] != L('\0'); + + if (nrules == 0) + { + /* There are no names defined in the collation + data. Therefore we only accept the trivial + names consisting of the character itself. */ + if (c1 != 1) + return FNM_NOMATCH; + + if (!is_range && *n == startp[1]) + goto matched; + + cold = startp[1]; + c = *p++; + } + else + { + int32_t table_size; + const int32_t *symb_table; +# ifdef WIDE_CHAR_VERSION + char str[c1]; + int strcnt; +# else +# define str (startp + 1) +# endif + const unsigned char *extra; + int32_t idx; + int32_t elem; + int32_t second; + int32_t hash; + +# ifdef WIDE_CHAR_VERSION + /* We have to convert the name to a single-byte + string. This is possible since the names + consist of ASCII characters and the internal + representation is UCS4. */ + for (strcnt = 0; strcnt < c1; ++strcnt) + str[strcnt] = startp[1 + strcnt]; +#endif + + table_size = + _NL_CURRENT_WORD (LC_COLLATE, + _NL_COLLATE_SYMB_HASH_SIZEMB); + symb_table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_TABLEMB); + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_EXTRAMB); + + /* Locate the character in the hashing table. */ + hash = elem_hash (str, c1); + + idx = 0; + elem = hash % table_size; + second = hash % (table_size - 2); + while (symb_table[2 * elem] != 0) + { + /* First compare the hashing value. */ + if (symb_table[2 * elem] == hash + && c1 == extra[symb_table[2 * elem + 1]] + && memcmp (str, + &extra[symb_table[2 * elem + 1] + + 1], c1) == 0) + { + /* Yep, this is the entry. */ + idx = symb_table[2 * elem + 1]; + idx += 1 + extra[idx]; + break; + } + + /* Next entry. */ + elem += second; + } + + if (symb_table[2 * elem] != 0) + { + /* Compare the byte sequence but only if + this is not part of a range. */ +# ifdef WIDE_CHAR_VERSION + int32_t *wextra; + + idx += 1 + extra[idx]; + /* Adjust for the alignment. */ + idx = (idx + 3) & ~4; + + wextra = (int32_t *) &extra[idx + 4]; +# endif + + if (! is_range) + { +# ifdef WIDE_CHAR_VERSION + for (c1 = 0; c1 < wextra[idx]; ++c1) + if (n[c1] != wextra[1 + c1]) + break; + + if (c1 == wextra[idx]) + goto matched; +# else + for (c1 = 0; c1 < extra[idx]; ++c1) + if (n[c1] != extra[1 + c1]) + break; + + if (c1 == extra[idx]) + goto matched; +# endif + } + + /* Get the collation sequence value. */ + is_seqval = 1; +# ifdef WIDE_CHAR_VERSION + cold = wextra[1 + wextra[idx]]; +# else + /* Adjust for the alignment. */ + idx += 1 + extra[idx]; + idx = (idx + 3) & ~4; + cold = *((int32_t *) &extra[idx]); +# endif + + c = *p++; + } + else if (symb_table[2 * elem] != 0 && c1 == 1) + { + /* No valid character. Match it as a + single byte. */ + if (!is_range && *n == str[0]) + goto matched; + + cold = str[0]; + c = *p++; + } + else + return FNM_NOMATCH; + } + } + else +# undef str +#endif + { + c = FOLD (c); + normal_bracket: + + /* We have to handling the symbols differently in + ranges since then the collation sequence is + important. */ + is_range = *p == L('-') && p[1] != L('\0'); + + if (!is_range && c == fn) + goto matched; + + cold = c; + c = *p++; + } if (c == L('-') && *p != L(']')) { @@ -434,23 +608,19 @@ FCT (pattern, string, no_leading_period, flags) various characters appear in the source file. A strange concept, nowhere documented. */ - int32_t fseqidx; - int32_t lseqidx; + uint32_t fcollseq; + uint32_t lcollseq; UCHAR cend = *p++; # ifdef WIDE_CHAR_VERSION + int idx; size_t cnt; # endif - if (!(flags & FNM_NOESCAPE) && cend == L('\\')) - cend = *p++; - if (cend == L('\0')) - return FNM_NOMATCH; - # ifdef WIDE_CHAR_VERSION /* Search in the `names' array for the characters. */ - fseqidx = fn % size; + idx = fn % size; cnt = 0; - while (names[fseqidx] != fn) + while (names[idx] != fn) { if (++cnt == layers) /* XXX We don't know anything about @@ -458,63 +628,210 @@ FCT (pattern, string, no_leading_period, flags) match. This means we are failing. */ goto range_not_matched; - fseqidx += size; + idx += size; } - lseqidx = cold % size; - cnt = 0; - while (names[lseqidx] != cold) + fcollseq = collseq[idx]; + + if (is_seqval) + lcollseq = cold; + else { - if (++cnt == layers) + idx = cold % size; + cnt = 0; + while (names[idx] != cold) { - lseqidx = -1; - break; + if (++cnt == layers) + { + idx = -1; + break; + } + idx += size; } - lseqidx += size; + + lcollseq = idx == -1 ? 0xffffffff : collseq[idx]; } # else - fseqidx = fn; - lseqidx = cold; + fcollseq = collseq[fn]; + lcollseq = is_seqval ? cold : collseq[(UCHAR) cold]; +# endif + + is_seqval = 0; + if (cend == L('[') && *p == L('.')) + { + uint32_t nrules = + _NL_CURRENT_WORD (LC_COLLATE, + _NL_COLLATE_NRULES); + const CHAR *startp = p; + size_t c1 = 0; + + while (1) + { + c = *++p; + if (c == L('.') && p[1] == L(']')) + { + p += 2; + break; + } + if (c == '\0') + return FNM_NOMATCH; + ++c1; + } + + if (nrules == 0) + { + /* There are no names defined in the + collation data. Therefore we only + accept the trivial names consisting + of the character itself. */ + if (c1 != 1) + return FNM_NOMATCH; + + cend = startp[1]; + } + else + { + int32_t table_size; + const int32_t *symb_table; +# ifdef WIDE_CHAR_VERSION + char str[c1]; + int strcnt; +# else +# define str (startp + 1) # endif + const unsigned char *extra; + int32_t idx; + int32_t elem; + int32_t second; + int32_t hash; + +# ifdef WIDE_CHAR_VERSION + /* We have to convert the name to a single-byte + string. This is possible since the names + consist of ASCII characters and the internal + representation is UCS4. */ + for (strcnt = 0; strcnt < c1; ++strcnt) + str[strcnt] = startp[1 + strcnt]; +#endif + + table_size = + _NL_CURRENT_WORD (LC_COLLATE, + _NL_COLLATE_SYMB_HASH_SIZEMB); + symb_table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_TABLEMB); + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_EXTRAMB); + + /* Locate the character in the hashing + table. */ + hash = elem_hash (str, c1); + + idx = 0; + elem = hash % table_size; + second = hash % (table_size - 2); + while (symb_table[2 * elem] != 0) + { + /* First compare the hashing value. */ + if (symb_table[2 * elem] == hash + && (c1 + == extra[symb_table[2 * elem + 1]]) + && memcmp (str, + &extra[symb_table[2 * elem + 1] + + 1], c1) == 0) + { + /* Yep, this is the entry. */ + idx = symb_table[2 * elem + 1]; + idx += 1 + extra[idx]; + break; + } + + /* Next entry. */ + elem += second; + } + + if (symb_table[2 * elem] != 0) + { + /* Compare the byte sequence but only if + this is not part of a range. */ +# ifdef WIDE_CHAR_VERSION + int32_t *wextra; + + idx += 1 + extra[idx]; + /* Adjust for the alignment. */ + idx = (idx + 3) & ~4; + + wextra = (int32_t *) &extra[idx + 4]; +# endif + /* Get the collation sequence value. */ + is_seqval = 1; +# ifdef WIDE_CHAR_VERSION + cend = wextra[1 + wextra[idx]]; +# else + /* Adjust for the alignment. */ + idx += 1 + extra[idx]; + idx = (idx + 3) & ~4; + cend = *((int32_t *) &extra[idx]); +# endif + } + else if (symb_table[2 * elem] != 0 && c1 == 1) + { + cend = str[0]; + c = *p++; + } + else + return FNM_NOMATCH; + } +# undef str + } + else + { + if (!(flags & FNM_NOESCAPE) && cend == L('\\')) + cend = *p++; + if (cend == L('\0')) + return FNM_NOMATCH; + cend = FOLD (cend); + } /* XXX It is not entirely clear to me how to handle characters which are not mentioned in the collation specification. */ if ( # ifdef WIDE_CHAR_VERSION - lseqidx == -1 || + lcollseq == 0xffffffff || # endif - collseq[lseqidx] <= collseq[fseqidx]) + lcollseq <= fcollseq) { /* We have to look at the upper bound. */ - int32_t hseqidx; + uint32_t hcollseq; - cend = FOLD (cend); -# ifdef WIDE_CHAR_VERSION - hseqidx = cend % size; - cnt = 0; - while (names[hseqidx] != cend) + if (is_seqval) + hcollseq = cend; + else { - if (++cnt == layers) +# ifdef WIDE_CHAR_VERSION + idx = cend % size; + cnt = 0; + while (names[idx] != cend) { - /* Hum, no information about the upper - bound. The matching succeeds if the - lower bound is matched exactly. */ - if (lseqidx == -1 || cold != fn) - goto range_not_matched; - - goto matched; + if (++cnt == layers) + { + /* Hum, no information about the upper + bound. The matching succeeds if the + lower bound is matched exactly. */ + if (idx == -1 && lcollseq != fcollseq) + goto range_not_matched; + + goto matched; + } } - } + hcollseq = collseq[idx]; # else - hseqidx = cend; + hcollseq = collseq[cend]; # endif + } - if ( -# ifdef WIDE_CHAR_VERSION - (lseqidx == -1 - && collseq[fseqidx] == collseq[hseqidx]) || -# endif - collseq[fseqidx] <= collseq[hseqidx]) + if (lcollseq <= hcollseq && fcollseq <= hcollseq) goto matched; } # ifdef WIDE_CHAR_VERSION @@ -553,6 +870,7 @@ FCT (pattern, string, no_leading_period, flags) /* Skip the rest of the [...] that already matched. */ do { + ignore_next: c = *p++; if (c == L('\0')) @@ -568,12 +886,52 @@ FCT (pattern, string, no_leading_period, flags) } else if (c == L('[') && *p == L(':')) { - do - if (*++p == L('\0')) - return FNM_NOMATCH; - while (*p != L(':') || p[1] == L(']')); + int c1 = 0; + const CHAR *startp = p; + + while (1) + { + c = *++p; + if (++c1 == CHAR_CLASS_MAX_LENGTH) + return FNM_NOMATCH; + + if (*p == L(':') && p[1] == L(']')) + break; + + if (c < L('a') || c >= L('z')) + { + p = startp; + goto ignore_next; + } + } p += 2; - c = *p; + c = *p++; + } + else if (c == L('[') && *p == L('=')) + { + c = *++p; + if (c == L('\0')) + return FNM_NOMATCH; + c = *++p; + if (c != L('=') || p[1] != L(']')) + return FNM_NOMATCH; + p += 2; + c = *p++; + } + else if (c == L('[') && *p == L('.')) + { + ++p; + while (1) + { + c = *++p; + if (c == '\0') + return FNM_NOMATCH; + + if (*p == L('.') && p[1] == L(']')) + break; + } + p += 2; + c = *p++; } } while (c != L(']')); diff --git a/posix/regex.c b/posix/regex.c index eb5a6b3d47..0af5283505 100644 --- a/posix/regex.c +++ b/posix/regex.c @@ -2690,7 +2690,7 @@ regex_compile (pattern, size, syntax, bufp) PATFETCH (c); /* Now add the multibyte character(s) we found - to the acceptabed list. + to the accept list. XXX Note that this is not entirely correct. we would have to match multibyte sequences diff --git a/posix/tst-fnmatch.input b/posix/tst-fnmatch.input index 9c3ae1f167..7c79ddc3ab 100644 --- a/posix/tst-fnmatch.input +++ b/posix/tst-fnmatch.input @@ -70,23 +70,34 @@ C "]" "[!a]" 0 C "]]" "[!a]]" 0 # B.6 012(C) -# *** implement [. .] +C "a" "[[.a.]]" 0 +C "-" "[[.-.]]" 0 +C "-" "[[.-.][.].]]" 0 +C "-" "[[.].][.-.]]" 0 +C "-" "[[.-.][=u=]]" 0 +C "-" "[[.-.][:alpha:]]" 0 +C "a" "[![.a.]]" NOMATCH # B.6 013(C) -# *** implement [. .] +C "a" "[[.b.]]" NOMATCH +C "a" "[[.b.][.c.]]" NOMATCH +C "a" "[[.b.][=b=]]" NOMATCH -# B.6 014(C) -# *** implement [. .] # B.6 015(C) C "a" "[[=a=]]" 0 C "b" "[[=a=]b]" 0 C "b" "[[=a=][=b=]]" 0 +C "a" "[[=a=][=b=]]" 0 +C "a" "[[=a=][.b.]]" 0 +C "a" "[[=a=][:digit:]]" 0 # B.6 016(C) C "=" "[[=a=]b]" NOMATCH C "]" "[[=a=]b]" NOMATCH -C "a" "[[=b=]]" NOMATCH +C "a" "[[=b=][=c=]]" NOMATCH +C "a" "[[=b=][.].]]" NOMATCH +C "a" "[[=b=][:digit:]]" NOMATCH # B.6 017(C) C "a" "[[:alnum:]]" 0 @@ -225,6 +236,10 @@ C "a" "[[alpha]]" NOMATCH C "a" "[[alpha:]]" NOMATCH C "a]" "[[alpha]]" 0 C "a]" "[[alpha:]]" 0 +C "a" "[[:alpha:][.b.]]" 0 +C "a" "[[:alpha:][=b=]]" 0 +C "a" "[[:alpha:][:digit:]]" 0 +C "a" "[[:digit:][:alpha:]]" 0 # B.6 018(C) C "a" "[a-c]" 0 @@ -236,9 +251,28 @@ C "B" "[a-c]" NOMATCH C "b" "[A-C]" NOMATCH C "" "[a-c]" NOMATCH C "as" "[a-ca-z]" NOMATCH +C "a" "[[.a.]-c]" 0 +C "a" "[a-[.c.]]" 0 +C "a" "[[.a.]-[.c.]]" 0 +C "b" "[[.a.]-c]" 0 +C "b" "[a-[.c.]]" 0 +C "b" "[[.a.]-[.c.]]" 0 +C "c" "[[.a.]-c]" 0 +C "c" "[a-[.c.]]" 0 +C "c" "[[.a.]-[.c.]]" 0 +C "d" "[[.a.]-c]" NOMATCH +C "d" "[a-[.c.]]" NOMATCH +C "d" "[[.a.]-[.c.]]" NOMATCH # B.6 019(C) -C "b" "[c-a]" NOMATCH +C "a" "[c-a]" NOMATCH +C "a" "[[.c.]-a]" NOMATCH +C "a" "[c-[.a.]]" NOMATCH +C "a" "[[.c.]-[.a.]]" NOMATCH +C "c" "[c-a]" NOMATCH +C "c" "[[.c.]-a]" NOMATCH +C "c" "[c-[.a.]]" NOMATCH +C "c" "[[.c.]-[.a.]]" NOMATCH # B.6 020(C) C "a" "[a-c0-9]" 0 @@ -394,23 +428,34 @@ de_DE.ISO-8859-1 "a" "[[=a=]b]" 0 de_DE.ISO-8859-1 "â" "[[=a=]b]" 0 de_DE.ISO-8859-1 "à" "[[=a=]b]" 0 de_DE.ISO-8859-1 "á" "[[=a=]b]" 0 +de_DE.ISO-8859-1 "ä" "[[=a=]b]" 0 de_DE.ISO-8859-1 "b" "[[=a=]b]" 0 de_DE.ISO-8859-1 "c" "[[=a=]b]" NOMATCH de_DE.ISO-8859-1 "a" "[[=â=]b]" 0 de_DE.ISO-8859-1 "â" "[[=â=]b]" 0 de_DE.ISO-8859-1 "à" "[[=â=]b]" 0 de_DE.ISO-8859-1 "á" "[[=â=]b]" 0 +de_DE.ISO-8859-1 "ä" "[[=â=]b]" 0 de_DE.ISO-8859-1 "b" "[[=â=]b]" 0 de_DE.ISO-8859-1 "c" "[[=â=]b]" NOMATCH de_DE.ISO-8859-1 "a" "[[=à=]b]" 0 de_DE.ISO-8859-1 "â" "[[=à=]b]" 0 de_DE.ISO-8859-1 "à" "[[=à=]b]" 0 de_DE.ISO-8859-1 "á" "[[=à=]b]" 0 +de_DE.ISO-8859-1 "ä" "[[=à=]b]" 0 de_DE.ISO-8859-1 "b" "[[=à=]b]" 0 de_DE.ISO-8859-1 "c" "[[=à=]b]" NOMATCH de_DE.ISO-8859-1 "a" "[[=á=]b]" 0 de_DE.ISO-8859-1 "â" "[[=á=]b]" 0 de_DE.ISO-8859-1 "à" "[[=á=]b]" 0 de_DE.ISO-8859-1 "á" "[[=á=]b]" 0 +de_DE.ISO-8859-1 "ä" "[[=á=]b]" 0 de_DE.ISO-8859-1 "b" "[[=á=]b]" 0 de_DE.ISO-8859-1 "c" "[[=á=]b]" NOMATCH +de_DE.ISO-8859-1 "a" "[[=ä=]b]" 0 +de_DE.ISO-8859-1 "â" "[[=ä=]b]" 0 +de_DE.ISO-8859-1 "à" "[[=ä=]b]" 0 +de_DE.ISO-8859-1 "á" "[[=ä=]b]" 0 +de_DE.ISO-8859-1 "ä" "[[=ä=]b]" 0 +de_DE.ISO-8859-1 "b" "[[=ä=]b]" 0 +de_DE.ISO-8859-1 "c" "[[=ä=]b]" NOMATCH |