Question

I have found a case where e.g. 500 letters/digits with whitespace makes PCRE blow up when using \w in the regex. I have tried boh TPerlRegEx and Delphi XE2 pcre/obj implementations. Same for both.

I then tried calling

pcre_config(PCRE_CONFIG_MATCH_LIMIT, @vSysStrRegex_MatchLimit_Value); 

But setting match limits does not seem to have any effect. The way I am using it is that I call it once for each thread. (Note: Others have set this to get not set such settings)

I really need the regex library to drop out of parsing instead of continuing until it overflows the stack. (It seems next to impossible to recover the thread/program from that.)

How do I prevent stack overflows in this situation? I can’t control the content parsed or the regular expressions. Thus I am specificly looking for ways to avoid PCRE running into stack overflows through a setting or similar.

Solution by editing TPerlRegEx code:

function TPerlRegEx.Match(AStartOffset: Integer = 0): Boolean;
...
if FHints <> nil then // set by "study" call
  begin
    PPCREExtra(FHints)^.flags := PPCREExtra(FHints)^.flags or PCRE_EXTRA_MATCH_LIMIT_RECURSION;
    PPCREExtra(FHints)^.match_limit_recursion := 750 // 1000 too much in tests
  end
;
OffsetCount := pcre_exec(FPattern, FHints, ...)

Answer 1

You’ve cited some PCRE documentation that describes setting the recursion limit at compile time with the --with-match-limit-recursion configuration option. You can exercise that option if you build the PCRE library yourself. If you read elsewhere in that same document, you’ll find a description of the match_limit_recursion field of the pcre_extra block passed to pcre_exec:

The default value for match_limit_recursion can be set when PCRE is built; the default default is the same value as the default for match_limit. You can override the default by suppling pcre_exec() with a pcre_extra block in which match_limit_recursion is set, and PCRE_EXTRA_MATCH_LIMIT_RECURSION is set in the flags field. If the limit is exceeded, pcre_exec() returns PCRE_ERROR_RECURSIONLIMIT.

So, set the recursion limit to something lower than the default. The default is evidently even higher than your actual available stack space; if it were lower, then you’d already be getting the PCRE_ERROR_RECURSIONLIMIT error instead of an OS-raised stack-overflow exception.

How the Delphi-specific wrappers represent this setting, I have no idea.