[00:00] <cjwatson> I suspect lazr.restfulclient may not be sufficiently careful, although that probably doesn't matter for this
[00:08] <cjwatson> though actually that might be fine if it's only opening the filename and possibly stuffing it back into a header as a UTF-8-encoded bytestring
[00:33] <tsimonq2> Hmph, so it seems I can no longer `make schema`...
[00:33] <tsimonq2> While:
[00:33] <tsimonq2>   Installing scripts.
[00:33] <tsimonq2>   Getting distribution for 'pytz==2017.2'.
[00:33] <tsimonq2> Error: Couldn't find a distribution for 'pytz==2017.2'.
[00:33] <tsimonq2> And that's installed.
[00:34] <tsimonq2> Oh. Duh. 2017.3 is installed. Let's manually edit that file...
[00:35] <wgrant> tsimonq2: bzr up download-cache
[00:35] <wgrant> LP doesn't use many system Python dependencies.
[00:35] <wgrant> It gets them from download-cache
[00:35] <tsimonq2> wgrant: Oh. Thanks.
[00:36] <tsimonq2> wgrant: Did I overlook this in the docs somewhere or should it go in there? :)
[00:37] <wgrant> tsimonq2: rocketfuel-get does it
[00:41] <tsimonq2> Oh, gotcha.
[01:01] <tsimonq2> cjwatson: Ok, so we agreed on doing the factory operation for bug 439470. So should someone be able to call CveSet via the API with a CVE number and it will automatically set it, should I give each (valuable) function its own thing to be able to access via the API, or should the database addition function be the only one that's publicly accessible?
[01:01] <mup> Bug #439470: Cannot attach currently-unknown CVEs via linkCVE() <api> <lp-bugs> <platform-want> <Launchpad itself:In Progress by tsimonq2> <https://launchpad.net/bugs/439470>
[01:01] <tsimonq2> cjwatson: I *could* be forgetting a little bit of our convo but I don't have the logs to grep any more and I don't remember when it was...
[09:29] <cjwatson> tsimonq2: I'd probably be economical: the only new thing exposed on the API would be CveSet.new (as cves.new).
[09:31] <cjwatson> tsimonq2: Webservice type declarations would get difficult if you tried to make e.g. linkCve take either an actual Cve object or a string.  If you wanted to do that I think you'd need to give linkCve an extra string-flavoured argument and require exactly one of cve or the new arg to be passed.
[09:32] <cjwatson> tsimonq2: But I'd be inclined to do just the minimal functional thing as a first pass.
[11:52] <tsimonq2> Ack cjwatson, thanks!
[12:29] <cjwatson> wgrant_: codetree seems to work well; just needs copying into the relevant places and a shim in update-sourcecode to handle forcing HTTP (useful e.g. for buildbot).  Will poke as time permits.
[12:30] <cjwatson> Parses our config file unchanged.
[12:31] <wgrant> Handy.