Test case failure; with (very rough) proposed fix

[spoonguard]

The following test case seems to fail -- specifically, the invocation of has_trailing? seems to incorrectly report false:

s = Mongrel::BMHSearch.new('--X', 512)

s.find('--X') # No trailing

s.find('B--') # Trailing; works correctly

s.find('X--') # Trailing

raise 'fail' unless (s.has_trailing?) # Fails

The correctness of s.has_trailing? makes buffer management in the caller much simpler and/or more efficient in some cases.

Commenting out the following line in ext/bmh_search/bmh_search.c (around line 199) and ending brace:

// don't bother if the string ends in the needle completely

if(S->nfound == 0 || last != hlen - S->nlen) {

...

}

...seems to fix the problem for this test case, and all of the ones in the application I'm currently working on. I haven't had a chance to run the fuzzers/tests that come up on pastie.org yet, but I plan to.

Dave
(who is writing stream-based multipart mime handling code, and happened to run in to this...)