linuxppc trees, what is going on ?

[Tom Rini]

On Mon, Jan 19, 2004 at 06:07:41PM +0100, Sven Luther wrote:
> On Mon, Jan 19, 2004 at 09:48:50AM -0700, Tom Rini wrote:
> >
> > On Mon, Jan 19, 2004 at 05:38:57PM +0100, Sven Luther wrote:
> > > On Mon, Jan 19, 2004 at 08:58:01AM -0700, Tom Rini wrote:
> > > > On Mon, Jan 19, 2004 at 04:47:58PM +0100, Sven Luther wrote:
> > > >
> > > > > On Mon, Jan 19, 2004 at 08:29:54AM -0700, Tom Rini wrote:
> > > > > > > I did pull the linux-2.4 tree, and exported the v2.4.24 tag. I did the
> > > > > > > same for the linuxppc-2.4 tree, and diffed the two exported trees. Both
> > > > > > > were exactly the same.
> > > > > >
> > > > > > That is correct. A tag has the same values (set of ChangeSets)
> > > > > > regardless of the tree (simpilication).
> > > > >
> > > > > Ok, how are you then supposed to make a diff of the powerpc relative
> > > > > changes ? You use the changeset just above the merge or something ?
> > > >
> > > > The process I described about (that you snipped) is how you find the
> > > > ChangeSet of when 'linux-2.4' at a certain point is merged into
> > > > 'linuxppc-2.4', and that ChangeSet will point to something like 2.4.24 +
> > > > powerpc changes.
> > >
> > > Yeah, but i am not sure i fully understand. In this case, bk changes in
> > > linuxppc-2.4 yields :
> > >
> > > ChangeSet at 1.1058.1.312, 2004-01-04 19:34:56-02:00, marcelo at logos.cnet
> > >   Import 2.4.24 final tree
> > >
> > > ChangeSet at 1.1058.1.311, 2004-01-04 19:32:02-02:00, marcelo at logos.cnet
> > >   Cset exclude: laforge at netfilter.org|ChangeSet|20031204183256|31723
> > >
> > > ChangeSet at 1.1058.1.310, 2004-01-04 19:31:23-02:00, marcelo at logos.cnet
> > >   Cset exclude: jt at bougret.hpl.hp.com|ChangeSet|20031213132008|01226
> > >
> > > ChangeSet at 1.1058.1.309, 2004-01-04 19:31:05-02:00, marcelo at logos.cnet
> > >   Cset exclude: trini at mvista.com|ChangeSet|20031210203050|36304
> > >
> > > ChangeSet at 1.1058.1.308, 2004-01-04 19:28:46-02:00, marcelo at logos.cnet
> > >   Cset exclude: bjorn.helgaas at hp.com|ChangeSet|20031218183339|13120
> > >
> > > ChangeSet at 1.1058.235.6, 2004-01-04 17:53:50-02:00, marcelo at logos.cnet
> > >   Change EXTRAVERSION to 2.4.24-rc1
> > >   TAG: v2.4.24
> > >   TAG: v2.4.24-rc1
> > >
> > > None of them seem related to when the linux-2.4 changeset got merged
> > > into linuxppc-2.4.
> >
> > You also get:
> > ChangeSet at 1.1179.1.1, 2004-01-06 12:37:52-07:00, trini at kernel.crashing.org
> >   Merge in recent changes by hand.
>
> Oh.
>
> > Which is as close to 2.4.24 + powerpc changes as you're going to get,
> > automatically.  You can use 'bk export -tpatch' to back out the changes
> > which follow that you don't want.
>
> Yeah, but the point is, i am not sure if i am the person most
> appropriate to checking which of all those changeset are needed or not.
> A proper tagging would be much more appropriate, and make for easy
> reference when getting feedback and such.

A proper tagging, for distribution use, isn't possible 100% of the time.
For example, if the tree goes from v2.4.29 to v2.4.30-pre1 to
v2.4.30-pre2 all while myself/Paul/Ben are on vacation, theres no way
to bring the bitkeeper tree, as of v2.4.30-pre1 and v2.4.30-pre2 into
linuxppc-2.4 and make a tag for each.  Likewise, if we don't grab the
bitkeeper tree at exactly v2.4.30-pre2, we can't make a tag that
corresponds to exactly that.

On the other hand, it's not that hard to decide which changes are
appropriate and which are not.  Using the real example of v2.4.24, all
of the changes inbetween are for generic things, or possibly ppc64.  All
of those can be backed out.
<hat=former LinuxPPC, Inc kernel packaging guy>
We all started digging into the kernel someplace :)
</hat>.

--
Tom Rini
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/