Mercurial > vim
changeset 5255:3c6e2b89875f v7.4b.004
updated for version 7.4b.004
Problem: Regexp crash on pattern "@\%[\w\-]*". (Axel Kielhorn)
Solution: Add \%(\) around \%[] internally.
| author | Bram Moolenaar <bram@vim.org> |
|---|---|
| date | Thu, 01 Aug 2013 18:38:26 +0200 |
| parents | d17b182a73a4 |
| children | 74301ef4297a |
| files | src/regexp_nfa.c src/testdir/test64.in src/testdir/test64.ok src/version.c |
| diffstat | 4 files changed, 30 insertions(+), 4 deletions(-) [+] |
line wrap: on
line diff
--- a/src/regexp_nfa.c +++ b/src/regexp_nfa.c @@ -1166,6 +1166,15 @@ nfa_regatom() reg_magic == MAGIC_ALL); EMIT(NFA_OPT_CHARS); EMIT(n); + + /* Emit as "\%(\%[abc]\)" to be able to handle + * "\%[abc]*" which would cause the empty string to be + * matched an unlimited number of times. NFA_NOPEN is + * added only once at a position, while NFA_SPLIT is + * added multiple times. This is more efficient than + * not allowsing NFA_SPLIT multiple times, it is used + * a lot. */ + EMIT(NFA_NOPEN); break; } @@ -1641,7 +1650,7 @@ nfa_regpiece() * engine interprets the plus as "try matching one more time", and * a* matches a second time at the end of the input, the empty * string. - * The submatch will the empty string. + * The submatch will be the empty string. * * In order to be consistent with the old engine, we replace * <atom>+ with <atom><atom>* @@ -2242,13 +2251,13 @@ nfa_postfix_dump(expr, retval) else if (retval == OK) fprintf(f, ">>> NFA engine succeeded !\n"); fprintf(f, "Regexp: \"%s\"\nPostfix notation (char): \"", expr); - for (p = post_start; *p && p < post_end; p++) + for (p = post_start; *p && p < post_ptr; p++) { nfa_set_code(*p); fprintf(f, "%s, ", code); } fprintf(f, "\"\nPostfix notation (int): "); - for (p = post_start; *p && p < post_end; p++) + for (p = post_start; *p && p < post_ptr; p++) fprintf(f, "%d ", *p); fprintf(f, "\n\n"); fclose(f); @@ -3005,7 +3014,18 @@ post2nfa(postfix, end, nfa_calc_size) { int n; - /* \%[abc] */ + /* \%[abc] implemented as: + * NFA_SPLIT + * +-CHAR(a) + * | +-NFA_SPLIT + * | +-CHAR(b) + * | | +-NFA_SPLIT + * | | +-CHAR(c) + * | | | +-next + * | | +- next + * | +- next + * +- next + */ n = *++p; /* get number of characters */ if (nfa_calc_size == TRUE) {
--- a/src/testdir/test64.in +++ b/src/testdir/test64.in @@ -373,6 +373,7 @@ STARTTEST :call add(tl, [2, '\%[bar]x', 'xxx', 'x']) :call add(tl, [2, 'b\%[[ao]r]', 'bar bor', 'bar']) :call add(tl, [2, 'b\%[[]]r]', 'b]r bor', 'b]r']) +:call add(tl, [2, '@\%[\w\-]*', '<http://john.net/pandoc/>[@pandoc]', '@pandoc']) :" :"""" Alternatives, must use first longest match :call add(tl, [2, 'goo\|go', 'google', 'goo'])