Time Nick Message 12:54 rubenwardy merging in 10: https://github.com/rubenwardy/minetest/commit/d83fe16a29c231491183c890c5378c671525fc8e 12:54 rubenwardy !title 12:54 ShadowBot Fix macro warning due to incorrect define conjunction · rubenwardy/minetest@d83fe16 · GitHub 12:55 nore rubenwardy: lgtm 13:05 Krock [Autotimer] RRRIIINNNGGG rubenwardy 13:06 rubenwardy done 13:36 fwhcat hey, it would be nice to have a LOD on far objects so when I press R it wouldn't lag so much ?! 13:36 fwhcat is there any plan to implement such things? 13:38 Krock "Lord Of Destruction"? 13:38 Krock top result of abbreviations.com 13:38 rubenwardy fwhcat: farmap 13:38 rubenwardy it's a WIP feature by celeron55 13:38 rubenwardy well, it's not being worked on anymore 13:38 fwhcat nice. 13:38 rubenwardy #3502 13:38 ShadowBot https://github.com/minetest/minetest/issues/3502 -- Far map and improved map transfer and rendering (WIP) by celeron55 13:39 fwhcat I was thinking also about the client cache 13:39 rubenwardy previous versions of Minetest had something like this, but it didn't scale very well 13:39 fwhcat because everytime you disconnect, the map is totally reloaded from server 13:39 Krock tried it once - worked quite well despite funny looking mapblocks 13:39 fwhcat it would be nice to look at the cache first and compare to some kind of hash of the mapblocks 13:39 fwhcat to know that you have locally the "latest" version or not 13:40 fwhcat like a git commit xD 13:40 Krock well, it has to free the memory in order to have resources free on the next server 13:41 fwhcat yes for the memory, but isn't the map stored in cache files also ? 13:42 Krock no, that's only the media like textures, sound files and models 13:43 Krock however, you can save the map locally but that won't be used for the next rejoin 13:48 rubenwardy #7891 13:48 ShadowBot https://github.com/minetest/minetest/issues/7891 -- Fix ContentDB packages timing out by using download_file instead by rubenwardy 14:15 Krock how's #7892 high priority AND blocker? neither of them 14:15 ShadowBot https://github.com/minetest/minetest/issues/7892 -- Deprecate modpack.txt and use modpack.conf instead by rubenwardy 14:15 rubenwardy well, maybe not blocker 14:15 rubenwardy without that, ContentDB doesn't install modpacks correctly 14:17 Krock ah, I see 14:17 numzero Is that the cause of #7370? 14:17 ShadowBot https://github.com/minetest/minetest/issues/7370 -- Browse online content: Mods can't be deleted or renamed 14:17 rubenwardy simply because ContentDB needs to keep some tracking information in content dirs 14:18 rubenwardy ah yes, maybe 14:18 Krock -> result.package.type == "modpack" why not introduce a new type? 14:19 rubenwardy because ContentDB doesn't have that type 14:19 rubenwardy that would probably be a lot more consistent 14:19 rubenwardy maybe I'll do that, and just make the "mods" link actually do a mod and modpack union 14:19 rubenwardy basically, modpacks are mostly used for organisation purposes 14:20 rubenwardy as compared to minecraft, where modpacks are a catered collection of mods 14:20 rubenwardy so I didn't want to separate mods and modpacks in the UI for contentdb 14:21 rubenwardy It depends whether you want content and package consistency, or UI and thing consistency 14:21 rubenwardy tbh, this code has descended into a whole bunch of hacks 14:23 Krock modpacks themselves are yet not friendly due to name collisions 14:24 rubenwardy yeah 14:29 rubenwardy also, the issue with adding a modpack CDB type is that you have to trust users more with choosing the right type 14:29 rubenwardy or you have to have code to work it out 14:29 rubenwardy on the forums, modpacks are commonly called mods 14:30 rubenwardy which is correct from a user point of view really 14:36 rubenwardy so, T4im, how do you profile the profiler? 14:37 T4im for one it instruments the globalstep used for sampling like any other globalstep 14:37 T4im and then it instruments an empty function to estimate overhead 14:37 T4im instrumentation overhead that is 14:38 rubenwardy yeah, just found that 14:39 T4im it's a hybrid between an instrumenting and a sampling profiler; instead of sampling data directly, it samples instrumentation collected data; keeping the instrumentation lean yet the sampling precise 14:39 T4im i guess that's important to keep in mind when reading the whole thing :) 14:41 * rubenwardy checks out the branch 14:42 ANAND What's the best way to test PRs? I download the fork, checkout the branch, and compile. Is there a better, more efficient way of doing this? 14:43 T4im depends on the size of the PR, if it's small you could probably just download the patch and apply it locally 14:43 rubenwardy ANAND: git hub 14:44 rubenwardy hub.github.com 14:44 rubenwardy then git checkout https://github.com/minetest/minetest/pull/4343 14:44 rubenwardy https://hub.github.com 14:44 ANAND PRs can be checked out directly? 14:44 T4im oh interesting tool 14:44 rubenwardy with hub, yes 14:44 ANAND Cool 14:44 rubenwardy you can also do git clone awards to clone one of your own repos 14:45 rubenwardy no need to go to the page and copy the ssh url, because you can never remember 14:45 rubenwardy and also git clone minetest/minetest for someone else's 14:45 ANAND Nice! Thanks :) 14:45 T4im and available in the debian repos as "hub" nice :3 14:48 rubenwardy You inconsistently use . and : in that PR 14:48 rubenwardy and also function Class:thing() vs Class.thing = function(self) 14:48 rubenwardy That's a fairly minor nit pick though 14:52 p_gimeno gosh, GIT does that without the need for hub. Just add the repository of the PR creator as remote, and fetch it. Then you have the branch the PR is made from. 14:53 Krock download the patch. git am -3. another possibility 14:54 rubenwardy p_gimeno: that takes longer though 14:54 rubenwardy patchs and git am also always fail 14:55 p_gimeno rubenwardy: longer than re-cloning the whole repository?!?! 14:55 Krock they fail when the PR needs rebase 14:55 rubenwardy nope 14:55 rubenwardy hub doesn't reclone the whole repo 14:55 rubenwardy it adds a remote for the pull author, then fetches and checkout the branch 14:56 p_gimeno rubenwardy: hub seems like a convenience tool for those who can't use git properly 14:57 rubenwardy no, it's a tool for people who don't won't to get the remote URL for every single contributor there is, or type it manually 14:57 rubenwardy it saves me time 14:57 rubenwardy *want 14:59 * p_gimeno bites his tongue 14:59 p_gimeno also, there was some discussion earlier about modpack name collisions, exactly what #6898 was intended to address (it's still pending in https://notabug.org/pgimeno/minetest/pulls/1 and since it's a breaking change, if it doesn't make it to 5.0.0 it will have to wait for 6.0.0) 14:59 ShadowBot https://github.com/minetest/minetest/issues/6898 -- Allow distinguishing mods by modpack by pgimeno 15:01 p_gimeno ^ example of a PR for which hub won't help because it locks you in to github 15:48 ANAND Tested #7891. Packages are retrieved correctly *every single time*, so #7447 is indeed fixed by this. 15:48 ShadowBot https://github.com/minetest/minetest/issues/7891 -- Fix ContentDB packages timing out by using download_file instead by rubenwardy 15:48 ShadowBot https://github.com/minetest/minetest/issues/7447 -- Content store does not retrieve any packages 15:48 ANAND Thanks for the fix, rubenwardy :) 15:51 ANAND How exactly does this fix the issue though? Packages aren't fetched all-at-once now? 15:51 rubenwardy :) 15:52 rubenwardy ANAND: it downloads the file to the disk instead of fetching it into memory using curl 15:52 rubenwardy curl has a time out 15:52 ANAND Oh neat... 15:52 rubenwardy the better fix may have been to add an option to curl to increase the time out 15:53 ANAND Since the code has been moved into Lua from C++, won't efficiency take a hit? 15:58 Krock for those few microseconds I doubt it matters at all 15:59 rubenwardy Lua isn't that slow 16:00 rubenwardy attempt 2: #7896 16:00 ShadowBot https://github.com/minetest/minetest/issues/7896 -- Fix ContentDB packages failing to download by increasing timeout by rubenwardy 16:00 rubenwardy maybe another setting is warranted? 16:01 nerzhul lua is faster than PHP :p 16:02 ANAND Testing 16:03 p_gimeno luajit is faster than anything https://github.com/trizen/language-benchmarks 16:03 rubenwardy ANAND: make sure you get the right version 16:03 rubenwardy just pushed 16:03 ANAND Ok 16:03 ANAND hub is *very* convenient here... 16:05 p_gimeno ahh, MS's favorite strategy, to lock you in to using only their products, so they can abuse you later 16:06 rubenwardy hub was started before Minetest 16:06 ANAND rubenwardy: The packages' short descriptions aren't displayed... 16:06 rubenwardy oops 16:06 rubenwardy forgot I changed that 16:06 ANAND Oh, lol :P 16:07 rubenwardy updated 16:07 ANAND Re-testing 16:08 p_gimeno rubenwardy: timing doesn't matter here, it's still something that MS will take LOTS of advantage from, regardless of who started it 16:10 rubenwardy the short description thing was because I changed CDB's API an hour or so ago 16:18 ANAND rubenwardy: I got a couple of timeout errors downloading screenshots 16:19 ANAND Only once, though. 16:19 ANAND https://github.com/minetest/minetest/pull/7896#issuecomment-441452303 16:20 ANAND The UI is very unresponsive for quite some time after the content store is opened. 16:21 ANAND The time taken for the content store to open since the button press is considerably longer than the corresponding time taken using the previous PR. 16:21 ANAND By 2-3 seconds at times 16:22 ANAND at all times* 16:23 rubenwardy so, C++ is slower? 16:23 ANAND I have no idea... :/ 16:24 ANAND Taking into consideration the end-user experience, the first PR seems to be better, IMHO. 16:24 ANAND It feels much snappier 16:28 rubenwardy have many times have you compared them? 16:28 rubenwardy because it could just be an inconsistent network 16:29 rubenwardy also, try again 16:29 rubenwardy I've changed CDB to return https links 16:29 rubenwardy which should cut out a request 16:30 ANAND rubenwardy: 10 times for each PR 16:31 ANAND Compared to the first, the second feels much slower to open, and unresponsive after being opened, all 10 times. 16:32 rubenwardy the issue is that the older PR makes more work down the line, if I ever want to move the code all into C++ 16:32 rubenwardy I guess that work has already been spent in making that PR 16:33 rubenwardy the alternative is to just not move the code into C++ 16:33 ANAND Good idea 16:33 rubenwardy and make the CLI package manager a completely separate program 16:33 ANAND the CLI what? 16:33 ANAND Sounds interesting... 16:33 rubenwardy command line interface package manager 16:34 rubenwardy so server owners can install from CDB 16:34 ANAND Neat! 16:35 rubenwardy I can't really reproduce this issue 16:35 ANAND The original packages-not-being-fetched-at-all one? 16:41 rubenwardy from 6 samples each, the Lua parsing method is 16% faster 16:41 rubenwardy so yeah, you're right 16:41 rubenwardy this is from copying from C++ to Lua, probably 16:49 ANAND rubenwardy: How'd you measure the difference accurately? 16:49 rubenwardy manual timing 16:49 rubenwardy lol 16:49 rubenwardy not very reliable 16:50 ANAND Oh, lol 16:50 rubenwardy there is a better way to do it 16:51 ANAND Profiler? 16:51 rubenwardy manually profiling it by having hooks around the download 16:51 ANAND Ahh.. 23:53 paramat Shara any opinion on game#2260 ? 23:53 ShadowBot https://github.com/minetest/minetest_game/issues/2260 -- Improving the glass door handle 23:54 paramat and whether to do for obsidian glass door too 23:58 Shara paramat: sorry, but though I'm not going to block it, it's not a change I like. I prefer glass door as it already is 23:58 paramat ok 23:59 Shara I understand why you want to change it and the logic is sound but... I just don't like how it looks