Re: generation numbers
- From
Junio C Hamano <gitster@pobox.com>
- Date
- Jul 7, 2011, 19:34 UTC
- Message-ID
- <7vliw9hoky.fsf@alter.siamese.dyndns.org>
- In-Reply-To
- <20110707185908.GB12044@sigill.intra.peff.net>
Jeff King <peff@peff.net> writes:
> You could "cheat" and instead of storing the sha1 of a blob object in > the notes tree, use the lower 32 bits to store an actual value. I don't > think that currently breaks any assumptions in the notes code, but it > definitely is against the intent of it.
I highly suspect that it would break fsck rather badly. You may not even be able to repack a repository with such a notes tree.
> For a local lookup cache, I would use a fixed-size binary integer just > to keep the lookup data structure simple (then you know the width of > each record ahead of time). For a generation commit header, obviously we > would go with the ascii representation as we do for other headers.
Yes.