- Wednesday, June 18, 2014

release-team@conference.openafs.org

Wednesday, June 18, 2014< ^ >

Room Configuration

Room Occupants

GMT+0
[12:16:50] wiesand joins the room
[12:17:02] <wiesand> test
[13:57:39] Marc Dionne joins the room
[14:01:05] <wiesand> Hello
[14:01:16] <Marc Dionne> hi Stephan
[14:01:31] <wiesand> How's Linux 3.16 doing?
[14:01:31] deason joins the room
[14:02:28] <Marc Dionne> just pushed 2 commits a bit earlier to gerrit for 3.16
[14:03:45] <Marc Dionne> however 3.16-rc1 has a broken btrfs, so I currently can't verify that it works with -rc1, but it tested ok with mainline from a few days before
[14:05:02] <wiesand> Ok, thanks. These don't look too invasive.
[14:05:19] <wiesand> ETA for 3.16 is ~ 2 months from now?
[14:05:42] <Marc Dionne> that's the usual timeframe, ~9 weeks
[14:06:56] <wiesand> That should allow us to get out 1.6.10 ready for 3.16. But headroom's not comfortable.
[14:09:37] <wiesand> From August 20 to September 30, I'll have a bit more time for openafs than usual. Except for a block of ~3 weeks in there where I'll probably be travelling. When and if the latter happens isn't clear, and there won't be much advance notice. Whether I'll manage to be online frequently, as last year in Iceland, isn't clear either.
[14:11:35] <wiesand> Marc, a few meetings ago Simon mentioned your CheckTokenCache speedup changes which should go into 1.69 (back then...).
[14:11:44] <wiesand> I think those haven't been pulled up yet?
[14:13:34] <wiesand> BTW I'm working on making the RHEL/Fedora 1.6.9 binaries available.
[14:13:41] <Marc Dionne> i can push that to 1.6 - don't think anyone has yet
[14:13:49] <wiesand> Marc: Thanks.
[14:14:08] <wiesand> I'm a bit unsure whether to upload RHEL7 binaries yet.
[14:14:53] <deason> from a previous meeting, I thought we were considering a change, but I thought I saw rhel7 binaries were already uploaded for some version
[14:15:27] <wiesand> If/once gerrit 11300 works (it's untested) and is agreed, it would probably ok to upload what we have now.
[14:16:33] <wiesand> Yes, for 1.6.8 and 1.6.7, but only for the RHEL7 rc.
[14:17:51] <deason> "a change" meaning, we either don't upload rel7 rpms, or use non-transarc paths
[14:17:58] <deason> er, rhel7
[14:18:24] <wiesand> That was proposed, with no feedback.
[14:20:14] <deason> during the meeting, I thought I supposed phasing out our rpms, and at least one other person also supported it
[14:20:20] <deason> er, I supported
[14:20:53] <Jeffrey Altman> My recollection of the discussion in this room was that phasing out openafs rpms was a good idea
[14:20:59] <shadow@gmail.com/barnowl853202E0> killing off non-FHS rpms, whether we fix ours or switch to rpmfusion's
or whatever, is fine with me, at least in the general sense
[14:21:37] <wiesand> I think going completely FHS is rather something to do with the change to 1.8.
[14:21:55] <Jeffrey Altman> I believe there was a question about whether rpmfusion's rpms had issues we would want to ask downstream to modify
[14:22:26] <deason> as I believe was brought up earlier, it makes more sense to do the transition on an RHEL7 version boundary than an openafs version boundary
[14:22:47] <Jeffrey Altman> and whether there is a downstream maintainer since Jack is no longer working with AFS
[14:23:14] <wiesand> I looked at the rpmfusion ones. They're really different.
[14:23:17] <Jeffrey Altman> Ken is/was a co-maintainer
[14:23:49] <wiesand> No answer from Stephen yet whether they could easily switch to using FHS.
[14:24:08] <Jeffrey Altman> I agree with Andrew that switching to FHS should be done on a distribution boundary
[14:24:17] <wiesand> I think Ken still is. But priorities are likely to shift a bit.
[14:24:45] <wiesand> Yes, I tend to too.
[14:25:53] <deason> we'd still need/want the option of a -transarc transitional package in that case (which just has compat symlinks); you shouldn't need to move files from e.g. /usr/afs/etc, but you still may have some tools that refer to /usr/afs/etc
[14:26:32] <deason> which I assume doesn't exist now in the existing packaging
[14:26:45] <Jeffrey Altman> can there be a separate transition package that has the FHS package as a dependency and adds the symlinks ?
[14:27:57] <deason> that is (or, can be) what I mentioned; yes, I think that should exist
[14:30:15] <shadow@gmail.com/barnowl853202E0> i assume so
[14:30:36] <shadow@gmail.com/barnowl853202E0> long ago i did a revised spec file that spit one out, but i never got
that works and it's probably bitrotted to death
[14:30:41] <shadow@gmail.com/barnowl853202E0> "one spec to rule them all"
[14:31:20] <wiesand> I'd object to having this in the same spec.
[14:31:38] <wiesand> That's sufficiently messy as it is.
[14:33:23] <wiesand> A spec producing FHS compliant packages should be separate, an do away with everything pre-EL7/F21 IMO.
[14:33:53] <wiesand> But it could of course create a subpackage with compat symlinks.
[14:34:36] <wiesand> BTW would we still want to be compatible with today's openafs-compat?
[14:34:53] <wiesand> openafs-compa-compat?
[14:35:42] <wiesand> (bw compat symlinks in /usr/afsws)
[14:36:14] <deason> openafs-transarc
[14:36:22] <wiesand> gerrit 11300 would be the cheap way out...
[14:37:13] <deason> if we don't do this now, we're stuck with providing packages for rhel7 forever, or we have to have a transition period where we wean people off of our packaging
[14:37:25] <wiesand> We could ship the usual binaries for RHEL today and fix 131859 with 1.6.10 for EL7 only.
[14:38:25] <wiesand> Andrew: right.
[14:39:25] <wiesand> gerrit 11300 also demonstrates what kind of rpm voodoo is required to make updates work that replace files/dirs with symlinks
[14:40:43] <wiesand> The rpmfusion spec seems to assume a lot of rpmfusion infrastructure. So it's not just "dump it into gerrit".
[14:41:02] <deason> ? I don't want to dump anything into gerrit
[14:41:09] <deason> I want us to stop shipping rpms for rhel7; is that not clear?
[14:41:24] <deason> (or rather, "stop shipping rpms, starting with rhel7")
[14:42:48] <wiesand> Fine, let's stop. I can do that ;-)
[14:43:41] <jhutz@jis.mit.edu/owl> > going completely FHS is rather something to do with the change to 1.8
Nope.  This needs to coincide with a major distro version, not an upstream
openafs version.
[14:44:50] <wiesand> You're right. But unless we just stop shipping RPMs for EL7 now, I don't see this getting done in time.
[14:45:00] <jhutz@jis.mit.edu/owl> And ==deason; I think if we're going to switch to telling people to use the
rpmfusion packages, then we should just do that, and not try to produce
them ourselves.
[14:45:24] <deason> well, can we? I'm not sure if you're clear that that's a serious suggestion; I would actually rather us not be shipping packaging; downstream is supposed to do that
[14:45:38] <jhutz@jis.mit.edu/owl> Bear in mind that I neither produce nor consume openafs RPMs, and never
have.
[14:45:42] <Jeffrey Altman> and if we want to produce FHS packages ourselves (which we don't) then we should not ship anything until we have packages we are happy with
[14:45:43] <deason> I'm not joking or saying "let's all just stop doing anything" or whatever
[14:45:58] <wiesand> NB there are no rpmfusion packages for RHEL.
[14:46:27] <wiesand> Only the elrepo ones, handling the kernel module in a way disapproved by the experts.
[14:46:28] <Jeffrey Altman> Its the EPEL packages for RHEL
[14:46:34] <Jeffrey Altman> elrepo
[14:46:35] <wiesand> And I think unmaintained too.
[14:47:16] <wiesand> You want to recommend using weak-updates kmods?
[14:47:47] <Jeffrey Altman> No, I recommend that we advise downstream to fix the packaging.
[14:48:27] <wiesand> Well, I tried that a couple of years ago.
[14:48:32] <deason> yes, you can still work on the packaging and all; it should just be part of "rpmfusion" not "openafs.org"
[14:49:02] <wiesand> EL5 times. It didn't even have a complete list of ABI deps.
[14:49:10] <deason> (is weak-updates the mechanism that tries to use libafs in different linux kernel versions without recompiling?)
[14:49:23] <Jeffrey Altman> yes
[14:49:24] <deason> because if so, yeah, we should be disabling that as much as we can
[14:49:29] <wiesand> Yes, that one.
[14:51:44] <wiesand> I don't have time to get involved in rpmfusion. Who does?
[14:53:06] <deason> well, I thought we'd be asking ken if he's still doing that
[14:53:25] <deason> or in general mention this to the community (since they as end-users need to know anyway)
[14:54:15] <deason> I'm not entirely sure I see the difference in time commitment in fixing things here vs another packaging repository...
[14:55:46] <wiesand> Well, I do.
[14:56:46] <deason> okay; can I mention this to -info and -devel and we can come back to this topic next week?
[14:57:26] <deason> and anyone can speak up if they want to work on the rpm packaging, or ask around if someone they know wants to
[14:57:36] <deason> (I suppose specificaly, RHEL rpm packaging, specifically RHEL7+)
[14:57:41] <wiesand> Proposal: I won't upload EL7 binaries. I'll announce the other RH binaries on -announce, and use that occasion to announce that we plan not to build for EL7+ and F21+ at all anymore.
[14:57:43] <deason> er, rhel/fedora
[14:58:09] <wiesand> And why.
[14:58:18] <Jeffrey Altman> I think that is a good plan.
[14:59:27] <wiesand> Me too :)
[15:00:08] <deason> yes, the "why" people would want :) and it's not set in stone yet
[15:00:28] <wiesand> The last item I put on the agenda was "what's not yet in gerrit that was agreed to be considered for 1.6.10"
[15:01:08] <wiesand> We already talked about Marcs token cache changes.
[15:02:12] <wiesand> I think everything else is ready for review. Missing something?
[15:03:15] <deason> I suppose it is too soon to try and get a fixed tmay-caching back in?
[15:04:02] <wiesand> Well, that was reverted in an emergency.
[15:04:29] <Jeffrey Altman> and no fix has been approved on master.
[15:05:01] <wiesand> We can try. But if we want to get 1.6.10 out in time (mid August), we can't wait long with pre1.
[15:05:09] <wiesand> And as Jeffrey said...
[15:05:41] <wiesand> I hope I can resume merging tomorrow.
[15:06:07] <wiesand> There are a few that clearly had sufficient review (and are running on my test systems).
[15:06:27] <wiesand> But many changes need more review. And we'll be missing Ben the next two weeks.
[15:07:21] <wiesand> [uploading EL5 binaries, the last ones]
[15:07:48] <deason> is there any particular order or anything you want for reviews?
[15:07:58] <deason> as in, priority; which to do first
[15:08:47] <wiesand> Last week's agenda had them grouped into stacks and misc other stuff.
[15:09:31] <wiesand> Please do them group after group, and start at the bottom of the stack.
[15:09:52] <wiesand> This helps merging changes faster.
[15:10:33] <wiesand> misc:            10946 11075 11112 11151 11154 11155 11159 11193
                  11195 11204 11207 11208 11209 11218 11219 11220
cmd:             11210 11211 11212 11213 11214
coverity:        11014+11206 11196
build:           11158 11205
parallel build: 11221 11222 11223 11224 11225 11226 11227 11228 11229
                  11230 11231 11232 11233 11234 11235 11236 11237 11238
                  11239 11240 11241 11242
stack reduction: 11162 11163 11164 11165 11166 11167 11203
messages:        11181 11182 11183
[15:10:51] <wiesand> But forget cmd, as discussed last week.
[15:11:06] <deason> yeah, I have the list; but any grouping before others?
[15:11:14] <deason> or is that what you meant by "bottom of the stack"?
[15:12:01] <wiesand> Groups are mostly orthogonal, no dependencies in between.
[15:12:20] <Jeffrey Altman> stephan, which are the series that you are testing in production?
[15:13:02] <wiesand> give me a second
[15:14:23] <wiesand> - next test build, based on HEAD at 9492d2a, +
- misc:            10834 10946 11075 11112 11151 11154 11155 11159 11193
-      11195 11204 11207 11208 11209 11218 11219 11220
- cmd:      11210 11211 11212 11213 11214
- coverity:    11014+11206 11196
- build:    11158 11205
- parallel build:  11221 11222 11223 11224 11225 11226 11227 11228 11229
-         11230 11231 11232 11233 11234 11235 11236 11237 11238
-    11239 11240 11241 11242
- stack reduction: 11162 11163 11164 11165 11166 11167 11203
- messages:    11181 11182 11183
[15:14:36] <wiesand> + 10834 reverted
[15:15:15] <Jeffrey Altman> everything in your list.
[15:16:00] <wiesand> No surprise.
[15:16:29] <wiesand> I feel best about merging a change when it has sufficient positive review and I've seen it run ;-)
[15:16:38] <Jeffrey Altman> understood
[15:17:14] <Jeffrey Altman> I have to run.
[15:17:17] <wiesand> And it's almost everything we had before the volscan pullups.
[15:17:25] <wiesand> Thanks Jeffrey. Bye.
[15:17:43] <wiesand> I think we're mostly done anyway.
[15:17:53] <wiesand> Anything else to discuss today?
[15:19:24] <deason> not here
[15:19:49] <wiesand> Ok. Thanks a lot for participating!
[15:19:58] Marc Dionne leaves the room
[15:22:29] <wiesand> I have to leave too. Bye.
[15:22:35] wiesand leaves the room
[15:22:38] deason leaves the room