| [08:52] <cjwatson> tomwardill: Could you have a look over https://code.launchpad.net/~cjwatson/launchpad/+git/launchpad/+merge/380350 and https://code.launchpad.net/~cjwatson/launchpad/+git/launchpad/+merge/380351 for OCI webhooks? | |
| [08:52] <tomwardill> oh, I meant to do that yesterday, had the tabs open and everything | |
| [08:52] <tomwardill> on it | |
| [08:52] <cjwatson> ilasc: Could you have a look at https://code.launchpad.net/~cjwatson/lp-signing/+git/lp-signing/+merge/380014 ? It's a prerequisite for the Juju charm | |
| [08:53] <cjwatson> Also, yesterday I found https://forge.softwareheritage.org/T1734 which relates to my recent GitRepositorySet.getRepositories changes, and updated it with a slew of details | |
| [08:56] <ilasc> cjwatson: +1 | |
| [08:59] <ilasc> have mp 380391 open for study on the to do list for today, above detail is great for that! thanks for sharing Colin! | |
| [08:59] <wgrant> Ah, excellent, that hopefully solves much of the problem. | |
| [09:00] <cjwatson> wgrant: Which does? | |
| [09:00] <wgrant> That softwareheritage.org ticket, and your reply to it. | |
| [09:00] <cjwatson> Ah yes | |
| [09:01] <cjwatson> There's a SH bod on #debian-uk who said that the people working on this are students who just kind of ... show up. Must be a nice problem to have :) | |
| [09:01] <tomwardill> can we borrow some :) | |
| [09:03] <ilasc> :) | |
| [09:07] <wgrant> That does make some degree of sense. | |
| [09:08] <wgrant> And would indeed be quite handy. | |
| [09:08] <wgrant> Though I guess a bunch of us were just students who showed up at one point. | |
| [10:37] <cjwatson> Hmm I think I can possibly reproduce the turnip-pack-virt memory leak locally | |
| [10:40] <cjwatson> Doesn't seem to reproduce when looking up nonexistent repos, but if I clone the same existent repo 100x in a loop then I see virtserver's VIRT growing by a MB or so in top. (sshserver gets bigger too, but I think that's keepalive connections and goes away again) | |
| [10:40] <cjwatson> Attacking it with meliae now to see if I'm imagining it | |
| [10:40] <tomwardill> ... I've made the wifi slightly faster in my study by replacing the cable | |
| [10:40] <tomwardill> one of us is having a productive morning. I suspect it's not me. | |
| [10:43] <cjwatson> you're optimising your environment! :) | |
| [10:50] <cjwatson> suspicious growth in function objects | |
| [10:54] <cjwatson> all in the guts of inlineCallbacks I think | |
| [10:57] <SpecialK|Canon> Investing in your infrastructure is generally worthwhile! | |
| [11:00] <cjwatson> I don't think PackVirtServerProtocol.requestReceived ever returns in the successful case | |
| [11:00] <cjwatson> I stuck a log entry at the end that I'm not seeing | |
| [11:04] <cjwatson> Oh. Well nothing ever calls .callback() on that particular deferred, only .errback(). That would probably do it | |
| [11:05] <cjwatson> (see PackClientFactory) | |
| [11:07] <cjwatson> is it a one-liner | |
| [11:08] <SpecialK|Canon> hah | |
| [11:12] <cjwatson> Need to figure out how to test it but I'm pretty sure it's just https://paste.ubuntu.com/p/vRJsZtMhcW/ | |
| [11:12] <cjwatson> meliae++, made this a great deal easier to track down | |
| [11:27] <tomwardill> cjwatson: is it worth adding the date_* columns to the OCI credentials and push rules? | |
| [11:35] <cjwatson> tomwardill: I don't think so really. They're not going to be displayed on their own pages with the "Registered by <foo> on <bar>" boilerplate or anything; they're more like settings. | |
| [14:50] <tomwardill> cjwatson: the credentials field, what do I need to do to make it an encrypted field? | |
| [14:58] <cjwatson> tomwardill: There isn't yet any prior art for making an entire column be encrypted (as opposed to one value in a larger JSON column). But you should be able to work it out, more or less, by grepping for EncryptedContainer | |
| [14:58] <tomwardill> ah, that's the magic term i was missing | |
| [14:58] * tomwardill tries | |
| [14:59] <cjwatson> tomwardill: You'll need an IEncryptedContainer implementer (based on NaClEncryptedContainerBase) registered with some appropriate name and that plugs in the correct configured key pair; and you'll need a property on the model that does getUtility(IEncryptedContainer, whatevernameyoupick).{encrypt,decrypt} | |
| [15:00] <cjwatson> tomwardill: Or if you wanted to try slightly harder you could write a custom Storm variable that does this | |
| [15:00] <cjwatson> Shouldn't be very difficult, we just haven't needed it so far | |
| [15:00] <tomwardill> the latter feels more resuable | |
| [15:00] <cjwatson> Well, variable and property, since they usually come as a paid | |
| [15:00] <cjwatson> *pair | |
| [15:01] <cjwatson> e.g. have a look at the implementation of JSON and JSONVariable | |
| [15:01] <cjwatson> And yes, I think that would be more correct | |
| [15:27] <cjwatson> tomwardill: I think you can go ahead with https://code.launchpad.net/~twom/launchpad/+git/launchpad/+merge/380200 now - the linked code change is on prod | |
| [15:28] <tomwardill> ah, cool | |
| [15:28] <tomwardill> landing | |
| [16:08] <tomwardill> cjwatson: slightly trivial question, would the storm variable live in storm, or launchpad? | |
| [16:19] <cjwatson> tomwardill: Launchpad | |
| [17:12] <wgrant> I don't much like the Storm variable approach. It would make accidental leaks easier, and leave the object unloadable on most services | |
| [17:13] <tomwardill> wgrant: I don't follow either objection there | |
| [17:14] <wgrant> Accidentally printing the object's fields would leak the secret, and examining the object's fields on most LP hosts, that don't have the wrapping key, would crash. | |
| [17:15] <tomwardill> aha, right, that makes sense | |
| [17:17] <cjwatson> I suppose | |
| [17:18] * tomwardill closes storm checkout | |
| [17:18] * cjwatson finally persuades turnip tests to pass with only a small hack | |
| [17:27] <tomwardill> excellent, DB patch landed with only minor buildbot poking | |
| [17:28] <SpecialK|Canon> Nice | |
| [17:33] <cjwatson> Could I have a review of https://code.launchpad.net/~cjwatson/turnip/+git/turnip/+merge/380646 please? As far as I can tell this fixes the turnip-pack-virt memory leak on production that's been taking down the service every couple of days | |
| [17:34] <SpecialK|Canon> Excellent! | |
| [17:36] <tomwardill> looking | |
| [17:39] <tomwardill> test_git.py is inconsistent as to assert(expected, received) or assert(received, expected) | |
| [17:40] <cjwatson> Is it? | |
| [17:41] <cjwatson> Oh, I got this backwards, silly me | |
| [17:41] <tomwardill> test_git_receive_pack_calls_spawnProcess appears to be reversed from test_git_upload_pack_calls_spawnProcess | |
| [17:41] <tomwardill> so there's both in there already | |
| [17:41] <cjwatson> receive is using assertThat not assertEquals | |
| [17:42] <cjwatson> and assertThat requires the matcher as second arg | |
| [17:42] <tomwardill> oh, so it is | |
| [17:42] <tomwardill> sorry | |
| [17:42] <cjwatson> so it was just my new code that was wrong, and I've fixed it now | |
| [17:42] <tomwardill> cool | |
| [17:42] <tomwardill> cjwatson: what fails in the test without the callback change? | |
| [17:43] <cjwatson> it hangs | |
| [17:43] <cjwatson> since yield requestReceived never returned | |
| [17:43] <tomwardill> ah... that'd do it :) | |
| [17:43] <cjwatson> *returns | |
| [17:43] <cjwatson> (though doesn't hang indefinitely since AsynchronousDeferredRunTest has a timeout) | |
| [17:43] <tomwardill> have a +1 | |
| [17:44] <cjwatson> thanks | |