Mike Kupfer wrote:
>>>>>> "Jan" == Jan Setje-Eilers <setje at smack.Eng.Sun.COM> writes:
>
> Jan> I don't think anyone was suggesting allowing actually multiple
> Jan> deltas. The case in which we might have considered talking about an
> Jan> exception would have involved a create (so no pre-existing gate
> Jan> rev), in essence introducing one bit of external history from the
> Jan> beginning.
>
> Okay, I misunderstood the suggestion. But then it seems like merging
> with upstream gets harder over time, since the only baseline
> ("ancestor") file in the repository is the first one. Even if local
> changes are propagated upstream, I'm not aware of any merging tools that
> make it easy to skip over those changes, so that you only see the latest
> ones (i.e., the ones that are interesting). So, as you say, it would
> cease to be useful after awhile.
>
> The other options that I've seen so far (unless I missed one in the
> flood of email) are:
>
> 1. keep diffs between the ON source and the upstream source in the ON
> tree (the project team's current proposal).
>
> 2. keep diffs between the ON source and the upstream source somewhere
> else, probably the project web space on opensolaris.org.
>
> 3. keep a snapshot of the upstream source somewhere else (again,
> probably the project web space)
>
> I'd reject 2 off the top. Keeping the diffs separate makes it much more
> likely that they will get out-of-date.
>
> Choice 3 seems okay; it's better than choice 2 in that the snapshot does
> not become stale if additional local changes are made. Updates from
> upstream could be handled with a 3-way merge tool; or by generating
> diffs between ON and the snapshot, and applying those to the upstream
> version.
>
> I don't have a problem with 1, though I'm aware that others do. While I
> can sort of see their point, I don't understand why this seems to be a
> must-fix item. Yes, the diffs have to be manually maintained. But so
> do READMEs, and those seem to be okay (there are over 100 of them in the
> gate).
>
The majority of my problem was with (from Roland's initial mail):
>> ####----> All changes to the upstream sources ====> MUST BE <==== stored
>> in *.diff files (and a README per *.diff) in usr/src/lib/libshell/misc/
My objection is based on the fact that taking this as written, the number of
these diffs would grow with every change made by us to the ksh93 source base
from here on out.
Such that we aren't talking about "a diff", but a growing collection thereof
and a comment describing each. This is also why I suggested use of the SCM,
since the way I'm reading it, these diffs and READMEs would precisely
duplicate information already contained therein.
-- Rich
