Home
release-team@conference.openafs.org
Friday, April 27, 2018< ^ >
Room Configuration
Room Occupants

GMT+0
[00:20:03] kaduk@jabber.openafs.org/barnowl joins the room
[00:48:29] mvita joins the room
[04:20:58] mvita leaves the room
[11:43:02] Marcio Barbosa joins the room
[12:54:31] wiesand joins the room
[13:01:05] <wiesand> Good morning
[13:01:47] meffie joins the room
[13:01:54] <meffie> good morning
[13:02:05] <wiesand> Ben is supposed to be on  plane und thus not here
[13:02:12] <wiesand> Hello Mike
[13:02:26] <meffie> i wish him a safe flight.
[13:03:13] <meffie> thank you for the find-debuginfo.sh fix. i spent for too long trying to understand that script.
[13:03:16] <wiesand> I think anyone having to do with openafs does….
[13:04:17] <wiesand> I'm sorry that I couldn't firmly say "I'll fix it" last week.
[13:04:37] <wiesand> And thanks for looking into it!
[13:05:14] <meffie> i suspected it was the pattern, but i was trying to find out exactly why. so much changed in that script.
[13:05:57] <wiesand> Now we need Ben to merge 13030 (and probably 13031, though I think that's still non- fatal)
[13:06:23] <wiesand> Yes, lots of churn there.
[13:07:16] <wiesand> BTW did you notice they now create "debugsource" rpms too? I wonder whether we'll have the same kind of trouble with those as with the debuginfo ones.
[13:07:52] <meffie> hmm i did not look at that closely.
[13:09:53] <wiesand> EL6.10 beta was released yesterday.
[13:10:42] <wiesand> Given that they broke us with the 7.4 and 7.5 kernels, testing with the 6.10 kernel early seems a good idea.
[13:10:59] <meffie> ubuntu 18 lts was released yesterday too, i believe.
[13:11:10] <wiesand> Right
[13:11:44] <wiesand> My bionic beta VM has a 1.8.0pre5 client from universe :)
[13:12:24] <meffie> yay!
[13:12:59] <wiesand> I'll do some EL6.10 smoke testing before the next meeting - unless anyone else wants to.
[13:14:14] <wiesand> I'll probably also know the state of affairs in U18.04 final before the next meeting.
[13:15:02] <meffie> ok, i'm not sure offhand if i have access to EL6.10 yet.
[13:15:59] <wiesand> I do. We afford a subscription for such purposes.
[13:16:25] <meffie> ok, that's great.
[13:16:45] <wiesand> Though with one of those free developer subscriptions it shouldn't be a problem either.
[13:17:07] <meffie> i think we have something like that. i'll ask again.
[13:17:38] <meffie> centos is still at 6.9
[13:18:07] <wiesand> And BTW, the SL folks have been really quick with 7.5 - chances are they'll have the kernel packages rebuilt for testing within days.
[13:19:07] <meffie> ok.
[13:20:26] <wiesand> CentOS will be at 6.9 until they have 6.10 ready a couple of weeks after RHEL 6.10 GA. They don't issue betas. Packages may start showing up in the CR repository betwenn RHEL GA and CentOS GA. Bet last time I looked (this week), there was nothing for 7.5 in there yet.
[13:21:35] <meffie> ok, thanks, i knew there was a lag, but i did not realize it was so long.
[13:22:49] <wiesand> And Linux 4.17 seems to become the second one in a row for which we don't have to scramble. It's a bit hard to tell currently because now all the daily/rc builders seem to have trouble setting up the kernel…
[13:23:59] <meffie> ah, ok, i'll take a look a the buildbots. perhaps something changed in the ubuntu packaging
[13:24:25] <wiesand> In other words those daily reports turned out to be very very useful.
[13:25:05] meffie making an internal ticket for myself, cc'ing mark too.
[13:25:19] <wiesand> (remember I was a bit sceptic when you brought this up - sometimes being wrong is good)
[13:28:00] <wiesand> And then I managed to get much of my homework done this week. Quite a few new changes pulled up to openafs-stable-1_6_x (1.6.23) <https://gerrit.openafs.org/#/q/status:open+project:openafs+branch:openafs-stable-1_6_x+topic:1.6.23> - review review review…
[13:29:12] <meffie> thank you!
[13:30:40] <wiesand> But I stumbled over 12658. Unless we axe the pre-fedorakmod modules on the old stable branch too, I think that one isn't for 1.6 and we should rather spend the effort on getting 1.8. further along.
[13:32:25] <wiesand> And that's about what I had on my agenda for today.
[13:32:27] <meffie> 12658 is not too urgent, but i think it does fix a bug
[13:32:52] <wiesand> A longstanding one though, right?
[13:33:33] <meffie> yes. very long. i think the bug was people would specify --define build_modules 1 (without --define build_userspace 0)
[13:33:44] <meffie> and that would or somethink like that.
[13:34:16] <meffie> and that would fail. but i could be remembering wrong, since it was so long ago.
[13:34:29] <wiesand> I'd love to jettison the old module builds and get the 1.6/1.8 specs in sync again. But I have a feeling that some in the community would disapprove…
[13:35:30] <meffie> i dont know, but it seems are openafs.spec file is very overly complex.
[13:35:40] <meffie> s/are/our/
[13:36:04] <wiesand> agreed
[13:39:19] <meffie> i probably would be better to have separate spec files, one for userspace, and one for modules. or somesuch.
[13:40:49] <meffie> and a lot less conditional junk.
[13:41:13] <wiesand> It would already help to have it either build userspace or a single module at a time.
[13:42:50] <meffie> that's what i normally do. i have a python script that invokes rpmbuild several times; it builds userspace first, then one module at a time. it's just easier to debug.
[13:43:47] <wiesand> that's how I changed the SL packaging too - it's still messy, but a lot less messy than before.
[13:43:51] <meffie> e.g. it sets the --define "build_userspace x" --define "build_module y"
[13:44:31] <meffie> ok, great!
[13:44:38] <wiesand> 90% of the build time goes into the configure step for the kernel module anyway
[13:45:02] <meffie> that is a good point.
[13:45:10] <wiesand> at least if you do parallel builds, which has worked flawlessly for a long time
[13:45:57] <wiesand> I was thinking about submitting a change that changes "make" to "make %{_smp_mflags}" for a while.
[13:47:55] <wiesand> I did that for my tests for the find-debuginfo regex, so I'm pretty sure 1.8 is fine with that too, at least up to -j4.
[13:47:57] <meffie> i dont know that flag... it look like it sets the make -j flag?
[13:48:43] <wiesand> Yes, to the number of cores (or even HT threads) available on the build system.
[13:49:58] <meffie> and we could override it, to set it to -j1 for debugging?
[13:50:22] <wiesand> The SL spec has been doing this for many years, and I frequently build 1.6 like this on systems with up to 48 threads.
[13:51:27] <meffie> wow
[13:53:10] <wiesand> Yes, for debugging you can —define it on the rpmbuild command line. Or build in a single VCPU VM ;-)
[13:53:55] <meffie> sounds good then.
[13:54:19] <meffie> i'll have to give my builders more cores :)
[13:55:07] <wiesand> Alas, the's Amdahl's law :-(
[13:55:46] <wiesand> If you have that many cores available, the configure for the module becomes very very painfully dominant.
[13:56:15] <wiesand> s/the's/there's/
[13:56:47] <meffie> yes. most of the configure checks are independent. it's too bad there is not a way to make them in parallel.
[13:57:20] mvita joins the room
[13:57:27] <wiesand> I didn't dare to ask once more whether that's feasible.
[13:57:31] <meffie> also, configure -C can use a cache, we could save the caches and reuse them per platform type.
[13:58:01] <meffie> i use the configure cache a lot to save time while developing.
[13:58:46] <wiesand> I didn't know about -C. Sounds useful for us testing or developing, but hardly for production builds.
[13:59:40] <meffie> ageed, it would be riskier
[14:00:05] <mvita> gah, sorry I forgot again.
[14:01:07] <meffie> it would have been better to just lie and said you got stuck in traffic :)
[14:01:34] <meffie> off to another meeting, anything else Stephan?
[14:01:37] <mvita> I cannot tell a lie.
[14:01:49] <mvita> (which may be a lie itself)
[14:04:30] <wiesand> Participation in this meeting is voluntary…
[14:05:37] <wiesand> If it's forgotten about, that's my fault
[14:06:12] <wiesand> But let's adjourn for today. Thanks a lot!
[14:06:27] <meffie> thanks
[14:09:07] wiesand leaves the room
[14:25:03] meffie leaves the room
[15:58:49] mvita leaves the room
[16:02:13] mvita joins the room
[21:47:56] Marcio Barbosa leaves the room
[23:59:36] mvita leaves the room