[00:59:37] brion: https://phabricator.wikimedia.org/T50377 Seems to be about the play button being too translucent, but it has the same background-image high contrast mode problem. Comment in that task, create a new task, or roll it up into a TimedMediaHandler-Player High Contrast mode improvements? [01:06:41] Roll em up, I'd say [09:39:09] Morning [09:39:28] Anyone here know anything about the PDF handler? [09:39:43] It doesn't seem to like - https://en.wikisource.org/wiki/File:Representation_of_the_Peoples_Act_1918_(ukpga_19180064).pdf [09:39:53] despite the PDF being perfectly readable. [10:03:27] ShakespeareFan00: do you have a question? [10:03:41] andre__: Yes [10:03:54] what is your question? [10:03:57] I have a PDF that's generating illegible images [10:04:07] I'd like to know why [10:04:34] Elsewhere it was suggested that the current PDF handler code can't cope with embbeded images as well as it might [10:05:33] The relevant file is :- https://en.wikisource.org/wiki/File:Representation_of_the_Peoples_Act_1918_(ukpga_19180064).pdf [10:08:07] ShakespeareFan00: The JPG thumbnail itself is 365x258 pixels and has a lot of whitespace surrounding it. On the File: page it is shown as 365x599 pixels though. [10:08:19] ShakespeareFan00: to answer the "why": likely because there is some software bug somewhere. [10:08:30] Already raised a Phabricator ticket [10:08:39] Also the small JPG may be a preview image... [10:08:56] Because when I load the PDF directly, I get a VERY llegibal and large image [10:09:53] so you already know everything I wrote here... [10:10:10] ...hence not sure why you asked and let me duplicate your info. [10:10:34] * andre__ goes to a better use of their time [14:52:16] hi! [14:52:27] just learned about cumin - i had some questions, is that okay? [14:52:56] anarcat: sure [14:53:18] hi! :) [14:53:20] hi :) [14:53:33] small world :) [14:53:41] eh :) [14:54:00] so i was wondering why cumin was written considering the alternatives like Ansible and Fabric that already existed (and were in Python, nothing less) :) [14:54:19] you can assume i know how Puppet works and have worked with Ansible and Fabric before, just to give you an idea [14:54:27] i also looked at the fosdem slides, so i have *some* idea [14:54:48] anarcat: ok, thanks for looking at it and the context [14:54:55] (e.g. maybe it's because cumin taps into puppetdb / openstatck / etc) [14:54:58] stack* [14:55:12] so, first thing first is to be able to query multiple/different backends and combine their results [14:55:23] short answer is https://phabricator.wikimedia.org/T143306 :P [14:55:29] hehe [14:55:43] well that would be a great addition to the RTD [14:56:05] Ansible because through their Yaml configuration files is not possible to achieve our use cases and their API are not intended for direct consumption. [14:56:08] i see [14:56:19] i was also frustrated by Ansible's YAML and poor documentation [14:56:27] that thing is, unfortunately, quite a mess [14:56:45] i mean people sure love YAML and do awesome stuff with it, but i'm a python programmer and would have loved to be able to program Ansible directly [14:56:58] yeah so the phab task linked is the analysis we did at the time [14:58:13] and we didn't find anything that fitted our needs, and also was not using paramiko ;) [14:58:57] i'm afraid to ask, but what's wrong with paramiko? :) [14:59:08] i've heard a few bad things with it, but i've always assumed it was fixable stuff [14:59:12] anarcat: btw the video of the fosdem talk is online, in case you're interested, the slides were pretty graphical so by themselves don't say much [14:59:28] yeah, i noticed but i am lazy :p [14:59:54] usually far behind security issues, and we're trying to follow the "don't re-implement your crypto/ssh/tls" approach ;) [14:59:55] i take it you explain that reasoning there? [15:00:01] eh [15:00:15] so cumin takes the "bundle a bunch of code and dump that over a ssh pipe" approach like Ansible? [15:00:43] no, so far we decided that we didn't want the capability to drop random code over the hosts, and preffered the approach [15:01:06] to just run ssh commands, if you want a more complex script is better to have it reviewed and deployed by your configuration management solution [15:01:13] and just run it via cumin for example [15:02:57] i see [15:03:02] so it's more like fabric, iirc [15:03:12] the two main "Features" of cumin are basically: be able to select dynamically from multiple sources of truth and mix them [15:03:42] like I want all the hosts that have this puppet class with that parameter, that are debian jessie, that in my inventory are marked FOO and that are depooled from cluster X [15:04:19] yeah, a mix between ansible, mcollective and fabric ;) [15:04:35] lol [15:04:43] interestingly, mcollective wasn't explicitly dismissed from that phab issue [15:04:48] (some) people use ansible for configuration management -- that's not the intention here [15:04:55] and the second feature is to execute the things in a controlled way [15:05:00] of course [15:05:07] i wouldn't use ansible for CM either :p [15:06:16] okay well that answers my questions - thanks! [15:06:32] i encourage you to write a summary of this somewhere, or at least a pointer to the phab issue in the docs [15:06:33] I can give you more details on teh "controlled" way [15:06:46] i would do it myself, but ... er... you know, the lazy part? yeah, that. ;) [15:06:59] it's pretty much all described in Cumin docs [15:07:06] (i would probably do it if it didn't mean registering on the wikimedia thing) [15:07:08] oh [15:07:09] it is? [15:07:17] and also in the wikitech page for the usage inside WMF: https://wikitech.wikimedia.org/wiki/Cumin [15:07:32] yeah well i read that wiki page [15:07:36] and perused the docs [15:07:45] i didn't clearly see a "cumin vs X" or "why we did this" [15:07:50] the phab issue is a great start though [15:07:55] yeah, that's a good point [15:08:11] "why not X" too :) [15:08:16] i mean it's one thing to explain "what this does"... [15:08:51] true, that is missing [15:08:55] so what are your use cases? [15:08:59] i mean the Features section in that wiki page is great, but then it begs the question "why not mcollective or ansible?" [15:09:03] paravoid: nothing at all :p [15:09:06] are you looking into alternatives? does cumin cover them, and if not, should it? [15:09:06] i'm just curious [15:09:28] i'm a dropout sysadmin and now i pretend i am one on LWN.net :p [15:09:42] e.g. https://lwn.net/Articles/744410/ [15:09:52] yeah, I've seen your articles [15:09:57] your blog post about kubecon too :) [15:10:53] cool [15:11:00] which reminds me, i need to put those two prom articles online now [15:14:07] anarcat: if you think it might be useful for you or is missing some features to be useful for you, let us know [15:14:20] cool, thanks! [15:14:22] it's always very useful to hear from different point of views and use cases [15:14:25] it fills an interesting space [15:18:03] anarcat: just out of curiosity, did you also look at the docs (in doc.wikimedia.org or readthedocs) or just that wikitech wiki page? [15:18:16] volans: i looked at both [15:18:23] volans: i found the wikitech page later [15:18:39] ok, just wanted to be sure the docs are easier to find than the wikiteck page :D [15:18:45] *wikitech [15:18:48] yeah, definitely [15:18:57] linked from github and the readme and all does it :) [15:19:27] hmm how important it's to fix those signatures (https://fi.wikipedia.org/wiki/Toiminnot:LintErrors/tidy-font-bug)? it's quite easy to fix using awb, but not sure is it really worth it [15:20:35] anarcat: also there is a quick demo on asciinema, the one I used in the FOSDEM talk: https://asciinema.org/a/PoV36MH4H85FsH6xLjHrMw7BN without the audio comment is a bit sterile and hard to get though