[00:24] <saiarcot895> Something in the builders doesn't like the 1.1-GB source file (https://launchpad.net/~saiarcot895/+archive/flightgear-edge/+build/5436966)
[00:24] <saiarcot895> Neither this version nor the previous version (which I've retried 3 or 4 times) succeeded, and all failed within minutes
[00:28] <wgrant> Hmm
[00:28] <wgrant> The build farm is really not designed to commonly build packages that large, but it should work.
[00:28] <cjwatson> ensurePresent seems to be synchronous
[00:28] <wgrant> I suspect things are just a bit too heavily loaded atm to fetch all the relevant files within a minute.
[00:29] <wgrant> cjwatson: It is, deliberately.
[00:29] <wgrant> And it has a 60 second timeout atm.
[00:29] <cjwatson> Why deliberately?
[00:29] <wgrant> cjwatson: So we know when it's done.
[00:29] <cjwatson> Well, right, I didn't mean it should be fire-and-forget
[00:30] <cjwatson> Bug 241646, I guess
[00:31] <wgrant> Mm, it doesn't really matter too much.
[00:31] <wgrant> That was mostly from back in the days when it blocked the master.
[00:31] <wgrant> We'd otherwise have to introduce a new call to check if a file existed in the cache, and poll it.
[00:31] <wgrant> Which doesn't seem like an improvement at all.
[00:31] <wgrant> The problem here is just that we set a 60 second timeout, and that's not enough.
[00:31] <wgrant> Even if it were async, we'd still have a timeout, and it would probably still be 60s.
[00:33] <wgrant> I'll increase the timeout to 5 minutes like we have in most other places.
[00:37] <cjwatson> I was thinking of some kind of "still going" callback from the slave, but OK
[00:37] <cjwatson> there does have to be a timeout somewhere, but I would've thought it would be better phrased as "slave download from librarian stalled"
[00:38] <saiarcot895> Perhaps a check to see if the downloaded size has changed?
[00:38] <cjwatson> That's unnecessarily roundabout given that the slave knows what it's downloading.
[00:43] <wgrant> cjwatson: I'm not sure I see the benefit of polling.
[00:45] <cjwatson> I wasn't advocating polling
[00:46] <cjwatson> I don't know quite how to phrase what I want in the buildd slave architecture
[00:47] <wgrant> Unless we radically change the communication architecture, requests have to be initiated by buildd-manager.
[00:47] <cjwatson> But there are two things that might time out: one is master<->slave (e.g. slave falls over), the other is slave<->librarian.  I'm saying those should be independent timeouts; if slave<->librarian times out then it should fail the ensurePresent call (or analogue), but otherwise I'd like it to be able to refresh the master<->slave timeout as long as it's still in progress
[00:48] <wgrant> We probably don't want an unlimited timeout on master<->slave, as slaves more than occasionally break in terrible ways and need to be killed.
[00:49] <wgrant> But it would indeed make sense to have a smaller timeout on the slave<->librarian call
[00:49] <wgrant> And in the normal case terminate it from there.
[00:49] <cjwatson> I get that it's hard with the master using xmlrpclib though, since AFAIK we can't stream a response
[00:49] <wgrant> Right.
[00:55] <cjwatson> We could break the xmlrpc spec and send a response without content-length, and push a byte every so often ... not sure that's the best option :)
[00:55] <wgrant> Heh
[00:55] <cjwatson> it'd require bodging both sides, since neither twisted.web.xmlrpc nor xmlrpclib supports that extension (though it's not unheard of elsewhere)
[00:57] <cjwatson> xmlrpc can't be the best protocol for this
[01:00] <lifeless> s/ for this/./
[03:24] <FourDollars> Hi, I want to register a new project for ibus-chewing, but Launchpad told me that ibus-chewing is already used by another project. And then I open https://launchpad.net/ibus-chewing, there is no such project. Do you know how to fix this?
[03:26] <TheLordOfTime> FourDollars: If it says it exists but you can't see it it *could* be a private project.  If you read the error page: "This page does not exist, or you may not have permission to see it."
[03:26] <TheLordOfTime> (just a thought)
[03:27] <FourDollars> TheLordOfTime: Yes, but it should be a open source project. And should not be a private project.
[03:27] <FourDollars> TheLordOfTime: Is there any way to know who owns this project?
[03:27] <wgrant> FourDollars: I've freed up the name for you.
[03:28] <TheLordOfTime> FourDollars: ask wgrant, who already fixed the problem :)
[03:28] <FourDollars> wgrant++
[03:28] <FourDollars> wgrant: Thanks a lot. :D
[03:29] <wgrant> np
[12:23] <apachelogger> wgrant: can I get a size bump to 10gib on https://launchpad.net/~neon/+archive/kf5-snapshot-weekly please :)
[12:24] <wgrant> apachelogger: Done
[12:24] <apachelogger> wgrant: thank you
[19:00] <KI7MT> Hello, noob question, but I'm trying to locate bugs for a package called yelp-tools  particularly, upstream bugs, but I can seem to fins any bug rpts at all  .. am I not looking in the correct location: https://launchpad.net/ubuntu/+source/yelp-tools
[19:00] <KI7MT> *can't
[19:08] <dobey> https://bugs.launchpad.net/ubuntu/+source/yelp-tools would be where any bugs opened against the package in ubuntu would be, it's not the upstream
[19:09] <dobey> the upstream is probably bugzilla.gnome.org
[19:20] <KI7MT> dobey, Ok, thank you.