On 04/12/2013 10:42 PM, Guenter Knauf wrote: > On 10.04.2013 23:21, Guenter Knauf wrote: >> On 10.04.2013 23:01, fua...@apache.org wrote: >>> Author: fuankg >>> Date: Wed Apr 10 21:01:51 2013 >>> New Revision: 1466669 >>> >>> URL: http://svn.apache.org/r1466669 >>> Log: >>> Put this backport for now on hold to get some more >>> time for testing ... > >> ok, onward with some more testing ..... > now on r:regex() ... > > 1) the sample in the docs is completely wrong ... > 1.1) the docu has r:regex(string, pattern) but current code implements > r:regex(pattern, string); I will change this in the code soon since I it > looks more Lua-like for me what is in the docs since all other Lua match > functions also have parameters in the order (string, pattern) > 1.2) the pattern "foo (%w+) (%S*)" doesnt work for me; it looks more > like a pattern for string.match(); instead a valid pattern for r:regex() > would be f.e. "foo (\\w+) (\\a+)"; backslash must be escaped in a Lua > string > > 2) r:regex() crashes the server if more than AP_MAX_REG_MATCH matches > are found; this is because we still get the real number of matches in > regex.re_nsub although passing in AP_MAX_REG_MATCH: > rv = ap_regexec(®ex, source, AP_MAX_REG_MATCH, matches, 0); > > when I add: > > if (regex.re_nsub > AP_MAX_REG_MATCH) { > ap_log_rerror(APLOG_MARK, APLOG_INFO, 0, r, > "regex returned %d matches; allowed %d.", > regex.re_nsub, AP_MAX_REG_MATCH); > return 2; > } > > and test with a 20 match regex, I get this in the log: > regex returned 20 matches; allowed 10. > > The docu of regexec() gives no hints about how to retrieve the real > number of matches stored: > http://ci.apache.org/projects/httpd/trunk/doxygen/ap__regex_8h.html > and I've not yet digged into the code; but we need to make sure that the > following loop doesnt loop over AP_MAX_REG_MATCH: > for (i = 0; i <= regex.re_nsub; i++) { > > BTW. what is the reason for limiting AP_MAX_REG_MATCH to 10 in httpd.h ? > > Gün. > > I think the reason for limiting it to 10 is legacy stuff, so that's what I've complied with.
Docs fixed, ordering of arguments likewise, and r.banner as well as r.port are now just values, not functions. With regards, Daniel.
