2020-08-24T02:13:30 *** okurz_ is now known as okurz 2020-08-24T06:10:21 dns problem? 2020-08-24T06:33:07 dns problem? 2020-08-24T08:25:14 pjessen: I found a way to unfuck the database with encoding, but it's gonna be a lot of work 2020-08-24T08:33:35 lcp: no surprise 2020-08-24T08:36:18 not particularly happy about having to do it tbh 2020-08-24T08:37:03 yeah, nobody really wants top touch it 2020-08-24T08:37:18 i tried and wasxted many hours 2020-08-24T08:37:54 also it's not a 100% solution as far as I can tell, because it still leaves some unencoded things behind, so that's annoying 2020-08-24T08:52:42 *** Martchus_ is now known as Martchus 2020-08-24T08:58:36 lcp: have you considered leaving the current forums as a read-only archive and then a clean start with discourse ? 2020-08-24T09:08:25 I haven't yet 2020-08-24T09:08:29 I will let you know when I do 2020-08-24T09:09:05 our forums might be way bigger than expected by the import script tbh 2020-08-24T11:21:55 it's worth noting, there are some greek subforums that conflict by name, and I wonder if we could somehow remove them 2020-08-24T11:22:22 they don't exist publicly but I also don't see it from admin control panel 2020-08-24T11:23:23 nvm, I found them 2020-08-24T11:26:23 malcolmlewis: I assume that READY TO DELETE forums are actually something that needs to be deleted? 2020-08-24T12:13:48 https://gist.github.com/gschlager/bad57293004c0bfcddba might be helpful 🤔 2020-08-24T12:15:30 well, more or less ;-) 2020-08-24T12:16:16 our problem is that we have a "mixed" charset - the old half of a table has different encoding than the newer half 2020-08-24T12:16:42 while that gist works on whole tables 2020-08-24T12:16:57 I can no problem deal with utf-8 to windows-1525 and again utf-8, but latin-1 is the problematic factor 2020-08-24T12:16:58 I wish this database wasn't double encoded but oh well 2020-08-24T12:17:46 yeah, I honestly doubt the second half is recoverable by the looks of it tbh 2020-08-24T12:18:15 because it's filled with broken characters, and no wrongly encoded characters 2020-08-24T12:19:08 somebody forcefully encoded it as something else which isn't recoverable imo 2020-08-24T12:21:04 although maybe 🤔 2020-08-24T12:23:25 no, it's �, so doesn't seem recoverable 2020-08-24T12:24:15 IIRC it could be double-encoded utf8, but that's really IIRC - I looked at it several months ago 2020-08-24T12:25:04 this whole database is so unfortunate 2020-08-24T12:25:19 no objections on that ;-) 2020-08-24T12:26:08 when we moved the forums from Provo to Nuremberg, we decided not to touch the charset "fun" and to leave those old posts with broken charset 2020-08-24T12:26:17 maybe you should just do the same? 2020-08-24T12:26:21 I'm not giving up yet, but I'm not that far from it 2020-08-24T12:26:54 mainly because fixing stuff on importer side was easy, now I just have to deal with latin-1 stuff inside the db itself 2020-08-24T12:27:29 also yeah, we will need 2 days for the importer script, it's taking its time 2020-08-24T12:28:09 we will probably have a touch slimmed down db after removal of old subforums, but that's not a lot 2020-08-24T12:30:04 (also it's not possible to import this db without removing those subforums anyway, because it squishes subsubsubforums into subsubforums, and the names have to be different, which isn't the case for the TO BE REMOVED subforums) 2020-08-24T14:11:38 3 days, not 2 days 2020-08-24T14:11:52 that's a lot of things to import 2020-08-24T14:40:41 the import of posts will take 26 hours alone 2020-08-24T15:03:25 *** okurz_ is now known as okurz 2020-08-24T17:36:29 lcp, if they are ready to delete, I say go for it :) 2020-08-24T17:40:34 malcolmlewis: I mean https://forums.opensuse.org/forumdisplay.php/946-READY-TO-DELETE and its subforums 2020-08-24T17:41:08 still no vpn, huh? 2020-08-24T17:42:37 confirmed, my vpn connection is also broken (the certificate gets VERIFY OK: , followed by a "Connection reset") 2020-08-24T17:42:43 what happened to the vpn 2020-08-24T17:44:20 no idea, and I don't have a way to check without, well, vpn 2020-08-24T17:47:42 I didn't play with freeipa, so not me this time 2020-08-24T17:48:05 ;-) 2020-08-24T17:49:48 klein: can you please check/fix the vpn? 2020-08-24T17:50:54 I will probably be attempting another freeipa upgrade soonish 2020-08-24T17:51:38 pjessen: FYI: provo-mirror is still outdated (including /repositories). I'm basically running while true ; do rsync ... ; done since some days, but it looks like there's a big backlog, and I wonder if OBS builds faster than rsync can catch up ;-) 2020-08-24T17:51:58 so when the vpn is back, please check if the rsync push really works as expected 2020-08-24T18:12:43 cboltz: repositories should be pushed out as they change, maybe leave those till last 2020-08-24T18:13:00 sorry, didn't see your last msg 2020-08-24T18:16:22 last I checked on the push, it was fine 2020-08-24T18:17:11 given how many different repos my manually rsync already updated, I slightly ;-) doubt it's still fine 2020-08-24T18:17:55 for bonus fun, it seems some files let rsync "explode" and abort the sync 2020-08-24T18:18:54 I've seen inflate returned -3 (0 bytes) a few times 2020-08-24T18:19:08 and I've also seen some rsync: read error: Connection reset by peer (104) 2020-08-24T18:20:24 I can check the push, I think we keep the logs per server. not sure for how long. 2020-08-24T18:21:43 does anyone know whats happening to the vpn? this morning, mine was still okay, but DNS was down. then I ressytarted the vpn ..... 2020-08-24T18:26:03 cboltz: inflate - should we be running with compression? 2020-08-24T18:27:22 the rsync cronjobs on provo-mirror (and also my manually started rsync) use rsync -v -H --compress -ar0 --stats --timeout=21600 --no-motd --fuzzy --exclude=... --delete 2020-08-24T18:28:09 I wonder if that compress isn't superfluous - the rpms are already compressed 2020-08-24T18:28:20 good question 2020-08-24T18:28:29 the rpm payload is indeed compressed 2020-08-24T18:28:39 but I'm not sure about the header and especially the changelog 2020-08-24T18:29:00 hmm, dunno 2020-08-24T18:30:30 just tested: strings apparmor-utils-lang-2.13.4-703.1.noarch.rpm gives me the changelog in plaintext, which means it isn't compressed 2020-08-24T18:31:21 interestingly I first get a list of authors, and then the changelog content without the authors - no idea what the reason for this technical detail is 2020-08-24T18:34:39 apparmor-utils-lang-*.rpm is 84 kB, and rpm -q --changelog apparmor-utils-lang | wc -c says ~55 kB, so compression could make sense