Vlado Boza added the comment:
One possible fix:
Look for StringHasher in google v8 code
(http://code.google.com/p/v8/source/search?q=stringhasher&origq=stringhasher&btnG=Search+Trunk).
Main loop looks like this:
raw_running_ha
Vlado Boza added the comment:
>I tried this script on Linux 32 bits and Linux 64 bits: I didn't see any
>>collision. What is your operating system and the version of your >operating
>system please?
uname -a
Linux 3.0.0-17-generic #30-Ubuntu SMP Thu Mar 8 20:45:39 UT
Vlado Boza added the comment:
For example take this script (on 32bit):
ha = hash("\x00\xcf\x0b\x96\x19")
hb = hash("\x00\x6d\x29\x45\x18")
if ha == hb:
print "collision"
And run following:
for i in `seq 0 25`; do echo $i; for j in `seq 0 100`; do ./python -
Vlado Boza added the comment:
My bad (I checked only function in C++, not result in python).
This should work on 32bit:
Prefix: anything ending on 0x00
Suffix: anything
Strings: "\x00\xcf\x0b\x96\x19", "\x00\x6d\x29\x45\x18"
--
__
Vlado Boza added the comment:
E.g this strings collide for every prefix ending on 0xcd:
0x27fd5a18, 0x26fe78fa
--
___
Python tracker
<http://bugs.python.org/issue14
New submission from Vlado Boza :
Fix of this http://bugs.python.org/issue13703 is broken.
tl;dr: There only 256 different hash functions (compare it to size of
_Py_HashSecret prefix and suffix). And whether keys collide or not depends only
on the last 8 bits of prefix.
Problem with current
