[Bug classpath/24876] Regex failure

[kaz at maczuka dot gcd dot org]

------- Comment #5 from kaz at maczuka dot gcd dot org  2006-01-22 03:52 -------
Created an attachment (id=10699)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10699&action=view)
Another testcase

This test shows the following.

GNU Classpath (with Kaffe):
$ java W 'address@hidden' '(.+?)!((.+?)@(.+?))'
false
$ java -DFIND=1 W 'address@hidden' '(.+?)!((.+?)@(.+?))' 
true
G0 = address@hidden
G1 = rabbit78_
G2 = address@hidden
G3 = n=roman
G4 = 2

Sun's JDK:
$ java W 'address@hidden' '(.+?)!((.+?)@(.+?))'
true
G0 = address@hidden
G1 = rabbit78_
G2 = address@hidden
G3 = n=roman
G4 = 212.126.219.82
$ java -DFIND=1 W 'address@hidden' '(.+?)!((.+?)@(.+?))'
true
G0 = address@hidden
G1 = rabbit78_
G2 = address@hidden
G3 = n=roman
G4 = 2

Other languages:
Perl:
$ perl -e 'print (("address@hidden" =~ /(.+?)!((.+?)@(.+?))/
? $4 : false)."\n")'
false

Ruby:
$ ruby -e 'print (("address@hidden" =~ /(.+?)!((.+?)@(.+?))/
? $4 : false) + "\n")'
2

The behavior of GNU Classpath resembles Ruby and I think that is the
correct behaviour in the light of the meaning of "+?".

Sun's find() correctly finds "address@hidden" but Sun's matches()
matches "address@hidden" more greedily than expected
for "+?".

I am not sure whether this greedy match of Sun's is correct.

Programmers are not encouraged to use such ambiguous expression
whose meaning depends on the implementation.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24876