Update: Hmm. Well then, try (?:.|\n) or [\000-\377]. Maybe?

_____________________________________________________
Jeff[japhy]Pinyan: Perl, regex, and perl hacker.
s++=END;++y(;-P)}y js++=;shajsj<++y(p-q)}?print:??;

While I wouldn't advocate moving back to 5.6.0, I don't think \C avoids this particular bug...

% perl -le 'print "($1) " while "this_is_broken_" =~ /(\C*?)_/sg'
(this)

% perl -le 'print "($1) " while "this_is_broken_" =~ /(\C*?)_/g'
(this) 

% perl -le 'print "($1) " while "this_is_broken_" =~ /(.*?)_/sg'
(this) 

% perl -mre=debug -le 'print "($1) " while "this_is_broken_" =~ /(\C*?
+)_/sg'
Freeing REx: `,'
Compiling REx `(\C*?)_'
size 10 first at 5
   1: OPEN1(3)
   3:   MINMOD(4)
   4:   STAR(6)
   5:     SANY(0)
   6: CLOSE1(8)
   8: EXACT <_>(10)
  10: END(0)
floating `_' at 0..2147483647 (checking floating) anchored(SBOL) impli
+cit minlen 1 
Guessing start of match, REx `(\C*?)_' against `this_is_broken_'...
Found floating substr `_' at offset 4...
Guessed: match at offset 0
Matching REx `(\C*?)_' against `this_is_broken_'
  Setting an EVAL scope, savestack=9
   0 <> <this_is_brok>    |  1:  OPEN1
   0 <> <this_is_brok>    |  3:  MINMOD
   0 <> <this_is_brok>    |  4:  STAR
  Setting an EVAL scope, savestack=9
   0 <> <this_is_brok>    |  6:    CLOSE1
   0 <> <this_is_brok>    |  8:    EXACT <_>
                              failed...
                           SANY can match 1 times out of 1...
   1 <t> <his_is_brok>    |  6:    CLOSE1
   1 <t> <his_is_brok>    |  8:    EXACT <_>
                              failed...
                           SANY can match 1 times out of 1...
   2 <th> <is_is_brok>    |  6:    CLOSE1
   2 <th> <is_is_brok>    |  8:    EXACT <_>
                              failed...
                           SANY can match 1 times out of 1...
   3 <thi> <s_is_brok>    |  6:    CLOSE1
   3 <thi> <s_is_brok>    |  8:    EXACT <_>
                              failed...
                           SANY can match 1 times out of 1...
   4 <this> <_is_brok>    |  6:    CLOSE1
   4 <this> <_is_brok>    |  8:    EXACT <_>
   5 <this_> <is_brok>    | 10:    END
Match successful!
(this) 
Guessing start of match, REx `(\C*?)_' against `is_broken_'...
Not at start...
Match rejected by optimizer
Freeing REx: `(\C*?)_'
[download]

Let me know if you'd like more debugging info (its from the Cobalt setup I mentioned earlier)

UPDATE:

However, using /.{0,}?/s instead of /.*?/s seems to fix it for me...

%  perl -le 'print "($1) " while "this_is_broken_" =~ /(.*?)_/sg'
(this) 

%  perl -le 'print "($1) " while "this_is_broken_" =~ /(.{0,}?)_/sg'
(this) 
(is) 
(broken)
[download]

UPDATE 2

Both of japhy's updated suggestions seem to work for me... with and w/o the /s modifier...

(Broken Base Case...)
% perl -le 'print "($1) " while "this_is_broken_" =~ /(.*?)_/sg'
(this) 

% perl -le 'print "($1) " while "this_is_broken_" =~ /([\000-\377]*?)_
+/sg'
(this) 
(is) 
(broken) 

% perl -le 'print "($1) " while "this_is_broken_" =~ /([\000-\377]*?)_
+/g'
(this) 
(is) 
(broken) 

% perl -le 'print "($1) " while "this_is_broken_" =~ /((?:.|\n)*?)_/g'
(this) 
(is) 
(broken) 

% perl -le 'print "($1) " while "this_is_broken_" =~ /((?:.|\n)*?)_/sg
+'
(this) 
(is) 
(broken)
[download]

-Blake