I get one tiger tree hash when I run bitcollider in linux (compiled from source), I get different results from when I run in windows.

For example, if I make an empty file, on linux I get:

bitprint=3I42H3S6NNFQ2MSVX7XZKYAYSCX5QBYJ.GKJ2YYYMCPYCIX4SXOYXM3QWCZ5E4WCJFXPHH4Y

on windows, I get

bitprint=3I42H3S6NNFQ2MSVX7XZKYAYSCX5QBYJ.LWPNACQDBZRYXW3VHJVCJ64QBZNGHOHHHZWCLNQ

For a file with a single zero byte, I get

linux: bitprint=LOUTZHNQZ74T6UVVEHLUEDSD63W2E6CP.LWPNACQDBZRYXW3VHJVCJ64QBZNGHOHHHZWCLNQ

windows: bitprint=LOUTZHNQZ74T6UVVEHLUEDSD63W2E6CP.VK54ZIEEVTWNAUI5D5RDFIL37LX2IQNSTAXFKSA

This seems to be a bug. What's the problem?

btw, the windows implementation agrees with the the information on http://www.open-content.net/specs/draft-jchapweske-thex-02.html.

But, my own independent implementation agrees with the linux version. In any case, it seems that all the bitprints you have from either windows or linux are invalid.

What's your linux build box like -- processor, distro, etc. ?

The 'LWP...' value from Windows/THEX draft is the correct value.

This could be a new variant of the problem discussed in this thread:

http://bitzi.com/bboard/message?message_id=19829&forum_id=4076

- Gordon @ Bitzi

Yes, its the linux one that appears to be wrong. (I was not prepending the zero-byte to my strings).

This was compiling bitcollider 0.4.0 on gentoo, pentium 4.