A bug in the -B option of gnudiff?

[Cory F. Cohen]

Hello,

It would appear that my interpretation of  "-B Ignore changes whose 
lines are all blank." is different than the one implemented in GNU 
diff.  I would have expected that the result would have been the same as 
removing all blank lines from both files and then diffing.

This isn't a big deal, but I thought you might want to know, in case you 
considered it a bug too.  The test case below should be pretty straight 
forward, but I included a couple of short hex dumps to make sure.  It 
looks like a context problem, but I see no reason why -B shouldn't work 
anyway...  Sorry I don't have a patch... :-)

$cat foo1
x

same string
y

same string
z

same string
done
$cat foo2
x
same string

y
same string

z
same string

done
$od -h foo1
0000000     780a    0a73    616d    6520    7374    7269    6e67    0a79
0000020     0a0a    7361    6d65    2073    7472    696e    670a    7a0a
0000040     0a73    616d    6520    7374    7269    6e67    0a64    6f6e
0000060     650a                                                        
0000062
$od -h foo2
0000000     780a    7361    6d65    2073    7472    696e    670a    0a79
0000020     0a73    616d    6520    7374    7269    6e67    0a0a    7a0a
0000040     7361    6d65    2073    7472    696e    670a    0a64    6f6e
0000060     650a                                                        
0000062
$gdiff -B foo1 foo2
4d2
< y
5a4
> y
7d5
< z
8a7
> z
$gdiff --version
diff (GNU diffutils) 2.8.1
Copyright (C) 2002 Free Software Foundation, Inc.

Thanks!
Cory