Re: Re: this causes a segfault cvs: php-src(PHP_4_3)

Jani Taskinen wrote:

    Fixed in CVS now. :)

nice try:) - something as simple as this still kills it.. <? echo preg_match('/abc/','fredabc',$matches); ?> Regards Alan

        --Jani
    

On Thu, 10 Jul 2003, Alan Knowles wrote:

@@ -420,7 +421,32 @@ size_offsets = num_subpats * 3; offsets = (int *)safe_emalloc(size_offsets, sizeof(int), 0); - /* Allocate match sets array and initialize the values */ + /* + * Build a mapping from subpattern numbers to their names. We will always + * allocate the table, even though they may be no named subpatterns. This + * avoids somewhat more complicated logic in the inner loops. + */ + subpat_names = (char **)safe_emalloc(num_subpats, sizeof(char *), 0); + memset(subpat_names, 0, sizeof(char *) * num_subpats); + { + int name_cnt, name_size, ni = 0; + char *name_table; + unsigned short name_idx; + + pcre_fullinfo(re, extra, PCRE_INFO_NAMECOUNT, &name_cnt); + if (name_cnt > 0) { + pcre_fullinfo(re, extra, PCRE_INFO_NAMETABLE, &name_table); + pcre_fullinfo(re, extra, PCRE_INFO_NAMEENTRYSIZE, &name_size); + + while (ni++ < name_cnt) { + name_idx = 0xff * name_table[0] + name_table[1]; + subpat_names[name_idx] = name_table + 2;

I'm getting name_idx = 23078, sizeof(subpat_names) = 4 here. backing out the patch solves the problem (sorry I've no idea what the code causing the problem is - it's part of a huge app...) Regards Alan

+ name_table += name_size; + } + } + } + + /* Allocate match sets array and initialize the values. */ if (global && subpats_order == PREG_PATTERN_ORDER) { match_sets = (zval **)safe_emalloc(num_subpats, sizeof(zval *), 0); for (i=0; i<num_subpats; i++) {