Time Nick Message 03:08 tyler-2 Hey folks... I have a network with basically no Internet... And I'm setting up a local Minetest server for the kiddoes. I have my DNS server handing out a local IP for `servers.minetest.net` so that local clients don't have to make changes. And I've got `minetest/master-server` running locally too. Oddly this all works great, and clients see my servers. But only for 5 minutes... Looking at Apache log 03:08 tyler-2 s it seems like the servers are sending START announces, but not UPDATE. I dropped a silly PR to get Minetest to actually say when it was trying to do an ANNOUNCE. I expect if there were a problem with UPDATE announces, I wouldn't be the first to notice. The only thing that comes to mind is whether `dtime` works on a different time-scale or something on a Raspberry Pi... 03:08 tyler-2 I am trying to drop a PR that logs when `announce` happens for UPDATEs and DELETEs as well as STARTs. 03:09 tyler-2 But... I don't speak C good and I'm already struggling with string concatenation. But was curious if anyone had a "Oh yeah there's a problem with the server announce code..." moment. I don't see anything conditional to it other than the `dtime` ticks. 07:39 hecks https://a.cockfile.com/VtdDWk_1k.png 12:36 tyler-2 Dropped a PR that might be helpful in troubleshooting... But... (shrug) i'm mostly clueless. 12:36 tyler-2 https://github.com/minetest/minetest/pull/10177 12:38 sfan5 might be useful for debugging but having those log messages every 5 minutes would be annoying in regular use 12:48 tyler-2 Yeah, if your server was empty your log would just be full of that. 12:49 tyler-2 Don't suppose there's a debug loglevel or something that would _already_ show when this is happening? 12:51 sfan5 adding another log message to infostream would be fine 12:51 sfan5 since that's not visible by default 13:26 tyler-2 So actionstream instead of infostream/ 13:26 tyler-2 Er... infostream instead of actionstream? 13:27 sfan5 yes 14:40 tyler-2 Alright, maybe fixed that PR. Is there a way to get a running server to dump its dtime? 14:42 tyler-2 I had set my time scale to 1, with the intention of making the day/night cycles the same as the real world... I'm curious if that also impacts this `dtime` var that is used to trigger the server announce every 300 secs. 14:44 sfan5 it doesn't 18:51 rubenwardy #9995 18:51 ShadowBot https://github.com/minetest/minetest/issues/9995 -- ContentDB: Add Update All and download queuing by rubenwardy 19:03 rubenwardy I'd like the ability to release patch releases for Android only, with less bureaucracy 19:05 sfan5 we could do patch releases without announcing them separately 19:05 sfan5 (adding the windows builds really isn't that much work) 19:05 rubenwardy would they still be tagged? What branch? 19:06 sfan5 definitely, commits would be on stable-5 19:06 sfan5 but must also exist on master 19:06 Krock backport branch & PR 19:06 sfan5 that also works 19:06 Krock like done in 0.4.17(.1). I think that would be the sanest approach to keep track of the changes 19:12 sfan5 what I mainly meant by "announcing them separately" is: just edit the existing forum post, perhaps no changelog, don't tweet about it etc 19:14 Krock changelog would still be helpful, even if just a footer note 19:15 Krock I'd say that's a general rule whenever the version number increases... if it's trivial for Android to not increase the number (sneaky patch) then it's probably fine 19:17 rubenwardy the overhead of releases are: 19:17 rubenwardy - feature freezes (not a problem with patch releases) 19:17 rubenwardy - making all the builds (I'd really like this to be done using GitHub actions at some point) 19:17 rubenwardy - updating things like PPAs and F-Droid (doesn't need to be done if desktop isn't changed) 19:17 rubenwardy - announcements 19:19 Krock so sneaky patch it is