[02:47] infinity: and that revert of sysvinit should be fine for yakkety anyway fwiw since the changes were strictly for an upgrade issue [02:49] slangasek: Yeah, but we also need to be able to build xenial dailies. :P [02:49] slangasek: So, need to sort this out better. [02:50] Well, build dailies, and also not break d-i netboot. === s8321414_ is now known as s8321414 [04:02] infinity: build dailies from -proposed? surely not needed right now [04:02] infinity: I do intend to take a long hard look at the SRU on monday [04:08] slangasek: No, no. I just meant eventually, not today. [04:09] Anyhow, going to head off and try to get into a vaguely correct timezone. [07:52] doko, please syncpackage giflib -s costamagnagianfranco [07:52] it should help feh [07:53] also libcec please accept [07:56] and please remove "xmbc" from yakkety, we don't needs such transitional package [07:56] "xbmc" [07:59] * LocutusOfBorg leaves, cheers! [08:02] doko, syncpackage libpng1.6 -s costamagnagianfranco <-- multiarch -dev package yeah! [08:09] LocutusOfBorg, doko: i'll sync giflib and libpng1.6 [09:52] slangasek, infinity, xnox, pitti: my hope was to fix some more things before opening the archive for general uploads. [09:53] - finish the libpng/boost and icu transitions (although libpng and boost are untangled from icu) [09:54] - find out about ghc and -fPIE, broken on amd64 and ppc64el, that will block everything in a very short time. although haskell packages seem to ftbfs everywhere [09:55] - let gcc-5/gcc-6 migrate so people can get and see the new PIE defaults [09:55] - I'm told that kernel modules are broken with -fPIE as the default [09:55] - a lot of the phone stack ftbfs [09:56] so maybe give people a chance for fixes and only open on Monday late evening or Tuesday? === maclin1 is now known as maclin [15:07] doko: I don't see any point in finishing the transitions (and some will go better with auto-sync back on). [15:07] doko: But gcc/pie investigations seem worthwhile. [15:09] pitti: There still seems to be a lot of 404s for yakkety autopkgtest pages. [16:40] $ reverse-depends node-stringprep [16:40] reverse-depends: Error: Unknown release [16:40] infinity, ^^^ what needs fixing for this? [16:41] doko: Whoever runs the ubuntuwire stuff needs to teach it about new suites. [16:41] wgrant, ^^^ [16:41] wgrant: ^ Do you have access to fix reverse-depends, or is that one of the other ubnutuwire people? [16:41] We really should move those into the Canonical DC some day, given how much we rely on those tools. [16:42] doko: You can use checkrdepends as ubuntu-archive@snakefruit instead. [16:43] doko: It's more accurate anyway (just annoyingly on a remote machine). [16:44] doko: http://paste.ubuntu.com/16034666/ [17:38] infinity: reverse-depends is tumbleweed [17:44] and yes, I'd love someone to look after it :) [18:21] tumbleweed: Where's the code live? [18:21] tumbleweed: I'm sure we could move it to snakefruit or lillypilly if all it needs is a dists mirror and some TLC. [18:21] (ie: to people.canonical.com) [18:22] tumbleweed: And perhaps cloudify it later for HA and shininess. [18:24] tumbleweed: Without looking, I'm assuming it's a combo cron job to populate a DB and CGI script to return the API? [18:25] (hopefully a flat file db of some sort, not SQL-backed madness) [18:39] infinity: WSGI not cgi. and a sqlite db [18:39] https://code.launchpad.net/~stefanor/+junk/reverse-deps [18:50] infinity: anyway, I did a git pull on my distro-info-data checkout on ubuntuwire [19:43] doko: hm, actually, this ghc thing might ring a faint bell from when I was bootstrapping s390x [19:44] doko: will take me a little while to build suitable environments to test in, but I have an idea [19:44] ta [19:45] (I vaguely remember that I may have had to do an extra cycle, the first with a wrapper in place of /usr/bin/ghc - will investigate that) [19:47] I suspect we have a lot places where people thought it was a good idea to do "if s390x { -fno-pie -no-pie }" and we need to stop making that conditional. [19:47] Since it's a no-op for no-pie arches anyway, I'd assume. [19:49] doko: Is there a plan to move to gcc-6 this cycle too, or is that for 17.04? [19:49] infinity: Yeah but that's not quite the problem here. I think it needs a slightly bigger hammer the first time round to cope with the old ghc not emitting no-PIE flags [19:49] cjwatson: Sure, I didn't mean for ghc, but rather in general. [19:50] cjwatson: Rebootstrapping compilers is a special kind of suck. [19:50] One reason it was conditionalised, IIRC, was that it hadn't been that long since -fno-PIE (or maybe -no-PIE) started existing at all. [19:50] In fact I think I remember at one point it may only have been valid with the set of patches applied on s390x (but I could be misremembering that). [19:51] Definitely remember running into something in that general area. [19:51] I wish they'd picked a different TLA for that. I get hungry every time it's mentioned. [19:51] Mmm, PIE. [19:51] * infinity considers a trip to the pie shop. [19:52] The pie shop that cleverly waited until the lease was up on 314 10th Street and then made the building owners an offer they couldn't refuse. [19:52] (Seriously) [19:53] infinity, I'd like to [19:54] doko: For opening, or later in the cycle when more ftbfs bugs have been fixed? [19:54] (Assuming the latter since you haven't changed -defaults yet) [19:58] infinity, not before 6.2 is released which will be late June/early July [19:58] doko: Check. [19:58] hrm, haskell packages are not using Built-Using attributes ... [20:00] The hashed runtime deps achieve basically the same result, I would think. [20:01] no, not if there aren't any [20:03] doko: When are there ever not any? [20:03] Every -dev package depends on the -dev packages it was built with. [20:03] infinity, haskell-text-icu builds a -dev package only [20:03] doko: Yeah, check its deps. [20:03] not for non-haskell libs [20:03] (base)adconrad@nosferatu:~$ apt-cache show libghc-text-icu-dev | grep ^Depends [20:03] Depends: libghc-base-dev-4.8.2.0-0d6d1, libghc-bytestring-dev-0.10.6.0-9a873, libghc-deepseq-dev-1.4.1.1-614b6, libghc-text-dev-1.2.2.0-2c09c, libc6 (>= 2.2.5), libicu-dev [20:04] Ahh, I see. Yeah, no Built-Using for the C libs. [20:04] That's probably something dh-haskell could fix. [20:04] s/could/should/ [20:09] And thanks to Haskell transitions happening every 3 weeks, I assume a dh-haskell change affects the entire rdep tree pretty quickly. :P [20:18] cjwatson, just in case this gives you more information: https://launchpad.net/ubuntu/+source/haskell-text-icu/0.7.0.1-4build2 [20:18] yeah, the fun with -fno-pie is that it was only added around the same time s390x started defaulting to pie [20:38] xnox, would be nice if you could look tomorrow at the boost related ftbfs, which are blocking icu: osmium and mpd [20:50] hi can an archive admin please explain that? https://launchpad.net/ubuntu/+source/libcec/3.1.0+dfsg1-4 [20:50] double publishing? cjwatson ^^ [20:55] That looks like an unpleasant race... [20:55] wgrant: ^ [20:55] :I* [20:55] seems so [21:15] cjwatson: the patches for -fno-pie were always present on all architectures, but it was only introduced in the 16.04 cycle, so if you wanted backportability to any arch that existed prior to 16.04, you would need to conditionalize [21:22] Ahh, gross. [21:23] I was told these races were harmless [21:24] doko: That race looks harmful, as it has duplicate build records, both of which succeeded. So, no way to reconcile disk and librarian. [21:24] infinity, well, then re-upload [21:24] (Most double build record races result in fail/success or fail-to-upload/success pairs) [21:24] doko: Sure, I just want wgrant/cjwatson to look at the current state first to see WTF. [21:24] doko: Reuploading it likely the short-term fix once they've had a look. [21:25] I think that's known [21:25] s/it likely/is likely/ [23:00] doko: basically identical, as expected. No point uploading haskell-* stuff until ghc is sorted. [23:01] infinity: Just looks like the usual race when something's double-accepted from the queue or similar [23:01] Indeed just reupload.