All times shown according to UTC.

Time Nick Message
00:01 knielsen joined #buildbot
00:15 github-buildbot joined #buildbot
00:15 github-buildbot buildbot: 07master 03Dustin J. Mitchell * e095585 (1 files in 1 dirs): revert debugging change, accidentally committed -
00:15 github-buildbot left #buildbot
00:15 knielsen left #buildbot
00:32 knielsen joined #buildbot
01:53 _wms joined #buildbot
02:10 _wms left #buildbot
02:25 [Arfrever] left #buildbot
02:25 [Arfrever] joined #buildbot
03:21 knielsen left #buildbot
03:37 knielsen joined #buildbot
03:53 dwlocks left #buildbot
04:52 _wms joined #buildbot
05:02 jhford left #buildbot
05:05 jhford joined #buildbot
05:36 _wms left #buildbot
05:53 Caemyr left #buildbot
07:05 sjlehtin left #buildbot
07:08 Caemyr joined #buildbot
07:08 sjlehtin joined #buildbot
07:36 mattock joined #buildbot
07:46 AlexanderS left #buildbot
07:48 lordross joined #buildbot
07:49 AlexanderS joined #buildbot
08:06 Mook left #buildbot
08:22 mjw joined #buildbot
10:11 bxc c
10:34 jochen__ left #buildbot
10:34 jochen__ joined #buildbot
11:21 bewst joined #buildbot
12:41 bewst left #buildbot
12:46 armenzg joined #buildbot
13:06 armenzg left #buildbot
13:19 armenzg joined #buildbot
13:54 armenzg left #buildbot
14:05 _wms joined #buildbot
14:27 armenzg joined #buildbot
14:47 Caemyr left #buildbot
14:49 catlee joined #buildbot
14:56 redheadphones d
14:57 jhegge joined #buildbot
15:00 tomprince e
15:00 catlee f
15:01 * djmitche hooks #buildbot up to his MIDI synth and hopes y'all play something more interesting
15:56 ayust|home d
15:56 a
15:56 b
15:56 f#
15:56 g
15:56 d
15:56 g
15:56 a
15:57 Arfrever joined #buildbot
15:58 djmitche the Buildbot theme song?
16:00 ayust|home actually that's the baseline from pachebel's canon in d, but sure
16:01 djmitche :)
16:03 ayust|home (related, if you've never seen the pachebel rant, search on youtube for 'pachebel rant')
16:28 lordross left #buildbot
16:29 redheadphones i am a fan of this channel
16:31 djmitche <-- click the "like" button
16:36 bewst joined #buildbot
16:53 dwlocks joined #buildbot
17:15 _wms left #buildbot
17:16 _wms joined #buildbot
17:45 dsal joined #buildbot
17:53 lordross joined #buildbot
18:00 Caemyr joined #buildbot
18:16 djmitche dwlocks: can you send verm__ a build and builder pickle from 0.7.12?
18:16 just check it doesn't have sekrits in it :)
18:37 dwlocks left #buildbot
18:58 bewst I don't understand how arguments get passed to a LoggedRemoteCommand.  Does that args dictionary get turned into —key=value or something?
18:58 djmitche the args dict on the master side gets passed over the PB connection
18:59 bewst Ohh, it's the slave step that knows how to interpret it?
18:59 djmitche yes
18:59 bewst Which means we can't add a new SCM without updating slaves?
18:59 djmitche correct
18:59 tomprince has been working on fixing that
18:59 and it was a GSoC project idea, too
18:59 bewst Was just going to ask
18:59 djmitche although nobody proposed it
18:59 bewst nice
19:13 _wms left #buildbot
19:22 github-buildbot joined #buildbot
19:22 github-buildbot buildbot: 07master 03Dustin J. Mitchell * 39cdbc3 (5 files in 4 dirs): Fix db.changes.addChange to not create or return a Change instance ...
19:22 buildbot: 07master 03Dustin J. Mitchell * a252ba2 (4 files in 3 dirs): Refactor db.addChange parameters ...
19:22 buildbot: 07master 03Dustin J. Mitchell * 56d87cb (10 files in 2 dirs): Update change sources to use the new addChange parameter names ...
19:22 buildbot: 07master 03Dustin J. Mitchell * a311234 (25 files in 10 dirs): Merge branch 'bug1905' ...
19:22 buildbot: 07master commits e095585...a311234 -
19:22 github-buildbot left #buildbot
19:27 catlee pickles are evil
19:46 dwlocks joined #buildbot
19:58 mjw left #buildbot
20:03 djmitche indeed
20:03 but we need to verify that the upgrades work and continue to work
20:06 catlee yeah
20:07 mattn joined #buildbot
20:07 catlee I looked at adding slots to Change again
20:07 you can't unpickle old Changes without changing __init__
20:09 djmitche that sucks :(
20:10 maruel catlee: you wanted to add slots to reduce memory usage or another reason?
20:10 djmitche memory usage
20:10 we have ridiculous memory usage right now
20:10 did you see #1941?
20:11 bewst Is there some good reason that the properties interface is not a more direct echo of that of dict?
20:11 weasel ayust|home: the scary part is that google finds it when one looks for your string of letters.
20:11 bewst This is why I ask:  cc ='cc', '')
20:11 djmitche the default parameter in particular?
20:12 bewst If I want a default, I can't just use p.get(k, d)
20:12 so I end up with stuff like the above
20:12 mattn left #buildbot
20:13 djmitche it stores more than just k:v, so it needs some extra interfaces
20:13 but it's not that well-designed in general :(
20:13 I'm happy to add a get() if that doesn't exist
20:13 bewst Sure, but is there any reason it couldn't have all the usual interfaces?
20:13 djmitche nope
20:13 bewst it doesn't.
20:13 djmitche k, feel free to patch that
20:13 bewst Is there any reason it couldn't derive from dict?
20:14 djmitche because it stores extra stuff
20:14 bewst That's allowed
20:14 at least, it was the last time I looked
20:15 djmitche if it works, I'm game
20:16 bewst
20:16 maruel bewst: it's hard to derive from dict correctly. It's better not.
20:16 bewst maruel: what's the problem?  I've done it often
20:16 djmitche that's my impression, too
20:17 if you don't override every method, then you have methods returning internal details
20:17 bewst ??
20:17 djmitche if you store k : (value, source, runtime) in self
20:17 and you miss e.g., __getitem__
20:17 bewst dict's __getitem__ does that job
20:17 djmitche then props['foo'] will not return 'bar', it will return ('bar', 'somesource', True)
20:17 which is not what you want
20:18 bewst Oh, I didn't realize you were storing extra stuff per key
20:18 djmitche yep
20:18 so feel free to add methods, but let's not refactor
20:18 bewst that's easy, you just store another dict inside it
20:18 maruel bewst: I'm mostly warning about adding subtle bugs, I don't mind much.
20:19 bewst maruel: I prefer something small that works to a whole bunch of forwarding boilerplate that's easy to get wrong
20:19 djmitche maruel: I agree - this feels like a case where I'd merge a seemingly innocuous patch and then spend a lot of time trying to solve a subtle problem later once it's released
20:19 bewst better not to fix this problem, then ;-)
20:19 djmitche sure
20:20 my risk aversion is annoying, sorry about that :)
20:21 bewst Sorry.  It's just that your "unmaintainable" looks way less maintainable to me
20:21 And it's hard for me to write code I don't believe in :-)
20:22 So someone else should do that job
20:45 darshie I have a question about server names: basically I want to use static IP for my master and slave; so when i want to point the devs to the buildbot, should i give them the name of the slave or the master?
20:46 from what i understand is the master that creates the webserver where the waterfall runs, right?
20:46 so I should point them to the master?
20:47 djmitche right
20:47 darshie awesome :) thanks!
20:48 always better confirm instead than assume :)
20:55 _wms joined #buildbot
21:02 dbs joined #buildbot
21:02 dbs left #buildbot
21:02 dbs joined #buildbot
21:15 dbs left #buildbot
21:25 armenzg left #buildbot
21:36 tomprince djmitche: I remain convinced that moving the slave logic to master would not add any signifigant work to in3xes project.
21:37 djmitche tomprince: it might even make it easier
21:37 let's see how it goes when he gets started (May 12th)
21:37 tomprince and bewst: I have a protyotype runnig on the server in my git-demo branch.
21:38 bewst a prototype of what?
21:39 tomprince doing the slave logic on the server.
21:39 s/slave/source/
21:40 darshie not sure what i am doing wrong here, but i cannot trigger anymore email on failed builds
21:40 bewst sounds interesting, but I don't know what it means!  Tell more?
21:40 darshie my multi steps builder is running, and if one of the steps fails, the whole build is marked as fail
21:40 but it does not trigger the mail notifier
21:40 and i cannot understand why
21:41 i didn't touched the mail notifier since months
21:41 and in the logs i do not see any call to the mailnotifier
21:41 (usually i would get an error if something would go bad, but this time i get nothing)
21:42 and the code is the same used for the example of the mailnotifier, where it shows the last 80 lines
21:42 tomprince bewst: You and djmitche where discussing that slaves needed updating to change the source step logic, or add new vc systems.
21:43 I wrote a proof-of-concept git source step that has all the logic on the server.
21:43 bewst Does it in terms of ShellCommand?
21:44 tromey left #buildbot
21:44 tomprince Yes.
21:44 no, RemoteShellCommand
21:44 bewst Haha, and the difference between those two is...?
21:44 tomprince[…]82f3dcb2b81435618
21:44 verm__ i had always assumed that ind3x was going to be moving the source steps to the master
21:45 what's the point in redoing them and leaving them on the slave? it makes no sense
21:45 tomprince verm__: That was my reasoning.
21:45 verm__ and i agree with djmitche that it may even make the project easier
21:45 bewst tomprince: this is really cool.  would love to chat more but will have to go very soon. thanks for showing it, though
21:46 verm__ at least i don't see any reason why it would add any complexity
21:46 tomprince The difference that ShellCommand is a BuildStep and RemoteShellCommand is a RemoteCommand.
21:50 darshie nm, found the issue :) pointing to a wrong category
21:51 verm__ tomprince: that single example should be enough to go on i think
22:04 monokrome left #buildbot
22:07 lordross left #buildbot
22:16 dwlocks I keep having problems with .8.3p1 where changes are being passed around as a NoneType object.
22:16 so when some unassuming code tries to get... oh say, a branch, or properties, there's an exception in th elogs.
22:17 catlee dwlocks: maybe problems with the change horizon?
22:18 dwlocks people have mentioned that...  I used to have that in my config.  I'll double check.
22:18 _wms_ joined #buildbot
22:18 dwlocks the config has "c['changeHorizon'] = 20"
22:19 quotes are mine.
22:20 _wms left #buildbot
22:20 _wms_ is now known as _wms
22:20 dwlocks in the sourcestamp, I added changed line 69 from "if changes:" to "if changes is not None:"
22:20 now I'm on another error where changes is None.
22:22 catlee so changeHorizon=20 means that buildbot will delete all but the latest 20 changes
22:22 so any time it has to re-create a sourcestamp from the db, you could end up with missing changes
22:22 could try bumping it up to a higher number or disabling it completely
22:23 djmitche dwlocks: is changes None, or is changes [ None ]?
22:23 catlee (my knowledge is based on 0.8.1, so the situation may be different in your version)
22:23 dwlocks oh... when people said problems with changeHorizon, I thought they meant *missing* changeHorizon.
22:24 djmitche: I also added if changes[-1] is not None:
22:24 djmitche ok
22:25 jhegge left #buildbot
22:39 dsal left #buildbot
22:47 dwlocks is it recommended to have one poller for each project?
22:47 that seems wasteful of resources, but I'll do it.
22:48 in a case where I'd be watching multiple projects in the same svn repository.
22:54 djmitche no, not necessarily
22:54 I think there's a parameter to SVNPoller that can figure out the project from the path, right?
22:56 dwlocks hmm, the inline docs just say project can be used as a filter.
22:56 it's a runtime string variable passed to __init__
22:57 there's nothing obvious that says "here's how you set project from the url"
22:58 and place I can see to have project vary.
22:58 djmitche it's used as a filter for the scheduler
22:59 dwlocks I just found that.  :-)
23:00 djmitche hm, maybe there's no such parameter
23:00 I thought the file_splitter was upgraded to support that
23:00 ok, well, that's a good bug (and an even better patch)
23:01 dwlocks what's a good patch?
23:01 I upgraded *my* file splitter to return a project when I wrote SVN_multi_project
23:02 djmitche yeah, so fixing that in the upstream SVNPoller would be a good patch
23:02 but at least file a bug :)
23:03 dwlocks but I had to change a bunch of stuff to make it work, returning 2 vs 3 things was not so great.
23:03 what about allowing project to be a list of strings instead of just a string?
23:08 actually... it looks like split_file was subtly changed such that it could be used as a filter.
23:08 sorta.
23:10 the problem is the change.project is always set to self.project.
23:11 if you use split_file to filter based on project, you risk losing the original project.
23:12 djmitche you might need to add a new parameter like split_file_with_project
23:12 and if the user only specifies split_file, then just use self.project
23:12 dwlocks yeah.  what I did would break old configs.
23:12 djmitche that's probably the easiest way to maintain backward compatibility, rather than trying to get split_file to support both old and new behaviors
23:13 you might also make the new split_file_with_project (or whatever you call it) return a dictionary instead, so that we can add keys to it later
23:14 dwlocks I'll call it "split_url" perhaps?
23:14 that could be used by other SCM's then, right?
23:15 do all pollers have a split_file like function?
23:19 djmitche I think just SVNPoller
23:19 but yes, split_url sounds good
23:23 _wms left #buildbot
23:34 verm__ djmitche: on BuildStatus, finish_watchers, results, updates, watchers, test_results
23:34 none of these need to be stored in the db correct?
23:34 djmitche results does
23:34 and test_results
23:34 verm__ the pickles store everything it took a while to sort through what did and did not need to be stored
23:34 really? hmm
23:35 oh sorry results shouldn't have been in that list
23:35 testResults though.. i have to look and see what that is again..
23:36 that's not related to the python tests step is it?
23:37 djmitche yes, and subunit
23:37 I don't know much more than that :/
23:37 verm__ right, that should go into it's own table then
23:37 no, that's fine
23:38 does metabot use it?
23:45 djmitche not sure
23:46 verm__ i'm going to leave it out until i get an example then
23:47 djmitche k - make a note, though
23:47 verm__ already did
23:47 i'll post this in an hour or three
23:48 djmitche k
23:48 yay, code!

← Previous day | Index | Server Index | Channel Index | Today | Next day → | Atom Feed | Search | Google Search | Plain-Text | plain, newest first