[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Nmh-workers] fgets bug in mhbuildsbr.c

From: David E. West
Subject: Re: [Nmh-workers] fgets bug in mhbuildsbr.c
Date: Thu, 12 May 2005 09:55:44 -0400


I messed up when  attempting to attach the test ZIP file  in my previous email.
I accidently used  the unpatched version of mhbuild  when composing the message
and I didn't inspect it before sending.   Also it has come to my attention that
it is common practice for spam  blockers to block attached ZIP files anyway, so
below is the ZIP file, in text, base64 encoded.


There is  a command line  utility called "base64"  that can be used  to convert
this back into binary.  For example, concatenate the above into one long string
and run the following:

$ echo -n "UEsDBBQACQ . . . awAAAAAA" | base64 -d > foo.zip

The md5 of the resulting binary should be "7a093565d1474332174bd6e884429d51". 

Also any  8-bit binary string  that is  less than 256  bytes long and  has only
7-bit characters  before the  first occupance of  a 0x00 character  would work.
And the string could  be longer than 256 bytes.  This pattern  would have to be
repeated for  each block of 255  bytes in the file.   (This is the  size of the
temporary buffer into which the fgets  call stores file data before the 8/7 bit
determination has been made.)  The trick to this bug is that the probability of
it occurring  approaches 0,  exponentially, based upon  the number  of 255-byte
blocks  there are in  the file.   When your  entire file  fits within  a single
block, then  the probability  is much higher,  because the detection  code only
gets one shot  at it.  But for  large binary files, it is  highly unlikely that
that a false negative would occur.

David E. West

reply via email to

[Prev in Thread] Current Thread [Next in Thread]