See perlre

m/(?:(?!somephrase).)*/;
[download]

"A zero-width negative look-ahead assertion. For example /foo(?!bar)/ matches any occurrence of "foo" that isn't followed by "bar". Note however that look-ahead and look-behind are NOT the same thing. You cannot use this for look-behind."

Maybe I'm missing something subtle here, but the answers so far seem too complicated.

If you want to stop at the phrase 'stop', just embed that at the appropriate point in the regex and use an non-greedy 'anything' match:

$s = "this is a load of junk to be consumed until we get to the word s
+top. 
After that nothing should stop it until another stop";;

print $1 while $s =~ m[(.+?)stop]g;;

this is a load of junk to be consumed until we get to the word
. After that nothing should
 it until another
[download]

And if you want the stop phrase to be a part of (say) the next capture, only then do you need to wrap it in a ZLA:

print $1 while $s =~ m[(.+?)(?=stop)]g;;

this is a load of junk to be consumed until we get to the word
stop. After that nothing should
stop it until another
[download]

That's fine so long as the regex ends at that point. Backtracking can easily cause problems if that pattern is part of some larger regex (or later becomes part of some larger regex).

For example, if I want to find only sections of "start...stop" that actually end in "stop now" then non-greedy doesn't try hard enough to not be greedy:

$_= "start this stop start that stop now start last stop";
print "non-greedy:\n";
print "  ", $_, $/
    for  /start(.*?)stop now/g;
print "look-ahead:\n";
print "  ", $_, $/
    for  /start((?:(?!stop).)*)stop now/g;
__END__
non-greedy:
   this stop start that
look-ahead:
   that
[download]

- tye        

...regex /([^:]+)/ which tells perl to eat everything until it reaches the ":".

...unless the input begins with a colon, when the regex will start its eating i. e. matching after the leading colon(s).

So the "exact" equivalent to /([^:]+)/ would be /((?:.(?!stop))+.)/, while the "eat up until" case might be something like /\G((?!stop)(?:.(?!stop))+.)/, corresponding /([^:]*)/, or so I hope.

No, it doesn't work. To work, it not only needs to match when it should match, it needs to not mach when it shouldn't match. Yours fails to do the latter.

/([^(stop)]+)/ is the same as /([^()opst]+)/. It doesn't look for "stop" at all, just individual characters.

You basically took the regexp the OP said didn't work, and made it not work in more places. For example, your solution stops wherever a paren is encountered.

$_ = 'nothing can stop the bus on the road';
print +(split /stop/)[0], "\n";
print +(split /bus/)[0], "\n";

# output (with trailing space)
nothing can
nothing can stop the
[download]

It's actually faster, but it doens't work. See Re^2: Tell regex to stop at "phrase", instead of char - how?.