[13:33:27] DrTrigon: I'm online if you'd like to communicate real time [13:33:45] DrTrigon2: To fix the zxing/opencv issues [13:37:59] AbdealiJK: yes? [13:38:26] Would you like to debug the OpenCV/ZXing issues you're facing ? [13:39:13] Yes but currently I do not have access to a running instance... [13:39:53] ...in about 1h again. Can I help you that way? [13:44:39] AbdealiJK: I can use the labs instance so in 30mins I can connect, hope that helps? [13:47:13] Sure [14:19:15] AbdealiJK: are you still here? [14:20:16] Yep [14:20:36] so let's go as long as my connection lasts... ;) [14:20:41] what can I do? [14:21:43] AbdealiJK: ^^^ ? [14:22:11] Let's start from the beginning - Can you try reproducing the errors you had and check if they still exist ? [14:22:23] Because the docker travis does not complain about OpenCV anymore atleast [14:23:34] I reinstalled everything a few hours ago - before the conpherence comment - didi you change anything after that? Also not the dockers are still using the old file-metadata version. [14:24:09] I have not pushed to file-metadata in the past 3 hrs [14:24:34] What do you mean by the dockers are using the old version ? They are using the master from git [14:25:05] My last comment is about 2hrs ago, right? [14:25:26] Look into the docker log I see a lot of 9's there... [14:25:31] In conpherence, Yes ^ [14:25:50] DrTrigon: I think there;s some misunderstanding. [14:25:57] The git master will always have 9s [14:26:05] DrTrigon: The versions on pip will not [14:26:21] DrTrigon: That's because the git master is a dev version [14:27:17] And inherently it's set to dev9999999999 because dev3 < dev999... - i.e. the git version is more recent dev version than the pypi [14:27:21] DrTrigon: ^ [14:27:57] Docker used 0.1.0dev999999... as you can see at https://travis-ci.org/pywikibot-catfiles/docker-file-metadata/jobs/149453367#L3409 [14:28:15] so there should at least be a 0.2.0 now, right? [14:28:35] we need a way to be able to compare github and pip versions [14:28:47] I mean compare version numbering [14:29:08] No, there is no 0.2.0 [14:29:31] All the pip versions are temporary to test if zxing gets installed correctly from pip [14:30:13] THe 0.2.0.dev3 is a temporary version I created to test if zxing gets installed correctly if installing directly from pip. It will be deleted from pypi as soon as 0.2.0 is created [14:30:31] DrTrigon: Note than 0.2.0.dev3 is a lower version as compared to 0.2.0 according to python version spec [14:30:37] even in Job 54 I only see 0.1.0dev9.... https://travis-ci.org/pywikibot-catfiles/docker-file-metadata/jobs/149483655#L3380 [14:30:59] DrTrigon: That is because git will always has 0.1.dev9... [14:31:10] I've never pushed a commit where that version changes into git [14:31:37] this is extremely confusing to me - so as long as you know whats going on... fine ;) [14:31:52] :D Yeah, python version spec does that [14:31:54] However what to do with the errors mentioned? [14:32:14] Right [14:32:22] So, the zxing issue I'm looking into [14:32:32] I cannot reproduce the OpenCV issue anywhere now [14:33:03] Can I just do a git pull on labs to "refresh"? [14:33:31] On master - ye [14:33:33] yes * [14:34:22] Already up-to-date. [14:34:52] Ok, and do you still get the error when running the test ? [14:35:13] You can run only the opencv tests with `py.test -k ImageFileFaceHAARCascadesTest` [14:35:13] (collecting 3 items ...) [14:35:29] takes a bit time on bastion-03 [14:35:47] nod, because of the NFS [14:37:47] yes - but not suprising as nothing changed, right? [14:37:57] Do I need to "refresh" the pip install too? [14:38:18] May be you can give me exact instruction on how you installed it to labs? [14:39:34] AbdealiJK: ^^^ [14:39:46] How did you install the file-metadata i labs ? [14:39:48] in* [14:40:19] I had just used https://wikitech.wikimedia.org/wiki/User:AbdealiJK/file-metadata [14:40:34] https://commons.wikimedia.org/wiki/User:DrTrigon/file-metadata#wikitech.wikimedia.org_.2F_ToolsLab (see "latest pip version") [14:40:34] With the folder path in `pip install` step [14:41:26] No, with file-metadata==0.2.0.dev3 [14:41:39] (actually tried both ;)) [14:42:23] Seems right enough. [14:42:34] I can't see any reason for OpenCV to not be detected correctly [14:42:55] DrTrigon: ^ [14:43:25] agree, especially if it works for you - are you on bastion-03 too? [14:44:00] I can investigate on the OpenCV issue myself a bit - but with the java thingy I have absolutely no idea... [14:44:04] nod [14:44:08] I think I have a lead on the java. [14:44:18] ...may be possible that one solution fixes the other too...?! (*hope*) [14:44:21] It's a package name clash, but I need to investigate a bit more [14:45:03] Yep - verified. It's a package name clash [14:45:09] It won't affect users [14:45:43] ok, perfect - can we stop here for a moment - I'll soon come into big traffic - I will look into the opencv thing and report my findings. [14:45:55] Oh wait. [14:45:58] What mean's it won't affect users? The code does run? [14:46:09] So, I believe I've figured both issue out [14:46:15] Ping when you get out of traffic [14:46:17] ??? :)) [14:46:17] I'll explain [14:46:33] As long as we only need IRC its fine - wait a sec... [14:46:37] Or I'll leave you a elaborate note in conpherence as I may be gone for dinner by that time [14:48:08] back again. [14:48:17] Oh. that was quick [14:48:28] Let me begin with the opencv issue [14:48:30] as you wish ... now here or conph. both fine ;)) [14:48:45] Essentially, ToolsLab does not have OpenCV installed. [14:48:54] Hence, the OpenCV cannot be found :D [14:48:59] ;))) was my guess [14:49:12] how did you solve that? [14:49:18] I have some magic lines in my bashrc which uses conda [14:49:36] And I can install opencv inside a conda environment - which is what is being used [14:50:13] Apologies. It slipped my mind that I am using conda ... Hence it took me a while to realize this ! [14:50:19] can we make that the official way of installing? [14:50:41] ;))) getting older, ha? ;)) [14:50:45] Well, at the moment opencv is not a "required" dependency. Plus conda is bulky. [14:51:05] we'd have to get the users to install conda first. And virtualenv is already there ! [14:51:26] So, I think I'll add a "with conda" and "with virtualenv" section in wikitech [14:51:37] but we also need a better way of dealing with opencv [14:52:00] ? What do you mean [14:52:16] stating it is not required will not make users happy when they want to use it [14:52:37] Well, I think opencv has too many nuances to get it installed [14:52:47] And that is probably a discussion for a later time [14:52:58] I would appreciate the with conda docu. [14:53:13] nod [14:53:23] Now, the second issue is zxing [14:53:43] opencv: ok, but somehow we need to fix that - it was an issue back then and now again... [14:53:54] DrTrigon2: Yep. Hence I dislike opencv :P [14:54:12] ...like a curse. ;) reminds me on monkey island... ;) [14:54:24] sorry yes 2nd go on [14:54:28] I'm currently thinking if possible, it may be worthwhile to implement haarcascades in skimage [14:55:31] https://groups.google.com/forum/#!searchin/scikit-image/abdeali%7Csort:relevance/scikit-image/PIpELINpmek/VXXwT_FAGQAJ [14:55:55] Anyway, back to zxing [14:56:28] In setup.py, we download jar files from maven (Java's pypi) and save them inside the datafiles/ folder. [14:56:44] (sounds very nice - still if there is a new feature in opencv we might like to test or use it...) [14:56:56] nod [14:57:17] Now, when we do `pip install /path/to/file-metadata` the datafiles are stored in the /the/final/storage/locations/file_metadata/datafiles [14:57:58] WHen we do a `py.test` inside the file-metadata folder, it first files the `file_metadata` in the same folder (it is higher on the path as compared to /the/final/storage/locations/file_metadata) [14:58:04] Hence, the datafiles aren't being found. [14:58:36] so we need to test from the subfolder? [14:58:47] Nope, there is no clean way to test it [14:59:11] I think in such cases we can only recommend to do a `pip install -e .` or `python setup.py develop` before running `py.test` [14:59:24] check the path structure at the beginning of the test? [14:59:38] DrTrigon2: That sounds very hacky [14:59:50] ;)) agreee like pwb.py [15:00:01] Well, so what I meant by it does not affect users [15:00:46] So that would mean the bot script should work for me anyways... will check that! [15:00:57] Yep - it should work. [15:01:20] Question: can we ask labs to install opencv? [15:01:42] PS: Ensure that you're not inside the file-metadata folder, otherwise the clash will happen even when running the scripts [15:01:56] I believe we can. Although jayvdb would know the procedure [15:03:04] I guess file it on conpherence, resp. ask to put in a puppet config. At least back in the old days... [15:03:51] ok perfect and thanks for your time so far will check what you mentioned! [15:04:02] Yeah, they seem to still be using puppet. Not sure where the request should be made though [15:04:20] nod [15:05:00] Alright. Cya, going to go for dinner [15:05:10] enjoy!!!