18:30:18 <sumpfralle> #startmeeting
18:30:18 <MeetBot> Meeting started Wed Aug  7 18:30:18 2019 UTC.  The chair is sumpfralle. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:30:18 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
18:30:33 <sumpfralle> Welcome to our weekly IRC meeting about munin!
18:30:40 <sumpfralle> #chair TheSnide h01ger
18:30:40 <MeetBot> Current chairs: TheSnide h01ger sumpfralle
18:38:05 <sumpfralle> My review of the last week: I did some packaging for Debian and a bit of style improvements in munin-2.0. Additionally there were a few pull requests for the contrib respository to be discussed.
18:38:39 <sumpfralle> Right now I am trying to fix #1190 - this will hopefully resolve other cdef-related issues, as well.
18:39:55 <sumpfralle> This part of munin's code feels really a bit ugly. I would be very impressed and grateful, if TheSnide cleaned up the mess for munin 3.0 ...
18:40:31 <sumpfralle> I am trying to decompose the relevant function a bit in order to make it testable.
18:40:56 <TheSnide> hi
18:41:15 <sumpfralle> hello!
18:42:00 <TheSnide> about #1190, does it work in 2.999 ?
18:42:30 <TheSnide> (as, this code part has been rewritten)
18:42:37 <sumpfralle> indeed, I did not try that.
18:42:56 <sumpfralle> I will do it.
18:43:08 <TheSnide> (but, i admit I didn't test it)
18:43:18 <sumpfralle> I tried to check, if/where you rewrote that part, but I lost track in the git history, somehow.
18:43:39 <sumpfralle> Maybe I will ask you some questions later ...
18:43:40 <TheSnide> most of munin-update & munin-graph got rewrote
18:44:01 <TheSnide> node & async wheren't
18:44:08 <sumpfralle> The "cdef/negative/..." handling is really screaming for small unittests. Maybe I can come up with some suggestions for 3.0 later ...
18:44:33 <sumpfralle> It is great, that you tackled that beast!
18:44:50 <TheSnide> "screaming for small unittests" not really. We should come up with a fully deterministic test node
18:44:53 <sumpfralle> Even if it should have some flaws now, it is surely more maintainable, than the old code.
18:45:09 <sumpfralle> "fully deterministic test node"?
18:45:20 * TheSnide is thinking about that since a while now, but... time.
18:45:39 <TheSnide> sumpfralle: yes. Something that emits a "fixed" output.
18:45:40 <sumpfralle> With unittests I meant lots of small tests for different variations of cdef definitions and so on. These would be nice input->output-tests
18:46:12 <TheSnide> well, if you mean 'unittest' done via node-test+munin-update, then i'd be ok.
18:46:29 <TheSnide> I really try to avoid testing directly the classes.
18:46:55 <TheSnide> as, it looks very easy at first. But then lack of time kicks in to maintain them.
18:47:12 <sumpfralle> For this specific detail, I think a direct test of one specific processing is more suitable. Somehow on the level of "does this sorting algorithm sort all items in different variations properly".
18:47:37 <TheSnide> maybe. I'm not concinved, but I won't stop you
18:47:44 <sumpfralle> Integration tests are nice and way more important. But tricky and hard-to-debug details can be suitable targets for small tests.
18:47:46 <sumpfralle> good
18:48:54 * TheSnide is aiming to write a small test suite that will be _super_ easy to understand and extend. So you can write them yourself
18:48:57 <TheSnide> )
18:48:58 <TheSnide> ;)
18:49:10 <sumpfralle> that would be awesome!
18:49:34 <TheSnide> indeed
18:51:31 <sumpfralle> Do you have anything else to share? Crazy ideas, beautiful experiences? Otherwise I would jump back at the cdef monster ...
19:01:09 <sumpfralle> #endmeeting