[Nix-dev] Re: Accidental store paths collisions

[Eelco Dolstra]

Lluís Batlle wrote:

>> Peter Simons <simons at cryp.to> writes:
>>
>>>  > I think the probability of an accidental collision between any two
>>>  > store paths (regardless of the ‘stdenv.system’ value) is that of an
>>>  > accidental collision between any two SHA256 hashes of two different
>>>  > inputs.
>> [...]
>>
>>> The question is: do the inputs differ when the same expression is built
>>> on two different systems, i.e. MacOS X and Linux?
>> Yes, at least one input differs, namely ‘stdenv.system’.
> I didn't recall the 'system' attribute is passed to any stdenv
> derivation through stdenv.mkDerivation. Also for builderDefs?

Yes, since it's one of only a few required attributes of the derivation primop
(along with "name" and "builder").

-- 
Eelco Dolstra | http://www.st.ewi.tudelft.nl/~dolstra/