Re: [cobalt-developers] diff usage

[Tim Hockin <thockin@xxxxxxx>]

Matthew Nuzum wrote:
> 
> Can anyone point to clear instructions on how to use diff, especially in
> understanding and using the output?

The only form of diff I use is -u.  This is called a unified diff.  Here is
an example of it's output, explained.  For this I used the -ruN options
(recursive, unified, new-files as diffs).  Sometimes the -b option is
useful to ignore blankspace diffs.


> diff -ruN dist-2.4.12+patches/Documentation/Configure.help cvs-2.4.12+patches/Documentation/Configure.help

the command line - when diffing multiple files, diff puts this in.

> --- dist-2.4.12+patches/Documentation/Configure.help	Mon Oct 15 10:20:16 2001
> +++ cvs-2.4.12+patches/Documentation/Configure.help	Mon Oct 15 10:34:55 2001

The --- line shows the name of the old file.
The +++ line shows the name of the new file.
All lines that start with a - have been removed.  All lines that start with
a + have been added.  In non-unified diffs, sometimes you'll see other
line-markers.

> @@ -14952,9 +15177,10 @@

This is the hunk-header.  It says that the hunk starts at line 14952 and
lasts 9 lines in the original file.  It also says that the new hunk starts
at line 15177 in teh new file, and contains 10 lines.  A net gain of one
line.

>  CONFIG_NVRAM
>   If you say Y here and create a character special file /dev/nvram
>   with major number 10 and minor number 144 using mknod ("man mknod"),

This is context.  By default diff uses 3 lines of context (a cmdline opt
can change that).  This helps patch find where to put a diff when teh line
numbers don't match exactly.  Always three lines at the top of a hunk and
three lines at the bottom.  Thus, no hunk is less than 7 lines (6 context,
at least one change).

> -  you get read and write access to the 50 bytes of non-volatile memory
> +  you get read and write access to the extra bytes of non-volatile memory

This shows that the line changed, the old form is removed and the new form
is inserted at the same place

>    in the real time clock (RTC), which is contained in every PC and

an unchanged line, but less than 3 lines, so no need for a new hunk.

> -  most Ataris. 
> +  most Ataris.  The actual number of bytes varies, depending on the 
> +  nvram in the system, but is usually 114.  

This shows that one line was removed and two inserted.  Here is our net
gain of one line.

>  
>   This memory is conventionally called "CMOS RAM" on PCs and "NVRAM"
>   on Ataris. /dev/nvram may be used to view settings there, or to

3 lines of context at the end.  So we have 3 context, 1 change, 1
unchanged, 1 change, 1 new, and 3 context = 10 lines.

> diff -ruN dist-2.4.12+patches/drivers/char/nvram.c cvs-2.4.12+patches/drivers/char/nvram.c
> --- dist-2.4.12+patches/drivers/char/nvram.c	Mon Oct 15 10:21:41 2001
> +++ cvs-2.4.12+patches/drivers/char/nvram.c	Mon Oct 15 10:21:41 2001

Another file.

> @@ -10,8 +10,9 @@
>  * "NVRAM" (NV stands for non-volatile).
>  *
>  * The data are supplied as a (seekable) character device, /dev/nvram. The
> - * size of this file is 50, the number of freely available bytes in the memory
> - * (i.e., not used by the RTC itself).
> + * size of this file is dependant on the controller.  The usual size is 50, 
> + * the number of freely available bytes in the memory (i.e., not used by the 
> + * RTC itself).
>  * 
>  * Checksums over the NVRAM contents are managed by this driver. In case of a
>  * bad checksum, reads and writes return -EIO. The checksum can be initialized

I'll leave the interpretation to you.  Any file can have multiple hunks,
each starting with @@


Clear now?  Maybe a bit :)
-- 
Tim Hockin
Systems Software Engineer
Sun Microsystems, Cobalt Server Appliances
thockin@xxxxxxx