Re: bug with diff 2.8.1

[jdeifik]

All the generated files are output from dd.

The first file (in hdd1 directory) is the boot partition of a hard drive.
The second file (in hdd2 directory) is a @1gb partition of a hard drive.
The third file (in hdd3 directory) is another @1gb partition of a hard drive.
The rest of the files (in sda1 directory) are 2gb pieces (via dd) of a 50gb 
hard drive.
My original bug report was incorrect with respect to the size of the pieces.
They are 2gb each, not 1gb each.

The files being backed up are of a relatively standard distribution of linux.
I am 99% sure that they are all binary (being output of dd on a hard drive 
partition),
and there are no text files.
I can provide the output of 'file' on each file if desired.

I manually ran cmp -l on each file in each tree, and all of the files are 
identical.
diff -r worked on one or two files in the sda1 directory before producing the
out of memory error.

I provided the directory tree, with the file names, dates, and sizes.
I would hope any source of binary data should do to create the actual files.

I still have the tree in question, and can run pretty any much tests on it.

Thanks,
jdeifik



---- Paul Eggert <address@hidden> wrote: 
> <address@hidden> writes:
> 
> > Though it is a bit ugly, perhaps you should replicate the directory tree
> > that caused the problem for me.
> 
> I cannot replicate the directory tree with just the information you've
> supplied so far.
> 
> I suspect that your original hypothesis is incorrect, and that the
> original input does not consist of 1 GB chunks of identical binary
> files.  If some of the files are not binary, that might well explain
> the problem: when the files are not binary 'diff' needs to read them
> all into memory to check them.
> 
> Perhaps you can verify this hypothesis on your data.