Minetest logo

IRC log for #minetest-dev, 2022-01-30

| Channels | #minetest-dev index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:08 HuguesRoss sfan5: Regarding #12011. Despite my title I'd trying not to involve myself much in design / management discussions for the engine. With that said, yeah it may be better to nuke. I don't think the argument about new games holds much water, because it only really applies to multiplayer games that were developed since this PR *and* haven't launched
00:08 HuguesRoss yet (as far as I understand this code is client-side, the server couldn't have blocked older clients from seeing coordinates)
00:08 ShadowBot https://github.com/minetest/minetest/issues/12011 -- Grant `basic_debug` privilege by default
00:08 HuguesRoss ...in which case, is there harm?
00:08 sfan5 harm in nuking it? don't think so
00:09 HuguesRoss then I'd say we may as well, and re-implement differently for the next release
00:10 Wuzzy is there a point that there is a "Page X of Y" in the MTG book read interface?
00:10 sfan5 so you know the page number?
00:10 Wuzzy To me, it's always 1 page no matter how long the text
00:10 sfan5 oh
00:10 sfan5 ¯\_(ツ)_/¯
00:10 Wuzzy and the single page shows the whole text 😆
00:10 Wuzzy it works its just weird
00:11 MTDiscord <Jonathon> probably related to the fact you cant limit a text input field?
00:11 MTDiscord <Jonathon> idk
00:11 Wuzzy it seems pointless to show page number if its always 1 page 😀
00:12 MTDiscord <Jonathon> it is mtg after all, is it supposed to make sense?
00:12 MTDiscord <Jonathon> anyways, seems a simple fix to remove the line from the formspec?
00:13 Wuzzy oh wait
00:13 Wuzzy it seems its based on linebreaks
00:13 Wuzzy when I enter a text with linebreaks, it does pagebreaks as well
00:13 Wuzzy my test text had no linebreaks ... 😆
00:13 Wuzzy nvm then
00:24 Wuzzy strange. is it normal that going to bed in singleplayer no longer skips night?
00:24 MTDiscord <luatic> have you enabled bed night skip in your settings?
00:24 Wuzzy nvm. wrong setting
00:25 sfan5 bed night skip is the default
00:25 sfan5 and works here
00:27 MTDiscord <luatic> erlehmann: I'm starting to doubt that "awaiting" using coroutines is possible in this case
00:28 MTDiscord <luatic> You basically have to block the main thread until MT calls the callback on the main thread... which makes it impossible for MT to call the callback in the first place...
00:28 Wuzzy yes sorry it was my fault
00:30 sfan5 @luatic coroutines do not actually block
00:30 sfan5 consider this code which works https://github.com/minetest/minetest/blob/master/games/devtest/mods/unittests/init.lua#L133-L134
00:32 Alias2 joined #minetest-dev
00:33 sfan5 but you are correct that you cannot use this to influence the behaviour of "external" functions
00:33 sfan5 so patching the thing we've been discussing in that fashion is not possible
00:46 jordan4ibanez joined #minetest-dev
01:00 kilbith joined #minetest-dev
01:06 kilbith_ joined #minetest-dev
01:26 Alias joined #minetest-dev
01:30 v-rob joined #minetest-dev
01:37 Toothless joined #minetest-dev
01:50 asdflkj_sh joined #minetest-dev
01:57 HuguesRoss Anyone happen to know the setting or engine modif needed to make the final mapchunk border of the map visible? I don't happen to know that one, would help with testing #11866
01:57 ShadowBot https://github.com/minetest/minetest/issues/11866 -- Raise max mapgen limit constant to align with MapBlock by sfan5
02:01 erlehmann joined #minetest-dev
02:05 HuguesRoss erlehmann: Just who I wanted to see, actually. I'll repeat since I don't know if you can see past messages. Do you happen to know what the necessary config is to display the mapchunk border? You allude to it in a few places, but a quick scan of the regular settings turns up nothing so I assume it's either poorly-labelled or more complicated
02:06 HuguesRoss Also, I saw some interesting behavior in your second test. There's an engine bug there, but it's not what your test claims it is. Arguably this is also a bug in the test, but it's possible you did what I'm seeing deliberately so I wouldn't necessarily call it that without knowing
02:07 kilbith__ joined #minetest-dev
02:09 erlehmann let me read the log brb
02:11 erlehmann HuguesRoss i know no easy way to make the outermost mapblock visible in a way that it helps you with the PR you are looking at.
02:11 HuguesRoss alrighty
02:12 erlehmann i think it is possible by editing the map files, but the problem is that the errors habben on *generation* time mostly. so you already need to have a mapblock.
02:12 erlehmann HuguesRoss what is the engine bug you claim? i hope it is not “nodes can be placed out of bounds” again.
02:13 HuguesRoss No, it's an s16 overflow when reading numbers >32767
02:13 erlehmann wait, there still is one?
02:13 HuguesRoss As a result, your find nodes in area call is checking like, 2.8 trillion nodes?
02:13 erlehmann i thought that was fixed
02:13 erlehmann are you *sure* you are on the newest version
02:13 HuguesRoss Not on my end
02:13 erlehmann oof
02:13 HuguesRoss I can double-check, but I'm pretty sure there is one
02:13 erlehmann wait, did you do an incremental build maybe?
02:14 HuguesRoss Maybe the branch is missing a commit?
02:14 erlehmann (incremental builds with cmake in general and minetest in particular are broken and may yield the wrong binary)
02:14 erlehmann hmmmm
02:14 HuguesRoss I'll try rebasing on master and rebuilding
02:15 erlehmann HuguesRoss the issue is that this only came up because of the fix for the call. maybe the fix does not work?
02:15 erlehmann HuguesRoss, which commit are you on?
02:15 HuguesRoss I'm on wherever in the history 11866 is at, but not for long
02:17 HuguesRoss Anyway, on my end I'm not seeing anything that would prevent an overflow in the code even after rebasing on master. Check src/util/numeric.h:288
02:19 erlehmann HuguesRoss so does my test trigger an overflow there?
02:20 erlehmann (you can verify by compiling with ubsan)
02:20 HuguesRoss yes, because one of your coords is 32770
02:20 erlehmann does that happen inside *mineclonia* though?
02:21 HuguesRoss Haven't checked, I took your test directly
02:21 HuguesRoss as linked
02:21 erlehmann oh, okay so
02:21 erlehmann in mineclonia there is actually mcl_engine_workarounds
02:21 erlehmann which contains a lua shim for minetest.find_nodes_in_area
02:21 erlehmann the test case tests if *that* thing is correct
02:21 HuguesRoss I see
02:21 erlehmann i fixed it in mineclonia before the engine fixed it (or not fixed it?)
02:22 erlehmann regardless, does this influence your decision on the PR?
02:22 erlehmann not shim, wrapper
02:22 HuguesRoss no, absolutely not. It was meant as an aside
02:22 erlehmann it calles the real minetest.find_nodes_in_area with sanitized arguments
02:22 erlehmann this is arguably very hacky, but we had a guy crash the server
02:23 HuguesRoss The first test still fails on the before and passes on the after, which I believe is what we were looking to check here
02:23 erlehmann the one that checks if the engine can “still count” i guess?
02:24 HuguesRoss Yes
02:24 HuguesRoss Also, after testing I can now confirm that your version of find_nodes_in_area works so you're all good
02:24 erlehmann nice
02:25 HuguesRoss The engine could use some more checks when getting inputs from mods though though
02:25 HuguesRoss We shouldn't allow numbers that just go out of range like that
02:26 erlehmann that is a common response to that
02:27 erlehmann there are multiple problems with that though
02:27 erlehmann for example, because of the hilarity that is int ←→ float conversion in the engine, you can actually have values that do not roundtrip cleanly
02:28 erlehmann also almost everyone insiststed on using MAX_MAP_GENERATION_LIMIT without knowing the real boundaries hehe when i last brought up the S16 boundaries
02:29 erlehmann quite a few things do actually stop at MAX_MAP_GENERATION_LIMIT btw, like biomegen
02:30 erlehmann HuguesRoss i believe that the best solution is to make the underlying functions work well without overflow. but if i understand celeron55 correctly, not checking for overflow in many cases was a deliberate design decision that i have no chance of changing without finding a dozen practical overflows.
02:31 erlehmann there is some comment about this near where MAX_MAP_GENERATION_LIMIT is defined
02:34 erlehmann HuguesRoss how are you finding overflows btw, reading code or ubsan?
02:34 HuguesRoss Neither
02:34 HuguesRoss I find them by following code execution paths in a debugger
02:36 HuguesRoss Does mineclonia's `spawn_structure` use find_nodes_in_area? Or just the mapgen spawning?
02:36 HuguesRoss I suspect it does, but I'm getting towards that time of the night where it's hard to read code
02:40 HuguesRoss Ok, looks like it doesn't but the callback after does in the case of the desert temples I've been dropping into the border
02:41 HuguesRoss good
02:42 HuguesRoss I think I'm probably satisfied with this. I could dig deeper with some engine modifs, but between the test code and the basic checks I've done manually I think this is ok to merge.
02:42 HuguesRoss I'm not doing it tonight though, not a good idea to merge code right before disappearing
02:44 erlehmann HuguesRoss the spawning structure code has callbacks that use minetest.find_nodes_in_area()
02:44 HuguesRoss yes, I was able to find it thankfully
02:44 HuguesRoss Thank you for confirming it
02:45 erlehmann i appreciate the effort you put into this stuff
02:45 HuguesRoss I try to give a level of scrutiny roughly on-par with what would be expected of me at work
02:45 erlehmann very nice
02:45 HuguesRoss Which is to say, not amazing but still reasonably thorough
02:46 erlehmann well, it's probably better than whatever happened before at that code path (if it was ever reviewed)
02:46 erlehmann because you would not have found anything if it was, i guess
02:46 HuguesRoss Usually we have specialists for this sort of testing, it's too bad we don't have any of those here lol
02:47 kilbith__ joined #minetest-dev
02:48 HuguesRoss hm, come to think of it
02:49 erlehmann HuguesRoss i believe that josiah_wi is trying to restore unit tests for irrlichtmt
02:49 erlehmann maybe you wanna take a look at that
02:50 HuguesRoss I'm no expert, but I can take a look sometime. I'll put it on The List
02:50 erlehmann beyond that, i believe that tests are, unfortunately, only acceptable right now, if they are added as regression tests. i don't really understand why, but this is, for example, why my suggestion of using sanitizers in the CI to catch bugs was not done IIRC.
02:50 erlehmann HuguesRoss could you share your debugging setup for minetest btw?
02:51 erlehmann compile options and all that
02:51 HuguesRoss I'm just running it out of the box in CLion
02:51 HuguesRoss You could probably do the same with GDB or any other IDE providing a frontend to it
02:52 HuguesRoss I'm not usually an IDE guy, but for C++ I find their debugging tools useful
02:52 HuguesRoss You can do most everything they do in the CLI, but it's more of a pain
02:52 erlehmann https://www.jetbrains.com/clion/ this one i assume?
02:52 HuguesRoss yup
02:52 HuguesRoss Costs money though, and  not FOSS
02:52 erlehmann > Generate tons of boilerplate code instantly.
02:52 erlehmann oh lol
02:53 MTDiscord <Jonathon> Clion is pretty good
02:53 erlehmann that is the most hilarious start of a marketing text for a CVE++ IDE i have seen
02:53 erlehmann like who *wouldn't* want to generate tons of boilerplate code!?
02:53 erlehmann :D
02:53 HuguesRoss I will note that CLion feels pretty sluggish. It works well, but likes to take its time doing so
02:54 HuguesRoss I tried a couple FOSS alternatives, but I found they didn't work as well sadly. I'll probably keep looking just in case though
02:54 erlehmann i have 2GB ram on the machine i am writing on, do i need to look closer on this IDE or not?
02:55 rubenwardy Clion uses lldb
02:55 HuguesRoss Definitely not, right now I think it's using 4GB
02:55 erlehmann ok thx
02:55 rubenwardy I use clion, I like intellij IDEs
02:56 HuguesRoss Until recently the only Jetbrains product I'd gotten much use out of was dotPeek
02:56 rubenwardy I use VSCode(ium) for most things though, but the C++ support isn't very good
02:58 HuguesRoss Anyway, I'm gonna go sleep probably.
02:59 HuguesRoss Only thing left, rubenwardy do you happen to know how the 2 approval rule interacts with coredev PRs? I don't remember if there's an exception or not, but we should probably get a 2nd review on #11866 either way
02:59 ShadowBot https://github.com/minetest/minetest/issues/11866 -- Raise max mapgen limit constant to align with MapBlock by sfan5
02:59 rubenwardy self approval counts
02:59 rubenwardy so as a core dev, sfan only needs on other approval
03:00 HuguesRoss Cool, in that case we're probably good then
03:00 rubenwardy however, he can wait for more than one other if he wants to exercise caution
03:00 HuguesRoss I'll ask him tomorrow then
03:00 rubenwardy you shouldn't assume self approval
03:01 HuguesRoss Alright, I think I've got it. I could've sworn this was written down somewhere, but I couldn't find it
03:14 v-rob joined #minetest-dev
03:28 queria^clone joined #minetest-dev
03:34 queria^clone joined #minetest-dev
04:03 MTDiscord <MisterE> HuguesRoss, have you tried vscodium with the c++ extention?
04:29 [0] The official C++ extension can't (legally) be used with the OSS builds of code, but it works with the clangd extension.  You have to run cmake with -DCMAKE_EXPORT_COMPILE_COMMANDS=TRUE.  You'll also have to add --compile-commands-dir=... to the clangd arguments if you're doing an out-of-source build.
04:37 rogerm[m] joined #minetest-dev
04:37 queria[m] joined #minetest-dev
04:37 wsor4035 joined #minetest-dev
04:37 unexploredtest[m joined #minetest-dev
04:37 programmerjake joined #minetest-dev
04:37 zehka[m] joined #minetest-dev
04:37 mister-e[m] joined #minetest-dev
04:37 tochigi joined #minetest-dev
04:37 _Zaizen_[m] joined #minetest-dev
05:00 MTDiscord joined #minetest-dev
06:04 v-rob joined #minetest-dev
06:26 Baytuch joined #minetest-dev
06:53 Baytuch_2 joined #minetest-dev
07:32 v-rob joined #minetest-dev
07:56 tekakutli joined #minetest-dev
08:44 olliy joined #minetest-dev
08:47 calcul0n joined #minetest-dev
10:03 Baytuch hmm.. - "Spoofing yout IP"
10:03 wsor4035 joined #minetest-dev
10:03 queria[m] joined #minetest-dev
10:03 _Zaizen_[m] joined #minetest-dev
10:03 mister-e[m] joined #minetest-dev
10:03 tochigi joined #minetest-dev
10:04 Baytuch Good morning
11:26 proller joined #minetest-dev
11:50 sfan5 looks like I'll prepare a PR to disable basic_debug
11:51 Fixer joined #minetest-dev
12:05 sfan5 it appears none of the privileges (interact, shout or the new basic_debug) are documented anywhere
12:06 sfan5 except in-game but that's not "documentation"
12:12 HuguesRoss joined #minetest-dev
12:12 JordanL2 joined #minetest-dev
12:19 MTDiscord <x2048> Will a simple section in lua_api.txt work for this?
12:22 sfan5 sure
12:28 Baytuch Let it be the option if necessary
12:39 sfan5 merging #11866, #12013, #12014 soon
12:39 ShadowBot https://github.com/minetest/minetest/issues/11866 -- Raise max mapgen limit constant to align with MapBlock by sfan5
12:39 ShadowBot https://github.com/minetest/minetest/issues/12013 -- Fix Minetest logo when installed system-wide. by rollerozxa
12:39 ShadowBot https://github.com/minetest/minetest/issues/12014 -- Update builtin translation by Wuzzy2
12:56 appguru joined #minetest-dev
14:14 erlehmann HuguesRoss are you going to file an issue for the buffer overflow you found? or is it a known one?
14:15 HuguesRoss Well, for starters it's an integer overflow
14:15 HuguesRoss and yeah, once I'm properly awake
14:15 erlehmann sorry, i am not awake either
14:16 MTDiscord <Benrob0329> 12011 makes me sad
14:16 MTDiscord <josiah_wi> I haven't been able to come up with a way to avoid integer overflows.
14:17 erlehmann josiah_wi what do you mean?
14:17 erlehmann what is hard about avoiding them?
14:17 MTDiscord <josiah_wi> Unless you clamp values, every addition or multiplication operation is a possible overflow.
14:19 MTDiscord <luatic> sfan5: mind having a quick look at game#2924
14:19 ShadowBot https://github.com/minetest/minetest_game/issues/2924 -- Addition for translation (uk) by baytuch
14:20 erlehmann josiah_wi you can check input values before doing a calculation to prevent overflow
14:25 erlehmann josiah_wi you can find a lot of examples like this when you look for solutions, does that make sense to you?: int safe_add(int a, int b) { if (a >= 0) { if (b > (INT_MAX - a)) { /* overflow */ } } else { if (b < (INT_MIN - a)) { /* underflow */ } } return a + b; }
14:25 MTDiscord <josiah_wi> Thanks.
14:25 erlehmann if you use gcc, there are also some builtins https://gcc.gnu.org/onlinedocs/gcc/Integer-Overflow-Builtins.html
14:26 erlehmann josiah_wi i believe every example calls these safe_add, but some of them do the addition before checking. you must never do this! signed integer overflow is undefined behaviour, so any checks might be optimized out in such a case.
14:26 MTDiscord <luatic> Always using floats/doubles is not as stupid as it might seem: At least you get proper infinities instead of over/underflows.
14:27 erlehmann luatic i divided by zero in lua and got inf and it was fine because coras code tried to calculate time to next frame from fps setting and i input 0 ^^
14:27 MTDiscord <josiah_wi> Minetest needs to run on old machines that might not be as optimized for floating point arithmetic as modern machines, no?
14:28 erlehmann computers are fast
14:28 erlehmann let me again say that i am on a thinkpad t60, minetest runs fine there
14:28 MTDiscord <luatic> Floating point arithmetic is surprisingly fast, even on old machines
14:29 MTDiscord <josiah_wi> Minetest runs at 5 to 12 FPS singleplayer and 1 or 2 FPS multiplayer on my dev computer. Performance matters to me haha.
14:29 erlehmann josiah_wi i may have sent you this! https://randomascii.wordpress.com/2014/01/27/theres-only-four-billion-floatsso-test-them-all/
14:29 erlehmann josiah_wi oh, that is unfortunate. what kind of computer is it? also, is it on windows?
14:30 MTDiscord <josiah_wi> It's a Dell Inspiron 531 with 2 extra RAM chips. It has an Athalon processor with 2 cores, and I'm running 62 bit Debian Bullseye.
14:31 erlehmann 62 bit?
14:31 erlehmann :D
14:31 MTDiscord <josiah_wi> 64 bit, sorry
14:31 ROllerozxa aw man, is basic_debug going to be reverted last minute?
14:32 erlehmann i hope so
14:32 MTDiscord <luatic> we'll see, sfan5's PR doesn't seem quite ready yet and has just been removed from the 5.5 milestone?
14:32 erlehmann Wuzzy was right about it being messed up
14:32 MTDiscord <luatic> I'll try to make an alternative PR to turn it into a HUD flag
14:32 MTDiscord <luatic> Should I turn the debug priv into a HUD flag as well?
14:32 erlehmann uh
14:32 ROllerozxa I'd prefer if it was given to singleplayer by default and up to the games to explicitly disable them
14:32 sfan5 the removal from milestone has no meaning
14:32 sfan5 PRs to add anything new will not be accepted one day before release
14:33 sfan5 s/one day before/on the day of/
14:33 MTDiscord <luatic> it isn't really "new", it just swaps out the privs for HUD flags...
14:33 erlehmann luatic are you sure we can figure out all the downsides with your approach if it has taken forever to figure out the downside with this approach?
14:33 MTDiscord <luatic> the downsides with this approach are on the modders, that's the neat part :D
14:34 erlehmann you monster :P
14:35 MTDiscord <luatic> Anyways, I'll make a PR, may wait until 5.6 I guess ¯_(ツ)_/¯
14:38 erlehmann rubenwardy celeron55 DM me your IBAN and whatever else i may need (a name? transaction purpose? unclear) to donate for minetest/contentdb hosting pls (i'll use the info from whoever messages me first)
14:38 MTDiscord <luatic> ...
14:39 MTDiscord <luatic> your bribery isn't working erlehmann
14:39 erlehmann luatic the PR got merged. i intend to keep my work regardless of if it was the result of bribery.
14:39 erlehmann my word
14:42 erlehmann also minetest has given me much more fun than any other game i have played i think, so maybe i *should* donate something else than complaints :P
14:43 erlehmann josiah_wi super weird that minetest is so slow on your machine, wanna query me and i try to figure sth out?
14:46 celeron55 erlehmann: contact sfan5 directly to fund contentdb hosting. he probably won't accept it though. and i'm good for years of hosting for my part (which does not include contentdb)
14:46 celeron55 eh
14:46 celeron55 i mean rubenwardy
14:46 celeron55 he probably will accept it
14:46 celeron55 sfan5 won't, but you didn't ask to fund the serverlist
14:46 MTDiscord <luatic> honestly erlehmann should probably be donating to sfan5 if anything
14:47 sfan5 money cannot be exchanged for time so that'd be pointless
14:47 MTDiscord <luatic> money is exchanged for time all the time
14:47 celeron55 i have heard rubenwardy also good money wise regarding to hosting though
14:47 sfan5 not in those quantities
14:47 MTDiscord <luatic> true
14:48 Taoki joined #minetest-dev
14:48 celeron55 the other thing about converting money to time is that people usually have full time work elsewhere that they can't take time off to do other work even if they were paid for it
14:50 MTDiscord <luatic> wow, HUD flags were written with forwards-compatibility in mind!
14:50 MTDiscord <luatic> the server can send a mask that is
14:51 erlehmann celeron55 it's not about paying for time (way too little money for that), it's about sending a message. “this PR is worth as much as buying a full game is to me”. i could have also said i'd donate to a charity of your choice, but then i thought that there are definitely charities i do not agree with.
14:52 erlehmann nvm, i won't make such an offer again, apparently it is more upsetting than anything. (my existing 2 still stand though.)
14:52 celeron55 the thing is, you really can't vote with donations
14:53 erlehmann nah, i just thought maybe it gets someone's attention
14:53 celeron55 we don't have a system in place for that, and we haven't unofficially done such thing, ever
14:53 erlehmann i figured it's a curveball
14:53 erlehmann dw i doubt i can buy PRs lol
14:54 erlehmann i just thought it might be a better attention grabber than whining on irc (bc i have tried that)
14:54 erlehmann as i said, i won't do it again.
14:55 celeron55 if you want to make a post on social media that you wanted to pay full price for minetest and did so by donating to a charity of the developer's choice, my choice will be EFF
14:55 erlehmann ok, i'll donate to the eff if neither you nor rubenwardy wants it!
14:56 MTDiscord <luatic> what!?
14:56 erlehmann i believe i should stop taliking about this
14:56 celeron55 this is a bit silly but you do have a point
14:56 erlehmann wdym “have a point?”
14:56 erlehmann with what
14:57 celeron55 i mean, you're getting attention with money
14:57 celeron55 8)
14:58 celeron55 your other option is to save the money and offer it to me in the future when/if i or someone else has an actual use for it that benefits MT
14:58 celeron55 i don't want to bank donations, it's borderline illegal and doesn't promote trust
14:59 erlehmann oh, i see.
14:59 erlehmann makes sense
15:00 erlehmann years ago, when i collected donations for a thing where the hosting costs exceeded my income (huge download, got too popular), i also only asked for donations to cover current, not far future costs.
15:01 celeron55 i have enough for years of hosting. at this rate i'm considering removing the paypal donation option from the website. i got what i need to host the website, and don't have so much time to use for MT that "paying" donations to myself would feel fair
15:01 erlehmann understandable
15:01 erlehmann hmm, the CI is not a cost center?
15:02 ROllerozxa github actions is free
15:02 erlehmann oh, i didn't know
15:02 ROllerozxa for public open source repositories at least
15:02 ROllerozxa anyways put the money to a bounty for libopenmpt support in minetest :)
15:02 celeron55 the only official hosting that's coming from someone's pocket is sfan5's serverlist. but he refuses donations because he hosts other stuff on the same machine
15:04 celeron55 the increase of donations needed to pay someone full time is something like 100x compared to hosting costs
15:04 erlehmann ROllerozxa funny, but i am not a fan of sending obscure input formats over the network.
15:04 celeron55 hosting is tens of money units per month total, a monthly full time paycheck is thousands
15:04 erlehmann celeron55 i wanted to specifically avoid to give money to a person instead of a cause, bc it's a) way too little b) it's too close to actual bribery.
15:08 ROllerozxa erlehmann: aw come on, libopenmpt has seen a lot of testing to not crash on malformed files (fuzz-testing and whatnot). it's certainly better than the older libmodplug library which is essentially abandoned at this point
15:08 MTDiscord <Jonathon> c55: why not keep the donation link up, but use the money for core dev voted bug bounties?
15:09 ROllerozxa erlehmann: I just want to see the possibility for music in minetest that isn't absolutely massive. MO3 in particular would be awesome for minetest
15:09 erlehmann ROllerozxa i see the appeal, but minetest is so full of holes it is not funny
15:10 Yad joined #minetest-dev
15:10 erlehmann no need to add more rn
15:10 erlehmann ROllerozxa could you give me a small example program to fuzz?
15:10 MTDiscord <Jonathon> minetest cant even keep audio working that long on windows anyways lol
15:11 celeron55 Jonathon: we can think about core dev voted bug bounties, not sure. if such were implemented, then the donation link should go directly to the issue on bountysource or some such
15:11 MTDiscord <Jonathon> just a suggestion
15:12 celeron55 basically the donation link would be a link carousel of sorts, picking one of the issues from a list updated periodically by the core team
15:12 celeron55 or something like that
15:12 celeron55 then the money wouldn't go through anyone, avoiding that legal problem
15:15 celeron55 a more fleshed out proposal wouldn't hurt. someone would have to also implement it
15:15 erlehmann celeron55 since you have not replied to my proposal for maintaining a “stable” version of minetest that gets improved but none of the API breakage (like other projects do it), i guess you don't like it. i'll probably try to contribute that to debian then, adding new patches to old versions.
15:15 erlehmann (mostly security focused)
15:16 celeron55 my question regarding to that is, do we have the resources to do it
15:16 Wuzzy joined #minetest-dev
15:16 ROllerozxa erlehmann: https://github.com/OpenMPT/openmpt/tree/master/contrib/fuzzing
15:16 ROllerozxa this is their fuzzing suite, I'm sure you probably could get it to run yourself
15:17 erlehmann celeron55 no idea. i mean people are maintaining cheat clients alone and those are mostly based on 5.4.1. i just thought it would also take the heat off discussions. mcl2/mcl5 feature breakage was much more drama before ppl could say “just use mineclonia”.
15:18 erlehmann josiah_wi i suggest to read this paper on integer overflow btw https://www.cs.utah.edu/~regehr/papers/tosem15.pdf
15:19 erlehmann josiah_wi you can ofc also do a width extension overflow test. do the calculation in a wider datatype that can not overflow, then check the result.
15:19 erlehmann josiah_wi i doubt that it works as well as checking inputs though, bc at that point you have type conversions going on, so speed may be an issue
15:19 celeron55 erlehmann: what you propose would practically mean that almost all feature work done by the core team and contributors would go towards the next major version as most of the time features can't be added without *some* kind of imcompatibility when thought about in the strictest way possible
15:20 celeron55 there would be an almost separate team doing minor versions
15:20 erlehmann celeron55 you would be surprised how many game devs are basically fine with the engine except for 1 or 2 of their pet features which they can fake.
15:21 erlehmann fake, as in: tricks like the spiderweb which is a fluid (?)
15:22 erlehmann if 5.4.1 was feature frozen forever, there are still untold possibilities, as the game jam showed
15:22 erlehmann but security updates and performance improvements are important
15:22 erlehmann and no one is going to say “uh i want the less secure, less performant version. bc that is a breaking change”
15:22 celeron55 i would be fine with what i just described, it would be quite different to what's been done ever since MT has existed though
15:24 celeron55 basically, move the versions yet again to the left by one dot
15:24 celeron55 (and remove the first number)
15:24 erlehmann if i can get a list of stuff that is definitely not breaking things, i volunteer to apply that to 5.4.1 to make a 5.4.2 and maintain it for at least a year (as in: apply new stuff to it that is not API breaking)
15:24 MTDiscord <Jonathon> maybe you should do you research first
15:25 MTDiscord <Jonathon> cause 5.4.2 already exists
15:25 erlehmann the thing is, it becomes easier the more stuff is a breaking change ^^
15:25 erlehmann for android, yeah, lol
15:25 erlehmann sorry
15:25 MTDiscord <Jonathon> it would have to be 5.4.3
15:25 erlehmann hadn't consideret that
15:25 MTDiscord <Jonathon> there's a lot of things you dont consider
15:26 erlehmann which is why i talk to ppl who call me out on my bullshit, and i try to do it with theirs
15:28 celeron55 erlehmann: can you write a proposal? bringing this up freely worded in random places at random times doesn't help
15:28 erlehmann celeron55, i can not guarantee it, while writing it i might figure out it's a bad idea after all. but i'll look into it!
15:29 erlehmann Jonathon I had a coworker once who seemed to be very smart but rarely said anything. when i asked her if she was shy, she said: no, she just talks only when she is really sure, to appear super smart. unsurprisingly, i did not have many discussions with her where i learned something new.
15:29 celeron55 it needs to be a proposal that the core team can look at and either agree, change it up and then agree, or disagree
15:29 erlehmann yes, of course
15:30 mrkubax10 joined #minetest-dev
15:30 erlehmann i just think the “one person maintaining a fork” thing is unsustainable, bc then you get stuff where ppl like fleck or cora are a single point of failure.
15:30 celeron55 i could write it myself but i'm not sold on the idea and i'm not sure about the details, so it would feel weird
15:30 erlehmann so i'm going to ask them and kay27 about it
15:31 sfan5 if you want that branch to be compatible with normal minetest that's going to be a huge mess: you either get stuck on an old protocol version and new features may not work or you have to follow new protocol versions and are forced to implement the same stuff upstream does all while figuring out how to not "break" something
15:31 MTDiscord <Jonathon> ^
15:32 erlehmann sfan5 old protocol version obv
15:32 celeron55 it wouldn't work like that. the only way it would work is features are not to be added into it
15:32 celeron55 only bugfixes, security fixes and compatible performance fixes
15:32 sfan5 okay so abugfix only branch
15:32 sfan5 +" "
15:32 erlehmann yeah, that would also benefit linux distributions
15:33 celeron55 i can see it could work, but the core team has to understand it and agree to it
15:33 sfan5 s/liinux distributions/debian/
15:33 erlehmann sfan5 a lot of stuff is based on debian
15:33 MTDiscord <luatic> s/linux distributions/not arch
15:34 erlehmann sfan5 “tell me that btw you use arch without literally telling me that btw you use arch” ;)
15:34 celeron55 it would essentially be like an extended support release. there's a chance it would skip over some major releases
15:35 MTDiscord <Jonathon> tbh its probably at best just going to become a time sucker from actually working on mte
15:35 erlehmann why would backporting fixes suck time away
15:36 erlehmann if you got someone volunteering to do it
15:38 celeron55 some MT forks actually are in nature just like extended support releases, so the manpower might be out there. the questions is can it be collected into one place
15:40 celeron55 well, forks do have a tendency to devolve into adding incompatible features and then not being able to jump to the next upstream version
15:40 celeron55 that's the thing to avoid
15:40 erlehmann that's why i want it to be very constrained
15:46 celeron55 people hating new incompatible features is actually a compliment. it means for some purposes, MT is already good enough. so in theory those who like what's already there can start maintaining that, and those who want more can develop more
15:47 celeron55 in practice, i have no idea if it can work like that
15:49 celeron55 if we are to take that route, then this release might have to be 6.0.0
15:51 celeron55 basically every time people start complaining about new features it's an indication that the major version has to be incremented
15:53 erlehmann i still think that for a lot of things people should consider adding numeric suffixes to functions. linux has clone2 and clone3 and wait3 and wait4 …
15:53 sfan5 we are neither linux nor windows
15:53 erlehmann i'm just saying that is how you avoid complaints
15:53 erlehmann it's not the new features, most of the time.
15:53 erlehmann it's the breakage
15:54 sfan5 those two happens to be connected
15:54 sfan5 happen*
15:56 erlehmann you keep saying that and what i hear is “it would be *way* too much effort”, is that an interpretation you can agree with?
15:57 sfan5 obviously
16:01 MTDiscord <luatic> What's the point of the register_playerevent return value? It doesn't seem to be used at all
16:15 jordan4ibanez joined #minetest-dev
16:32 mrkubax10 left #minetest-dev
16:33 Krock @luatic none. ignored in ScriptApiEnv::player_event.
16:33 Krock maybe there used to be one? it's an internal function after all, hence I would expect some changes every now and then
16:34 MTDiscord <Benrob0329> I'd like to voice my support for labeling this release 6.0 and getting on with it
16:36 Krock are we doing polls? I'd vote against that.
16:38 MTDiscord <Benrob0329> Wait wait wait, don't engine-default privs have a conf setting? Just add it to that, and let games who don't want it set the default_privs setting in the game provided minetest.conf
16:39 MTDiscord <Benrob0329> Boom, everybody's happy
16:40 sfan5 that does not apply to existing users
16:43 MTDiscord <Benrob0329> Can we have an upgrade privs setting? If there must be a perfect solution where nobody has to do anything, maybe a setting that grants users a privilege the first time they join since a certain date?
16:43 MTDiscord <Benrob0329> This feels like an easy mod though
16:45 sfan5 enough bandaid could be put on the current impl to make it okay
16:45 sfan5 or we could just drop it and reimplement it using hud flags
16:45 MTDiscord <Benrob0329> In the end, somebody is going to be unhappy. If it must be me, then it must be me.
16:45 MTDiscord <Benrob0329> Wasn't there a rejected hud flags implementation?
16:46 MTDiscord <Benrob0329> Also, the priv makes it easy for admins to grant it to players
16:46 MTDiscord <Benrob0329> Maybe make it a disable_debug priv?
16:47 sfan5 both ways were discussed once
16:47 MTDiscord <luatic> No, reverse privs are dirty
16:47 MTDiscord <Benrob0329> I don't care
16:47 MTDiscord <luatic> modlib has utils for checking against them I believe, but the engine doesn't
16:50 rubenwardy I don't remember a HUD flags implementation
16:50 MTDiscord <Benrob0329> A priv, a property, a flag, positive, negative, we have something for this finally but it's going to be reverted because there might be servers which won't react quickly (not an engine problem), existing games might want it by default, and somebody complained.
16:50 rubenwardy I suggested it, but I don't remember seeing it be done
16:51 rubenwardy I either want:
16:51 rubenwardy - it to be left as is
16:51 rubenwardy - or for it to be reverted and replaced with HUD flags in the next release
16:51 rubenwardy Adding a hacky mod to MTG will just make it harder to mods to disable it
16:52 MTDiscord <Benrob0329> I'll say it again, not everyone can be made happy, devs have to be ok with that. I'll go back to my cave of trying not to bitch about the engine as much as possible now :-)
16:52 sfan5 btw that somebody who complained is the person who initially proposed, implemented and argued for the solution
16:53 MTDiscord <Benrob0329> I am aware
16:53 rubenwardy That hacky mod could be third party though
17:01 tekakutli joined #minetest-dev
17:12 ROllerozxa HuguesRoss: I saw you commented on my PR about how it's unfortunate texture pack main menu textures require a restart, so I made a fix for it in #12018
17:12 ShadowBot https://github.com/minetest/minetest/issues/12018 -- Apply texture pack main menu textures immediately by rollerozxa
17:12 MTDiscord <Benrob0329> @Sharp had a decent suggestion, /grant_to_all and /revoke_from_all commands
17:12 HuguesRoss So I saw! I'll review it later, thanks for the PR
18:11 MTDiscord <Warr1024> I'm just going to say that I'm very disappointed about the basic_debug revert.
18:12 MTDiscord <Warr1024> I'm not going to provide a suggestion for how to fix it because lack of suggestions has never been the problem.
18:13 sfan5 I suppose for nodecore it did the right thing out of the box?
18:14 MTDiscord <Warr1024> Not sure what you mean.  It has always done as close to the right thing out of the box as I'm able to achieve.
18:15 sfan5 if for nodecore you wanted the coordinate display to be hidden then the PR did that for you without needing to anything
18:15 MTDiscord <Warr1024> Yes, it would have worked.  And actually was working well for those users using 5.5-dev on 5.5-dev servers.
18:15 MTDiscord <Warr1024> There were actually enough people apparently using 5.5-dev that this is probably going to be a visible regression for them.
18:16 erlehmann providing debug info is a regression?
18:16 MTDiscord <Warr1024> Correct.
18:16 erlehmann in what way? they can simply not press f5
18:16 MTDiscord <Warr1024> They have no way to know that they're supposed to not press F5.
18:17 erlehmann whereas in the other direction, users who could press f5 to get their bearings, no longer can
18:17 sfan5 you out of all people should have no issue considering something a regression by a strict definition
18:17 MTDiscord <GreenXenith> Can we like, not rebuild the bike shed?
18:17 MTDiscord <GreenXenith> This is ridiculous
18:17 MTDiscord <Warr1024> I mean there was a reason why I said "not gonna propose a suggestion"
18:18 erlehmann i think that wuzzy, who proposed and implemented the thing, now has gone back on it, speaks for itself.
18:18 MTDiscord <Warr1024> Benrob came in here to make sure that his protest was registered, from what I could tell, so I just wanted to make sure that people were not unaware of mine as well.
18:18 MTDiscord <GreenXenith> Mostly directed towards erl
18:18 erlehmann wuzzy has always been an advocate of users. much more than any other dev i know.
18:18 MTDiscord <GreenXenith> Seems like the idea is to replace it with a hud flag anyway?
18:19 MTDiscord <Warr1024> I am not here to have the same argument in triplicate.  I'm here to advocate for MY users.
18:19 MTDiscord <Warr1024> A HUD flag would work for me.  It would have worked for me the first time it was proposed and then ultimately rejected :-|
18:19 erlehmann same
18:19 MTDiscord <Warr1024> I can only hope that changes have happened in the project and/or people involved that would make that past rejection no longer relevant to its future plight.
18:20 erlehmann people can change their opinions in light of new evidence, you know
18:20 MTDiscord <GreenXenith> From what I could tell, the priv is only being reverted on the condition that a hud flag is introduced immediately to replace it
18:21 MTDiscord <Warr1024> You mean in 5.5?
18:21 MTDiscord <GreenXenith> Yes
18:21 MTDiscord <LandarVargan> Also note there is a setting to enable debug info by default, I do a lot of modding so I have that set
18:21 MTDiscord <Warr1024> I could support doing that, so long as it doesn't mean I'm ALSO supporting further delay of 5.5's release...
18:21 MTDiscord <LandarVargan> A lot of CTF players set it for the coordinates
18:21 erlehmann LandarVargan now is that good or bad, hmm?
18:22 MTDiscord <Warr1024> Whether it's good or bad in general really depends on the game.
18:22 MTDiscord <LandarVargan> It means that you have to disable it manually every time you join a Nodecore world
18:22 erlehmann a lot of anarchy server players also set it for the coordinates, so they do not build on straights or diagonals. griefers like to go to coordinates like 1000,0,1000 or 5000,0,0 and TNT everything.
18:23 MTDiscord <LandarVargan> So I discount your 'just don't press f5' argument
18:23 erlehmann and before someone discounts minetest anarchy, it's like normal minetest with super smart mobs that may fight you or collaborate, who knows.
18:24 sfan5 I'll go and import the weblate translations soon
18:24 sfan5 then we can merge 12016
18:24 sfan5 and 12001
18:25 sfan5 and perhaps decide about 12006
18:25 sfan5 and then do the release
18:25 sfan5 if that's fine with everyone
18:25 Krock alright
18:26 Krock quickly testing 12006
18:26 MTDiscord <LandarVargan> Why was this added/removed? https://github.com/minetest/minetest/issues/11992
18:27 sfan5 I thought it was not worth delaying 5.5 to fix this, it's not a critical issue either
18:27 Krock is it a new issue? probably not.
18:27 sfan5 good that you mention it though, maybe someone disagrees
18:28 MTDiscord <GreenXenith> So whats happening with 12011? That is happening before 5.5 yeah?
18:28 sfan5 it's happening but the way you might think
18:28 sfan5 basic_debug is being removed
18:28 sfan5 not the way*
18:29 MTDiscord <GreenXenith> Well yes we know that
18:29 MTDiscord <GreenXenith> But it needs to be replaced by a hud flag yeah?
18:29 sfan5 after 5.5, yes
18:29 MTDiscord <GreenXenith> Oh, great
18:29 MTDiscord <LandarVargan> That's a long time to wait, servers you can update but not clients
18:29 sfan5 (misread your first question as "is 12011 happening before 5.5"...)
18:30 sfan5 well 5.6 doesn't have to take as long as 5.5 to release
18:30 sfan5 twice a year is our rough schedule, we can deviate from it in either direction
18:30 MTDiscord <GreenXenith> Never mind @Warr1024, dev regression time. I give up :/
18:31 MTDiscord <LandarVargan> Players don't update as soon as a new release comes out though
18:31 MTDiscord <Warr1024> Green: yeah, that's what it sounded like to me at first.
18:32 MTDiscord <Warr1024> I mean technically it works on an "it was never actually released" basis, but there will be people who'll notice still...
18:32 MTDiscord <GreenXenith> Inevitably it's gonna be bikeshedded for months and 5.6 will take another year
18:37 sfan5 a single issue is not going to delay a release by 6 months
18:37 MTDiscord <GreenXenith> The two don't have to correlate
18:38 sfan5 if they do not correlate there is no causal relation either
18:39 MTDiscord <GreenXenith> Ok
18:40 MTDiscord <GreenXenith> Put a period between them if it makes you feel better
18:40 sfan5 I misunderstood what you were saying then
18:41 sfan5 by the way
18:41 sfan5 who is writing the changelog?
18:42 Krock I can do that, but it'll have to wait about two hours
18:42 sfan5 take as much time as you need
18:42 Krock okay
18:44 Krock PROTOCOL_VERSION has been bumped earlier in 5.5.0-dev already. good so far
18:45 Krock FORMSPEC_API_VERSION was increased from 4 -> 5
18:45 MTDiscord <GreenXenith> I just mean the issue is going to be bikeshed, and when it is finally merged, the next version is still gonna take forever to release
18:45 Krock a minetest release is long overdue
18:45 Krock there's surely time for discussions after this one
18:46 Krock worst case is a stall of one feature, which however should not slow down the release cycle
18:48 v-rob joined #minetest-dev
18:49 appguru joined #minetest-dev
18:51 wsor4035 joined #minetest-dev
18:53 MTDiscord <LandarVargan> How hard would it be to quickly add a hud flag?
18:53 rogerm[m] joined #minetest-dev
18:53 queria[m] joined #minetest-dev
18:53 _Zaizen_[m] joined #minetest-dev
18:53 programmerjake joined #minetest-dev
18:53 unexploredtest[m joined #minetest-dev
18:54 Krock not hard, but error-prone when doing it without good testing
18:54 zehka[m] joined #minetest-dev
18:54 mister-e[m] joined #minetest-dev
18:54 tochigi joined #minetest-dev
18:54 MTDiscord <LandarVargan> If those who want it can't take the time to test it then I don't see why it needs to be added
18:54 MTDiscord <LandarVargan> *added for 5.5
18:54 MTDiscord <LandarVargan> I shall create a PR, hud flags sound pretty simple
18:55 Krock famous words before desaster
18:55 appguru I already worked on it. Remember to send the bits inverted so that they default to one if not sent ;)
18:55 sfan5 I understand the sentiment but hud flags, migration code or anything more than a simple revert is not going to be in 5.5
18:56 sfan5 weblate commits https://github.com/minetest/minetest/commits/ci, waiting for CI
18:56 sfan5 assuming it even runs
18:56 HuguesRoss I'd invite you to do it anyway though, the sooner we get a PR the sooner we can start the review process. Personally I think it'd be nice to aim for an accelerated dev cycle for 5.6, so we can re-instate the stuff cut out of 5.5 with the desired fixes at a reasonable pace
18:57 * Krock goes afk. back in approx. 1 hour
18:57 sfan5 absolutely
18:57 sfan5 I just don't want disappointment when someone tells you in an hour that your finished & tested PR will not be in 5.5
18:58 MTDiscord <luatic> The problem is that we get in hell's kitchen if the debug priv is not ripped out
18:59 MTDiscord <luatic> We're already adding an inconsistency :(
18:59 MTDiscord <luatic> i.e. if you revert the basic_debug priv, please revert the debug priv as well - just revert the entire darn PR
19:00 MTDiscord <GreenXenith> Pretty sure debug priv predates the pr
19:00 MTDiscord <Jonathon> ^this
19:00 MTDiscord <luatic> you're right
19:05 appguru1 joined #minetest-dev
19:05 MTDiscord <LandarVargan> I'll create the PR so I won't wish I did
19:05 MTDiscord <luatic> I'll do it already
19:05 MTDiscord <LandarVargan> I feel like the feature freeze was meant to stop situations like this though
19:05 MTDiscord <LandarVargan> Oh, you started it?
19:06 MTDiscord <LandarVargan> Feel free then
19:06 MTDiscord <Warr1024> PR for which thing?
19:06 MTDiscord <LandarVargan> basic_debug as a hud flag
19:06 MTDiscord <Warr1024> ah, okay
19:07 MTDiscord <Warr1024> I guess hopefully we could at least get it into 5.6-dev early in the cycle so people who were tracking dev before won't have much of a hiccup.
19:08 MTDiscord <Warr1024> I also want to see 5.5 get officially released sooner rather than later because I've been holding off on my servers at the last commit before the zstdpocalypse.
19:09 MTDiscord <Warr1024> Once it's been officially released then MT will actually officially support zstd maps and I can start expecting players to be able to use them.
19:13 MTDiscord <LandarVargan> @Luatic my PR works, not sure how far along you are
19:15 MTDiscord <luatic> Should I write lua_pushboolean(L, (player->hud_flags & esp[i].num) == 0); with or without the == 0=
19:15 MTDiscord <luatic> ?
19:15 MTDiscord <luatic> I was done until I realized I needed to put debug back in. I still need to wrap my head around sending the bits inverted.
19:15 MTDiscord <LandarVargan> Where is esp[i].num coming from?
19:16 MTDiscord <luatic> const EnumString *esp = es_HudBuiltinElement;
19:16 MTDiscord <luatic> I'm cleaning up some code in the process
19:16 MTDiscord <LandarVargan> Ah
19:16 MTDiscord <luatic> l_hud_get_flags currently isn't using es_HudBuiltinElement, but should be
19:17 wsor4035 joined #minetest-dev
19:21 MTDiscord <luatic> Let's recap all client - server cases: old - old: doesn't matter; new - old: debug should be show; old - new: doesn't matter; new - new: debug should be flag-controlled
19:21 MTDiscord <luatic> when not sending inverted: new - old: debug won't ever be shown I guess
19:23 programmerjake joined #minetest-dev
19:23 unexploredtest[m joined #minetest-dev
19:23 tochigi joined #minetest-dev
19:24 MTDiscord <luatic> Although, I believe the mask might render this obsolete
19:28 MTDiscord <luatic> Yep it shouldn't matter
19:29 rogerm[m] joined #minetest-dev
19:29 queria[m] joined #minetest-dev
19:29 _Zaizen_[m] joined #minetest-dev
19:29 zehka[m] joined #minetest-dev
19:29 mister-e[m] joined #minetest-dev
19:40 MTDiscord <luatic> #12020
19:40 ShadowBot https://github.com/minetest/minetest/issues/12020 -- Debug HUD flag by appgurueu
19:41 HuguesRoss re #12006, I've approved but left a note about it. I'm not 100% secure with this fix, and I'm interested in opinions about if my fears are correct or not
19:41 freshreplicant[m joined #minetest-dev
19:42 ShadowBot https://github.com/minetest/minetest/issues/12006 -- Abort raycasts that go out-of-bounds by sfan5
19:43 MTDiscord <savilli> i'm 100% secure with this fix in terms it won't freeze the server, but i still think raycast should be fixed to return the right points
19:45 MTDiscord <luatic> ^
19:50 v-rob joined #minetest-dev
20:10 MTDiscord <Warr1024> What happens if I try to set a HUD flag to false on a version before the HUD flag was added?  Is it just silent-ignored, or will it throw an error?  Will I have a way to detect whether I can use and/or rely on that flag?
20:10 rubenwardy It'll be silently ignored
20:13 MTDiscord <luatic> It's simple: Clients and servers up to and including 5.4 obviously won't care.
20:15 MTDiscord <Warr1024> Excellent.
20:28 sfan5 merging 12006, 12016, 12001
20:47 sfan5 starting release process now
20:48 sfan5 (for reference https://dev.minetest.net/Releasing_Minetest#The_process)
20:51 Krock 60% done with the changelog (local draft)
20:59 sfan5 thanks for doing that
21:03 appguru joined #minetest-dev
21:06 sfan5 I exit an MTG session on the win32 build and it crashes, this is not good news
21:06 Krock native Windows or Wine?   (70%)
21:08 sfan5 wine
21:08 Krock shall I test on Windows for comparison?
21:08 sfan5 I'll test that too but last time it was a problem on both
21:15 sfan5 https://github.com/minetest/minetest/commit/70df3d54f37c280f7afe60f6e964b8406577f39f this seems very similar
21:15 sfan5 I wonder how I found out the source last time
21:21 Krock changelog is done
21:22 Krock windbg?
21:22 Krock generate and examine a minidump file in windbg
21:24 sfan5 I'll try gdb
21:26 sfan5 310Mminetest.exe
21:26 sfan5 that's a lot of debug information
21:27 Krock gtg. good luck with hat.
21:31 sfan5 gdb says it's the thread_local zstd context deletion
21:31 sfan5 now this is tricky, because we can't just make it global
21:32 sfan5 is it okay if I introduce a memory leak?
21:33 sfan5 ah wait it's the unique_ptr itself
21:45 MTDiscord <LandarVargan> Oh, could we get installer builds please? https://github.com/minetest/minetest/pull/11176
21:46 sfan5 rubenwardy, HuguesRoss, v-rob: requesting opinions on http://sprunge.us/VNVXRz?diff
21:47 rubenwardy Installer builds are great for user experience
21:48 rubenwardy That fix is dumb, as mentioned
21:48 rubenwardy But the memory leak is limited if only one is created
21:48 sfan5 worse than dumb
21:48 HuguesRoss sfan5, If I understand your previous comments is the crash tied to std::unique_ptr itself?
21:49 v-rob Ouch. Thanks, MinGW
21:49 HuguesRoss ie. can we manage this one resource ourselves?
21:49 rubenwardy Do we need to continue supporting mingw when we make our official builds using BS
21:49 sfan5 mingw32 in whatever version we currently have seems to have a problem with destructing thread_local variables
21:49 rubenwardy *VS
21:50 sfan5 (in it's defense I haven't updated the compiler in at least a year)
21:50 sfan5 gdb reports that whatever is being destructed (before std::set, now std::unique_ptr) has its destructor called with this = nullptr
21:50 HuguesRoss hm, I see. I guess what I'd do then is implement the leaky fix but document it in an issue?
21:50 sfan5 in this case the resource cannot be shared between threads so the only other alternative is creating a new resource on every call
21:50 v-rob Is there an alternative to thread_local that we could use?
21:51 sfan5 rubenwardy: official builds haven't been VS in years
21:51 Wuzzy alright, I am going on a Minetest-dev testing spree now. wish me luck
21:51 HuguesRoss That way, if/when a fix occurs we can remove the hack
21:51 rubenwardy That doesn't make sense. Is this an upstream bug?
21:51 sfan5 yes
21:51 sfan5 it's not our fault
21:51 Wuzzy feel free to point me to anything important you want tested
21:51 sfan5 HuguesRoss: sure, I want to get rid of it asap anyway
21:51 sfan5 not sure where to best document this though
21:51 sfan5 maybe an issue?
21:51 HuguesRoss that was my idea, yeah
21:51 rubenwardy sfan5: could we update and use the GitHub MSVC builds in future?
21:52 rubenwardy Obviously not for this release
21:52 sfan5 we could yes
21:52 rubenwardy iirc it's just missing deps right? Like LuaJIT GC64
21:53 sfan5 true
21:53 proller joined #minetest-dev
21:53 sfan5 with mingw we (or rather I) have the entire dependency chain under control so I'd rather stay with it
21:53 sfan5 anyway I'll push the aforementioned awful fix
21:53 sfan5 and open an issue
21:54 sfan5 then re-do the windows builds and we can continue with the release
21:54 rubenwardy doesn't vcpkg give us control over the chain?
21:54 rubenwardy cross-compiling is suboptimal when making native builds
21:54 sfan5 "control" in this case meant that I wrote the scripts that build the dependencies myself
21:54 sfan5 just like android basically
21:55 sfan5 (https://github.com/sfan5/mingw-pkgs fyi)
21:55 Wuzzy Any translators hanging out here?
21:56 Wuzzy because it would be nice to have builtin in more languages 😉
21:56 sfan5 rubenwardy: btw can you quickly confirm that the new version code needs to be 38?
21:58 rubenwardy correct
21:59 sfan5 ty
22:01 v-rob Alternatively, is it possible to compile stuff with VS make make them official Windows builds for this release?
22:01 v-rob Or is there a problem with that I don't know about?
22:02 sfan5 temporarily switching to VS is too many unknowns IMO
22:03 sfan5 in other news I just ran out of running two compiles at the same time while having my windows VM running
22:03 sfan5 and I already have 16 gig o.O
22:03 sfan5 +RAM
22:05 lhofhansl joined #minetest-dev
22:08 sfan5 okay windows builds are fine this time
22:13 sfan5 looks like that's all, let's go
22:24 sfan5 website updated, news topic posted
22:24 sfan5 >Announce the release on the Twitter account. rubenwardy has access.
22:24 sfan5 and whatever this involves: Add the new version to the drop-down list of compatible Minetest versions that authors can select for their things.
22:25 sfan5 if someone has cool ideas which features to highlight in the topic (https://forum.minetest.net/viewtopic.php?f=18&amp;t=27754) please write them here
22:25 HuguesRoss Only thing I'd note in the changelog, the zstd compression is mentioned under deprecations and compatibility notes but doesn't mention that the conversion makes maps not loadable in 5.4.x
22:26 HuguesRoss Might be worth putting that down just to ensure folks are aware
22:26 Topic for #minetest-dev is now Minetest core development and maintenance. Chit-chat goes to #minetest. https://dev.minetest.net/ https://github.com/minetest
22:26 Pexin was rubenwardy serious about maybe releasing an external reverse migration util?
22:29 rubenwardy merging #11784 in 10
22:29 ShadowBot https://github.com/minetest/minetest/issues/11784 -- Use virtual paths to specify exact mod to enable by rubenwardy
22:29 rubenwardy and also making an Android build
22:29 erlehmann releasing an external reverse migration util sounds good. who of you has one?
22:29 erlehmann i mean, strictly for dev purposes
22:30 erlehmann or when you accidentally have no backup
22:30 sfan5 HuguesRoss: added
22:33 Pexin I'd consider making that Note <b>NOTE</b> but, shrug
22:38 Wuzzy "Chop game background in mainmenu" → this breaks the Tutorial mainmenu screen
22:38 Wuzzy Tutorial uses overlay.png to have some basic instructions on what to click
22:39 sfan5 hmm I guess I could download the mac build from github and link that on the forum
22:39 sfan5 or do those need to be signed off by apple
22:39 sfan5 can someone from the discord side ping Jordach on this?
22:40 rubenwardy Jordach:
22:40 rubenwardy that pings Discord users
22:41 sfan5 discord users have a tendency to rename themselves to random things
22:41 sfan5 so I wasn't sure if that'd work
22:41 MTDiscord <Jordach> uwot
22:41 rubenwardy true, I suspect it's based on username
22:41 rubenwardy but yeah, hard to check from this side
22:41 MTDiscord <Jordach> >be signed off by apple
22:41 MTDiscord <Jordach> nope
22:41 MTDiscord <Jordach> you can run unsigned on M1
22:41 asdflkj_sh joined #minetest-dev
22:42 MTDiscord <Jordach> app store yes, self made builds no
22:43 sfan5 so if I download the CI artifact and offer it for people to download they can just run it?
22:43 MTDiscord <Jordach> yes
22:43 sfan5 that's neither self made nor app store
22:43 MTDiscord <Jordach> with the correct security procedure
22:43 sfan5 ...
22:44 MTDiscord <Jordach> you have to manually allow non signed builds yourself
22:44 sfan5 that sounds bad
22:44 sfan5 can we get our build signed?
22:44 MTDiscord <Jordach> yes
22:44 sfan5 not for app store distribution, just file downloads
22:44 MTDiscord <Jordach> but do you have the functional equivalent of $99
22:45 sfan5 I thought notarization was free?
22:45 MTDiscord <Jordach> requires the dev license
22:46 sfan5 disappointed but not surprised
22:46 MTDiscord <Jordach> with Linux on M1 getting increasingly closer with full Mesa support
22:47 MTDiscord <Jordach> (then again, with the fact you're buying a M1 family to begin with, you're going to get trouble)
22:47 erlehmann well i expect the m1 aarch64 development to be good for reform2 customers (such as me)
22:47 erlehmann in some capacity
22:47 MTDiscord <Jordach> spoiler alert no
22:48 erlehmann why
22:48 sfan5 Linux on aarch64 is already solid
22:48 erlehmann oh, OS X is that bad?
22:48 MTDiscord <Jordach> the problem is booting it
22:48 erlehmann didn't know
22:48 erlehmann sorry
22:48 erlehmann wait it's bootlocked?
22:48 MTDiscord <Jordach> https://github.com/pipcet/pearl
22:49 Pexin Apple: Your Computer is really Our Computer
22:49 Pexin (also google)
22:49 rubenwardy please lets not start another flamewar
22:49 rubenwardy at least go to #minetest
22:49 Pexin sory
22:50 Pexin unrelated question: it's my understanding the IME integration only works for linux? yes/no?
22:50 MTDiscord <LandarVargan> bump on installer build question
22:50 sfan5 correct
22:50 sfan5 nobody wanted to bother with IME on windows apparently
22:51 sfan5 probably not hard to do
22:51 Pexin hm. not sure it should be hilighted in the release list then
22:51 Pexin if crossplatform then I'd definitely suggest it
22:52 sfan5 press F5
22:54 rubenwardy SDL should help with that, it has IME support
23:04 Wuzzy rubenwardy: what's a package alias?
23:04 rubenwardy Wuzzy: ContentDB uses author/name to uniquely identify a package
23:04 rubenwardy A package alias allows that key to be changed
23:04 Wuzzy how?
23:04 rubenwardy on ContentDB
23:05 rubenwardy by contacting me
23:05 fluxionary joined #minetest-dev
23:05 rubenwardy basically, it allows changing usernames on ContentDB in the future
23:05 Wuzzy Does any package use this so far?
23:05 rubenwardy not yet
23:05 Wuzzy ah
23:05 Wuzzy clever
23:05 MTDiscord <GreenXenith> re: discord pings: Its supposed to be based on usernames. Whether or not it is in practice is another story entirely 🙄
23:05 rubenwardy Mineclone2 could use this to move to a Mineclone2 user, rather than "Wuzzy"
23:06 MTDiscord <GreenXenith> Oh, ill be the first then: Can you move all of my packages to my name?
23:06 MTDiscord <GreenXenith> 😄
23:06 rubenwardy done
23:07 MTDiscord <GreenXenith> Awesome, thanks
23:40 erlehmann celeron55 here is my proposal https://mister-muffin.de/p/6WEU.html
23:41 erlehmann oh, i have a word duplication in it
23:41 erlehmann celeron55 here is my fixed proposal https://mister-muffin.de/p/6WEU.html

| Channels | #minetest-dev index | Today | | Google Search | Plaintext