On Tue, 2014-11-04 at 21:13 +0100, Zbigniew Jędrzejewski-Szmek wrote:
On Tue, Nov 04, 2014 at 08:30:32AM -0800, Adam Williamson wrote:
> An update has been submitted for systemd today:
>
>
https://admin.fedoraproject.org/updates/kmod-18-4.fc21,systemd-216-9.fc21
>
> with a fairly short description. I wanted to flag up that, in fact,
> systemd-216-9 is a major change from systemd-216-8 and is not really
> systemd 216 at all.
Hi Adam,
this annoucement misrepresents the situation quite a bit. Since you
are speaking from your position as QA chief
I don't hold that position, and there isn't such a position.
, your word carries a lot
of weight. We *were* in contact on IRC yesterday, I'm in #fedora-devel
semi-permanently, and it should not be a problem to show it to me
before you sent it out, since you are talking about updates I made. If
you disagreed with what I have to say and *then* sent the mail, that
would be fine, but not like this, out of the blue.
Sorry for not running it by you first, but I'm just trying to stay on
top of a whole bunch of stuff for F21 right now. I wasn't *complaining*
about anything. I just wanted to make sure people didn't rubber-stamp
216-9 on the impression that it contained a single bugfix vs -8 or
something.
Anyway, returning to the matter at hand, systemd-216-9 is fairly
close
to systemd-216-8, has patches over it to fix *known bugs*, the ones
listed in the update, a few listed on the freedesktop systemd bug
tracker, and a few small ones I found while testing the update. The
delta is not as small as I would like, but fits imho in the rules.
I didn't say otherwise, but I wouldn't characterize the diff posted
earlier as 'fairly close', it really isn't. Anything past a dozen LOC is
not 'fairly close', IMHO.
If
systemd upstream was doing point releases, this would certainly
qualify as one.
Well, I didn't want to bring it up, but that would rather clarify
things, wouldn't it? Right now we have a sort of messy situation where
what the Fedora package is is basically 'whatever's on the
systemd-stable 216 branch upstream', and what that is is 'whatever you
find most convenient right at present' - like the Fedora package, up
until a couple of days ago that branch was a few hundred commits added
to the 216 tarball, whereas now it's been completely changed to be 217
minus a few reversions.
This I guess makes sense to you as you have this kind of mental vision
of what you want '216 stable' to be, so it makes sense to at some point
say 'oh hey, my mental 216 stable is just 217 minus X, Y and Z, so let's
make the branch be that' but it's rather difficult for anyone else to
follow, there isn't much continuity. Things like point releases and
changelogs and consistent maintenance procedures and more
differentiation between upstream and downstream changes would make the
picture clearer. While I was researching the mail I kind of pieced
together all this history from the package changelog, the package SCM
commit log, the upstream SCM log, and the Bodhi update comments, but you
*do* have to go through all of that to actually figure out what the
hell's going on, because it's all just '216-x', there's no 216 point
releases, no release documentation, not a lot to differentiate package
changes and upstream changes.
Actually it was 216-2 which contained the biggest change. I built it
on Oct 7, before the alpha freeze. It was called "216" because 217
wasn't tagged yet, and I didn't want F21 to miss the bugfixes and
features which have accumulated in the upstream git. So 216-2 has most
of the post-216 commits, and 216-9 is fairly close to that.
Again, well, I posted the diff, and I disagree that it's 'fairly close'.
216-2 was certainly hugely different to 216-1, but 216-1 never went to
updates-testing, so it doesn't really signify.
> systemd-216-8 (and 216-1 through 216-5) and earlier) was more or
less
> identical to upstream systemd-stable 216:
>
http://cgit.freedesktop.org/systemd/systemd-stable/log/?h=v216-stable .
> systemd "216-9" is not built from 216 at all, it is in fact systemd-217
> with some particular changes (presumably intended to be the most
> disruptive ones) reverted. When I dropped build-related files and
> directories and documentation from the trees, did a context-free
> recursive diff, and filtered out the metadata from the diff, it still
> worked out at >7,000 lines worth of additions and removals between the
> underlying code of 'systemd-216-8' and 'systemd-216-9'. This is a
lot of
> change to land between Beta and Final.
Like I said on IRC yesterday, a large part of this is code which is
not compiled for Fedora, or unsupported [*], or tests.
> Testers, please take care to test the update thoroughly, despite the
> small bump and small description it is a major change to the package.
That I can agree with. I'd much prefer a concrete list of things
to
test in this update though, which would be *useful* and lead to a
better release. Right now you suggest that anything might be broken.
Well, I just said it's a big change and the update description doesn't
accurately encapsulate it.
I would of course be very happy with a concrete list of things to test
in the update. The update description would be a good place for it.
Another good place for concrete lists of changes would be in the
changelogs for the upstream point releases which we don't have...:)
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net