LG violates gpl? take 2

[Luca Olivetti]

En/na Joseph Heenan ha escrit:
> On 3/25/2010 8:11 AM, Luca Olivetti wrote:
> 
>> But in this case it says that the same sources, with the same build
>> system and the same configuration, produce the same binary kernel (minus
>> the timestamp).
>> And if it isn't so, it's not the product of the same, exact, source.
>> I.e. LG didn't provide me the same os that's running on the tv.
> 
> Whilst there is a good chance you're right about that, I wouldn't forget 
> that (iirc) the compiled binary will also differ depending on:
> 
> - the location of the source tree (affects __FILE__)
> 
> - the username & hostname used for compilation (as this is included in 
> the initial banner)
> 
> - compiler/libraries/options (though I think you already had that side 
> of things covered)

Those are good points, so I did the following:

1) I put the sources in the same location as the original kernel 
(obtained from "strings image_from_tv | grep /home")

2) I changed the hostname *and* the date so that the banner have the 
same length:

$ strings image_from_tv.vmlinuz | grep 'swfarm'
%s version %s (charism at swfarm-l7) (gcc version 3.4.3 (MontaVista 
3.4.3-25.0.70.0501961 2005-12-18)) %s
Linux version 2.6.26 (charism at swfarm-l7) (gcc version 3.4.3 (MontaVista 
3.4.3-25.0.70.0501961 2005-12-18)) #108 PREEMPT Fri Sep 4 11:15:36 KST 2009

$ strings vmlinuz | grep 'localhost'
%s version %s (luca at localhostxxx) (gcc version 3.4.3 (MontaVista 
3.4.3-25.0.70.0501961 2005-12-18)) %s
Linux version 2.6.26 (luca at localhostxxx) (gcc version 3.4.3 (MontaVista 
3.4.3-25.0.70.0501961 2005-12-18)) #118 PREEMPT Thu Mar 4 11:42:14 CET 2010

The result of the comparison (both with cmp and diffing the hexdump) 
almost didn't change.
As a reference, I compared the previous vmlinuz (the one in the "wrong" 
tree) and the new one and the differences, while still there, are much less.

> There's also perhaps one or two more subtle effects that could produce 
> changes, like if the release kernel was taken from a tree that several 
> builds had already been done in it rather than being completely clean 
> before the build was done.

Bummer, so there's no way to check that an embedded device integrator is 
really complying.

Bye
-- 
Luca