On Thu, 24 Jul 2008 12:22:56 -0700 (PDT)
Linus Torvalds <torvalds@linux-foundation.org> wrote:

quoted text
>=20
> Pulling doesn't change the SHA1 ID of a commit, so _if_ it had been stabl=
e=20
quoted text
> in the -mmc tree, it would have had the same commit in the kernel too.=20
> That's very much a design issue in git: the same commit always has the=20
> same ID. _ALWAYS_.
>=20
> So the fact that the ID changed implies that something odd happened in th=
e=20
quoted text
> -mmc tree. Most likely a rebase. And again, this implies that somebody=20
> rebased already-public commits, since otherwise nobody would have ever=20
> even seen the original SHA1 ID.=20

Indeed I did. Things got shuffled around a bit when I updated the
patch queue with more stuff and the old attempt hadn't gotten merged
yet.

quoted text
>=20
> Which is not good behaviour, _exactly_ because it makes it much harder fo=
r=20
quoted text
> people to work together - suddenly a commit that got reported (for=20
> whatever reason: bug-report or 'please backport', it's all the same issue=
=20
quoted text
> in the end) no longer exists.

I've always considered my "for-linus" branch to be primarily that, but I
guess I can be more careful so that other people can use it as well.

Rgds
--=20
     -- Pierre Ossman

  WARNING: This correspondence is being monitored by the
  Swedish government. Make sure your server uses encryption
  for SMTP traffic and consider using PGP for end-to-end
  encryption.