inventory sha1 is not checked during routine operation

[Robert Collins]

We check that the sha1 of a stored inventory is what the knit thinks it
should be, but AFAICT we don't check that the revision object expected
that sha1 ;).

So we're safe against bit-errors in the inventory store, but not against
deliberate manipulation.

 affects bzr
--
GPG key available at: <http://www.robertcollins.net/keys.txt>.

[John A Meinel (jameinel)]

Is it enough to check the revision text against the stored inventory sha? (Rather than re-hashing the full-text at that time?)

[Jelmer Vernooij (jelmer)]

is this still an issue with chk repositories?

[Martin Pool (mbp)]

On 2 February 2011 03:09, Jelmer Vernooij <email address hidden> wrote:
> is this still an issue with chk repositories?

I think this is now basically safe. Perhaps we should make sure that
any attempt to get an inventory by revision-id always first gets the
relevant revision, then from that gets the inventory base SHA1. I
don't think that is true at the moment. Are other changes needed?