I'm issuing a second request for comment here. IMO, this is a pretty
dramatic departure from prior release strategy, and (for a few users)
it will be a pain to grab an old client to clean out the stale logs.
(or to do a new checkout, and figure out what changes to port over)
We're getting *very* close to the point where we're going to start on
a major hack/slash against the loggy code. We need to be sure that
people are bought into the position of, "oops. you crashed at a bad
spot. then upgraded your client. you're in a tough spot".
I'll assume "it's a total edge case, so yah. any stale logs prevent an upgrade."
If the problem space is unclear to anybody, then please let me know,
and Hyrum/myself can explain further.
Thanks,
-g
On Thu, Apr 9, 2009 at 03:01, Greg Stein <gstein_at_gmail.com> wrote:
>...
> Regarding upgrading: I believe we should always be able to upgrade any
> production release. IOW, upgrade 1.0 working copies. I think it is
> simply part of our compatibility story across 1.x releases.
>
> The question at hand isn't so much upgrading as "should we expend the
> effort to run stale logs from within the 1.7 codebase, or send people
> back to a prior release to clear those out first?" (or they can just
> do a new checkout) Running stale logs would be a HUGE effort, and
> potentially affecting mere handfuls of people. Thus, the reason for
> Hyrum's email.
>
> Cheers,
> -g
>
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2033486
Received on 2009-05-02 18:59:22 CEST