[15:24:38] Urbanecm: perhaps I can update the the latest php versioning [15:24:47] and re-test [15:25:13] Hauskatze, ftr, I have php 7.2 and the script works even on a fresh clone. [15:25:26] my clone is fresh as well [15:32:23] Hauskatze, can you double check installation requirements are met on your system? [15:32:52] Urbanecm: I'm updating my php and will recheck [15:33:12] Okay [15:34:27] so far everything listed there I had [15:37:19] Hauskatze, okay. Collator's built in PHP, see https://secure.php.net/manual/en/class.collator.php, I think missing something required is causing this problem. [15:44:06] okay so I have now PHP7.3.0 but Fatal error: Uncaught Error: Class 'Collator' not found [18:37:42] Do you know about "editingRestrictionWidget is null" from new blocking system? This error is on wikis, where it is disabled. [19:12:49] how do I get a copy-paste-able list of all sites in a category? [19:19:23] Nudin_WMDE: how about https://en.wikipedia.org/wiki/Category:Readability_tests?action=raw that does not include subcategories though [19:20:55] that gives the text of the category-page not the list of pages in it [19:21:04] Petscan? [19:24:23] yeah, Petscan -> csv -> Libreoffice does the trick (Petscan directly gives you also some meta-infos) [19:47:38] eh.. but the text of the category-page is the list of pages in it [19:48:19] oh.. it's not , nevermind [21:28:16] Hauskatze: Collator comes from the intl extension [21:29:09] legoktm: thanks - an user in Conpherence helped me enable the intl extension, but still mw-core/maintenance/updateCredits.php does funny stuff [21:29:26] oh? what's it doing? [21:30:10] legoktm: for me, https://phabricator.wikimedia.org/P7953 [21:30:35] it outputs the contents on the bottom of the file, not respecting headers [21:30:39] o.O [21:30:50] and removes authors as well [21:30:58] or rather, does not list them all? [21:31:13] if I copy-paste the list to replace the current one, some are removed [21:31:22] sooo... I don't know what's going on [21:31:44] Urbanecm seems to have run it successfully so this is likely my computer or general uselessness [21:32:38] I can confirm what Hauskatze says [21:36:35] Hauskatze, I'm pretty sure you tried that already, but can you double check your repository is up2date? Nothing but an in-repo change comes to my mind... [21:37:09] Urbanecm: yes, my repo is up to date, but I can git reset --hard origin/master && git pull origin master again I guess [21:38:49] okay, good. [21:40:56] I don't see any obvious reason why that would happen... [21:43:10] Urbanecm: clean copy, ran the script, same issue [21:43:38] what do you mean by clean copy? clean copy of what? [21:43:46] the repo [21:43:57] up to date [21:44:28] aha. Even it's up to date, can you please try to clone it again (remove the clone and run git clone (...) again)? [21:44:55] that can take some time :) [21:44:59] but okay [21:45:08] that's true, but I don't see anything else you can do :( [21:46:32] I'll try that tomorrow if I have time [21:47:04] up to you [21:47:15] I'm going to bed, see you later [21:47:30] Sleep well. It's probably some Windows stuff [21:47:49] Do a shallow clone then [21:48:13] git clone --depth 1 [21:48:36] eh, if a repo is up to date, it's up to date [21:48:43] not much point in rebuilding it [21:49:28] tgr, well, no idea how I got into that state, but I had a state when local commit didn't match gerrit patchset and git review -R said "no new change". [21:50:24] git repos use a hash tree similar to blockchains, if two repos have the same commit in HEAD, either their full history is identical or one of them is inconsistent in a way that's trivial for the software to detect [21:51:18] or there was a collision (improbable, but can happen theoretically) [21:51:22] git review doesn't care much about the history anyway [21:52:12] only in a sense of "theoretically" that does not have much real meaning [21:56:41] git uses 160-bit sha1 hashes, so for a big repo with a billion or so commits the probability of collision is something like 1e-30 [21:57:00] being hit by a meteorite is a fair bit more likely than that [22:02:07] I admit it's improbable and probably not the cause here, but still, there's https://shattered.io/ [22:13:34] I'll just avoid that script Urbanecm & tgr [22:13:52] or all of them [22:15:56] breaking a cryptographic hash by exploiting weaknesses in the algorithm and the chance of random collision really have nothing to do with each other [22:16:28] and even so it involved Google computing over a quintillion hashes [22:16:58] a quintillion MediaWiki commits would be a nice problem to have [22:17:34] in reality we are probably somewhere in the million range [22:18:57] github reckons ~86K for mw core [22:19:03] (at least on master) [22:25:22] Hauskatze: re: the script, most likely you are using a shallow clone, in which case git log stops at your clone deoth [22:25:44] git log --oneline --no-merges | wc -l [22:25:44] 68652 [22:25:57] with merges is 86350 [22:26:43] plus 50 or so because I didn't pull first [22:27:30] in theory if it's a shallow clone there should be a file called 'shallow' in your .git directory [22:28:41] I guess the maintenance script could check for that [22:54:10] sorry tgr I was in a phone call with Melbourne, AU [22:54:27] no shallow copy afaics [22:54:35] it's a pure git clone [22:55:13] My guess it's Windows messing around or some missing php extension in addition to the intl one [22:55:17] idk [23:04:06] hard to imagine that being a php error [23:04:43] you should just check, though, the git command is git log --format="%aN" [23:05:00] you can count the result lines and see if it matches [23:05:16] tgr: about the missing authors, could it be that it does not fetch the SVN authors? [23:05:27] before we moved to git [23:05:47] because it's just a handful of them which ain't added [23:06:14] I don't think there is anything special about them, SVN revisions have been converted to proper git commits when the repo was migrated [23:06:40] you should just run the git log command and put it in a paste [23:07:02] once we compare it with a good result it will probably be obvious what's common in the missing authors [23:07:21] 86424 entries [23:07:35] sorting them A-Z [23:07:50] and now removing dupes [23:08:13] 675 unique authors [23:10:50] I get 86426, probably two commits just happened in the meantime [23:10:57] so your git setup seems fine [23:11:30] and 675 uniques as well [23:13:36] when I run the script it inserts 14 new names and does not remove any [23:13:43] ok so maintenance/updateCredits.php does funny things for me [23:14:49] on thing I can imagine is windows being crappy about large output from shell_exec and silently truncating it [23:15:16] I use git bash, and have tested it on powershell too [23:15:38] at least convertExtensionToRegistration.php works for me well as it's the one I used most [23:15:40] :) [23:15:57] sadly there are no "easy" conversions left (for me) [23:16:19] you could add a bunch of var_dumps and check the size of the $contributors array at each step [23:19:04] I'll just give up, tgr. I am just annoying people with this, which works for everyone else.