Time  Nick       Message
08:11 erlehmann  i find it ridiculous that this has no core dev support. many humans have TWO HANDS! https://github.com/minetest/minetest/pull/11016/files
08:11 erlehmann  are there technical/engine limitations here?
08:57 MTDiscord  <Sublayer plank> the only limitation preventing dual wielding in the engine is the fact that PR hasn't been merged :P
09:12 MTDiscord  <Sublayer plank> seriously though from the comments it seems like people think a special condition for a second wield hand is too hardcoded to be an engine feature
09:18 erlehmann  Sublayer plank yeah but why? look how many hands you have. i would understand criticism that the patch should be able to add infinite hands!
11:37 erlehmann  i am not sure if removing the software renderer is a good idea. is it because you can always use LIBGL_ALWAYS_SOFTWARE with mesa?
11:40 erlehmann  oh my
11:42 erlehmann  again, i think that “remove lots of unused features” PR should never ever have been approved. figure it out for yourself why it makes development harder, not easier, if you remove the *example* code, that shows how the library is used, and built.
11:43 erlehmann  “removing all unit tests” is also not a good idea. i am sure *some* of them tested features that are still in the library.
11:43 erlehmann  and how are you going to know these still work?
11:45 MTDiscord  <Sublayer plank> examples used a lot of the features unused by minetest and therefor removed
11:45 erlehmann  yeah you know the responsible thing is to NOT remove the tests that tests what minetest needs
11:47 MTDiscord  <Sublayer plank> I agree removing all unit tests was a bit too much though, but irrlicht is being progressively cut away until it's gone and absorbed into the engine, it's just a bit of a transitional period until that
11:47 erlehmann  uh nno
11:47 MTDiscord  <Sublayer plank> uh?
11:48 MTDiscord  <Sublayer plank> so it was a good thing all the unit tests got removed?
11:49 erlehmann  no, the “transitionary period” story is a story that devs tell themselves. i have never once in 15 years of development see it work out if a person refactoring rather deletes tests than putting in the effort.
11:50 erlehmann  here is what will happen: there will be no examples and no tests, for the foreseeable future. every preventable breakage will only be found by playing the game.
11:50 erlehmann  the reason for that is not that it is not doable
11:51 erlehmann  but cultural
11:52 erlehmann  basically, the only way to get test cases is to make anyone refactoring stuff write test cases … or have them already
11:53 erlehmann  it never directly hurts a program to ignore testing for new features
11:53 erlehmann  it hurts later
11:53 erlehmann  for example, i came up with a strategy to let a program spit out test cases for the texture bugs and integrate them into tests
11:53 erlehmann  fat chance i'll be able to do that when there aren't any tests
11:54 erlehmann  and the original once failed because of the gutting
11:56 sfan5      nobody even knew how to run the original tests or if they worked
11:57 erlehmann  i just got something runnign
11:57 erlehmann  but i am on 5.4.1
12:00 sfan5      what?
12:00 erlehmann  all the stuff is in irrlichtmt in the past
12:00 erlehmann  don't worry, you'll get more crashes at some point
12:01 sfan5      you're still not making any sense
12:03 erlehmann  i think i can use the old test cases, largely unnmodified, to find bugs at scale for everything tested.
12:03 erlehmann  and the old example code
12:04 erlehmann  sfan5 is hecktest planning to add new tests?
12:05 sfan5      I don't think so
12:05 erlehmann  welllllllllllllll
12:05 erlehmann  but didn't hecktest want to rewrite the renderer?
12:06 sfan5      and
12:06 erlehmann  well rewriting stuff *usually* needs exhaustive test cases for everything that worked before
12:07 erlehmann  i'm not sure how else it could work
12:07 sfan5      you test it by hand
12:11 erlehmann  uh
12:14 MTDiscord  <Sublayer plank> the rendering rewrite is going to be separate to irrlicht right? like, it'd use the opengl context that hecks implemented into irrlichtmt that exposes opengl to the engine
12:15 sfan5      afaik yes
12:16 erlehmann  you know what, i think “rewriting the renderer without tests” is a horrible suggestion, but i'm not going to try and change your mind right now. figure it out for yourself. good luck.
12:17 erlehmann  rationale: you approved hecks stuff so far, so you obviously thinking the breakage is worth something (haven't seen any of that, but i know how these stories go)
12:21 sfan5      I think it's okay to do like this, yes
12:23 sfan5      more tests would undoubtly be better but you have to consider the context and resources of an open-source project
12:23 sfan5      differently worded: I'd rather have a better renderer and no tests than tests an the same bad renderer
12:28 erlehmann  i don't see how you can even approve a refactoring of that scale without any tests
12:29 erlehmann  like how would you make sure everything works … not just on the machine that hecks and core devs have
12:29 erlehmann  i mean shadows already broke stuff and however approved that obv did not have the time to test out everything
12:29 erlehmann  however → whoever
12:30 erlehmann  “I'd rather have a better renderer and no tests than tests an the same bad renderer” – this only works if reliability is of no concern
12:31 erlehmann  well, whatever, i'll keep submitting bug reports
12:32 sfan5      where would your hypothetical tests run that they can cover more than the machines devs have access to?
12:36 erlehmann  sfan5 does not matter. i just managed to instrument the irrlicht md2 example to an extent i can now autogenerate models that should be able to crash minetest.
12:36 erlehmann  the helloworld thing
12:37 erlehmann  and yes, i am aware that md2 support has been removed. i just wanted to see if the technique is viable
12:37 erlehmann  it's pretty slow though
12:37 sfan5      which fuzzer are you using?
12:37 erlehmann  afl-fuzz
12:37 erlehmann  it's magic
12:38 erlehmann  but if you have ever used, you see why “test the entire program” does not work
12:38 erlehmann  you need to have proper unit tests
12:38 erlehmann  i mean it does work … very slowly
12:38 erlehmann  by which i mean it is infeasible
12:47 pgimeno    erlehmann: any luck crashing feh?
12:48 sfan5      pgimeno btw https://github.com/minetest/minetest/pull/11615#issuecomment-918802164
12:50 erlehmann  pgimeno yes, years ago, try it yourself. it's not that hard. i'd rather have others learn that. look into how afl-fuzz works!
12:52 adfeno_    Hi there, how do I disable itens from appearing on creative inventory? Is it possible to do via settings/settingtypes? Can it be done without depending on the mods explicitly?
12:54 adfeno_    I see that there is minetest.override_item, and I haven't tested it but I'm afraid that if I were to do something such as groups={not_in_creative_inventory=1}, Minetest would erase all the groups for that item and only enable the one I used.
12:55 sfan5      I don't think it does but test it
12:55 adfeno_    (Well, not that the others matter much given they are in creative, but it is still worrying)
12:59 adfeno_    Also, does minetest.registered_items include all other minetest.registered_{nodes,craftitems,tools} ? Or are them all separated sets?
13:00 sfan5      a {node,tool,craftitem} is also an item
13:00 sfan5      so yes
13:01 adfeno_    Hm I see, thanks ;)
14:12 pgimeno    sfan5: updated #11615 with Wuzzy's suggestion, but note it's probably easier for you to amend the commit than to pull the changes
14:12 ShadowBot  https://github.com/minetest/minetest/issues/11615 -- [MANUAL MERGE] Add an option `-t` to force text output in /help by sfan5
15:15 erlehmann  pgimeno, why do you write “[MANUAL MERGE]”
15:16 pgimeno    I don't, read the post
15:30 erlehmann  > don't squash when merging to preserve authorship
15:30 erlehmann  uh
15:30 erlehmann  why would anyone squash merge by default
15:31 erlehmann  i mean i have seen it already, but you need extreme discipline among committers and tiny PRs to have that work
15:31 erlehmann  otherwise it basically breaks any use of git revert and git bisect that are meaningful
15:32 erlehmann  i mean … the people who do that usually don't use bisect
15:32 erlehmann  so it just makes it horrible for someone else years later
15:36 erlehmann  i mean, it is kinda self-enforcing in a way
15:36 erlehmann  if you squash merge stuff that makes huge commits, git-bisect is useless and then anyone can say “why should we change it, it was never useful so far”
15:38 erlehmann  i mean, take irrlichtmt feature removal. if hecktest had made many small commits, you could give the unit tests to git bisect and let it figure out which commit breaks what and then decide if you want to revert it.
15:38 erlehmann  but right now bisect can only say “this big commit broke stuff” lol
15:38 erlehmann  which is something we already know
16:00 MTDiscord  <josiah_wi> I try to discipline myself to make small commits in order to preserve the specific purpose of the different changes, so I don't particularly like the squashing, but I don't think complaining about it would help anyone. I'm not that good at making commits yet; I still have to push fixes for typos etc. that clutter the commit tree. Maybe if we all worked together to be professional in our commits and develop a unified commit style,
16:00 MTDiscord  people would hear us out about whether we want them squashed or not.
16:01 erlehmann  josiah_wi if you learn to interactive rebase it helps a lot
16:02 erlehmann  it is a pain to amend commits otherwise
16:02 erlehmann  by now i usually interactively rebase my stuff so the typos were never there hehe
16:03 pgimeno    @josiah_wi if you want your PRs to preserve the commits, add [NO SQUASH] to the title, but be ready to manually squash the commits resulting from the review
16:03 erlehmann  josiah_wi if you want to lie about the history that it never contained any typos (which i think is ok), read this: https://www.sitepoint.com/git-interactive-rebase-guide/
16:04 erlehmann  but this is *never* ok on master, main or any other branch anyone relies on
16:04 erlehmann  not even if you warn them
16:04 erlehmann  so use it on your own branches to clean up or reorder commits only
16:05 erlehmann  and by “never ok” i mean, ofc you can get the consent of all involved, but that means “everyone who ever checked this branch out”, so it only works on a very short-term basis usually
16:05 pgimeno    @josiah_wi examples: #11253, #11262, #11289, #11584
16:05 ShadowBot  https://github.com/minetest/minetest/issues/11253 -- [NOSQUASH] CMake/Gui: organizational changes + some refacto commits by nerzhul
16:05 ShadowBot  https://github.com/minetest/minetest/issues/11262 -- [NO SQUASH] Joystick sensitivity for player movement by NeroBurner
16:05 ShadowBot  https://github.com/minetest/minetest/issues/11289 -- Minor fixes (don't squash) by sfan5
16:05 ShadowBot  https://github.com/minetest/minetest/issues/11584 -- [DONT SQUASH] CMake / CI maintenance by sfan5
16:05 erlehmann  ((it breaks git pull)
16:07 erlehmann  so next time you want to fix a typo on a branch (i.e. the typo is not on master yet), just use interactive rebase and rewrite the history from the point where it diverged from master
16:07 erlehmann  i mean rebase is useful in general
16:07 erlehmann  but many ppl submit PRs where at least a few of them are “fix typos” or “remove whitespace”
16:07 erlehmann  you know
16:09 MTDiscord  <josiah_wi> I'm pretty new to open source contribution yet, and working with other devs in general, but it seems like unity and consistency is important to work well together. Putting no squash on all my commits seems like it would abuse the the project's intended use of no squash, and make me stand out as doing things my own way.
16:09 pgimeno    not on all your commits, but in your PR
16:10 erlehmann  josiah_wi just do it, i doubt anyone will hold back telling you to do it differently if they think it is wrong.
16:11 pgimeno    Minetest's PRs are most often single-commit features, with the subsequent commits being amendments arising from the review. The natural action in these cases is to merge squashing.
16:12 erlehmann  pgimeno that is what i meant with discipline
16:12 pgimeno    erlehmann: why don't you read about minetest's procedures before ranting on what you ignore then?
16:15 MTDiscord  <josiah_wi> I've heard that the Minetest code is a mess, and I've observed and heard about a lot of disagreement that didn't seem professional to me. I've also heard that it's difficult for new developers without specific skills to get involved. I don't think it has to be this way. For this reason I don't intend to request PRs not be squashed that could perfectly well be squashed.
16:16 erlehmann  josiah_wi if you think it could be squashed, then do it!
16:16 erlehmann  sometimes development history is important, but projects differ on it.
16:20 pgimeno    I mean, you *do* want this squashed, for example: https://github.com/minetest/minetest/pull/10173/commits
16:20 pgimeno    it's most often easier for reviewers to look at what changed with respect to the previous commit, than to read the whole patch from the beginning again
16:21 MTDiscord  <josiah_wi> Slight change of topic, but how fast do you think we could get a working clang format script that we agree new commits should conform to, except by special exception; I know some people seem to feel strongly that there are times you should ignore the formatter if it improves readability. This doesn't include refactoring existing code.
16:22 rubenwardy josiah_wi: minetest's code style is way too particular, and there's way too much code to go and find all the edge cases
16:25 MTDiscord  <josiah_wi> So you're saying clang format isn't capable of representing our code style? Perhaps we could agree to remove clang format altogether.
16:26 MTDiscord  <josiah_wi> Sorry, that was rude. I'm just looking for a clarification of your point.
16:27 erlehmann  pgimeno you are right, i should read up on stuff before going on anyones nerves
16:28 erlehmann  pgimeno i guess i have focussed too much on the negatives bc the project is full of holes, i should look at what is working!
16:32 MTDiscord  <josiah_wi> Discussing holes doesn't have to be negative. Finding bugs is an opportunity to make the project even better, no?
16:34 erlehmann  josiah_wi the point is i am at the same time poking at the wounds while not knowing enough about How Its Done™ and asking around, while the first is helpful, the second one is not necessarily
16:35 erlehmann  i.e. if i want ppl to read what i write i should not waste their time
16:36 pgimeno    @josiah_wi yes but following due process, i.e. submiting bug reports, not ranting
16:37 erlehmann  just found a b3d file that seems to hang minetest 5.4.1 clients, bug report coming soon
16:44 sfan5      we don't take bug reports for 5.4.1
16:44 sfan5      unless the same bug also exists in 5.5-dev of course
16:45 sfan5      (don't take that literally, I'm just mentioning it because I know you have the ability to test on 5.5)
16:46 sfan5      pgimeno: pulled
16:48 MTDiscord  <josiah_wi> In any case, if I can get people interested I'd like to demonstrate that it's possible to clean things up (and I mean actually cleaning, not just axing stuff) by coming to a decision about clang format and implementing it. The benefit from having a code style tool comes from being able to double check your style and format it very quickly. We already have a code style, and to my knowledge style guidelines for contributing aren't
16:48 MTDiscord  supposed to be optional. It doesn't have to be clang format, and we don't even need to have a style tool, but having a broken one sitting around taking up space seems a shame.
16:50 erlehmann  sfan5 look i have very little energy rn, very little time … and the build system is extremely not helping. i am currently recompiling stuff, but i will report that bug probably against 5.4.1 and deliver a very simple test case you can stuff into your test client.
16:50 sfan5      that's okay
16:51 erlehmann  like a mod that hangs clients
16:52 erlehmann  also my bug-finding setup relies on irrlicht having its example code intact, so … well …
16:53 erlehmann  i'll probably ruin it if i build 5.5-dev rn
16:53 sfan5      I doubt yet setup requires all 30 or so example apps from Irrlicht to work
16:53 sfan5      your*
16:53 sfan5      https://0x0.st/-3Hm.png github does not like my usecase >.>
16:55 erlehmann  sfan5 i basically chose the first example and told afl-fuzz to work its magic on it – and that means, it needed to actually compile and be able to load the example b3d model. since the example code, the b3d model and probably a huge chunk of irrlicht that would make it compile … have been removed … i don't see myself doing that for any newer version.
16:56 erlehmann  sfan5, btw, removing the software renderer was entirely legit, it sucks big time
16:56 erlehmann  (just so you know i do agree with gutting stuff that does not work correctly)
16:57 erlehmann  (and does not have a superior alternative, like LIBGL_ALWAYS_SOFTWARE=1)
16:57 sfan5      https://github.com/minetest/irrlicht/blob/master/examples/AutomatedTest/main.cpp#L48-L50 loading a model on irrmt is just as easy
16:59 erlehmann  sfan5 i see, thx. is this the only test case that remains?
16:59 sfan5      yes
17:00 erlehmann  why?
17:00 erlehmann  what makes it so special?
17:00 erlehmann  is it a collection of all things that should still work?
17:00 sfan5      we wrote it ourselves
17:00 erlehmann  ah ok
17:00 MTDiscord  <Sublayer plank> it's automatically tested by CI on each commit
17:01 erlehmann  nice. if i wanted to extend that (in particular, make it so that there are X tests for X features so they can be instrumented), where to start reading up on it?
17:01 erlehmann  like what is the command the CI runs?
17:02 erlehmann  (if it is a shell script, i could start there)
17:02 sfan5      https://github.com/minetest/irrlicht/blob/master/.github/workflows/build.yml#L50
17:02 erlehmann  thx
17:02 MTDiscord  <Sublayer plank> https://github.com/minetest/irrlicht/blob/master/.github/workflows/build.yml
17:02 MTDiscord  <Sublayer plank> oh whoops
17:02 MTDiscord  <Sublayer plank> ninja'd :P
17:03 sfan5      it tests three extremely basic things right now, I don't see how you could easily expand it to test e.g. graphics related stuff
17:03 sfan5      where "tests" = doesn't crash and maybe it even checks the result
17:03 erlehmann  sfan5, why would it not be easily expandable? have a bunch of binaries that either render a single frame or crash?
17:04 sfan5      well how do you check the frame for correctness
17:04 erlehmann  well, that's a question that comes after “does it even run before crashing”
17:05 erlehmann  i am mostly concerned with someone ripping out something and not noticing
17:05 erlehmann  if they have to delete the corresponding test case for a feature to be removed so the CI lets them, then they DO know.
17:06 sfan5      in what way? if you remove a function definition code that does use it won't compile
17:06 sfan5      and the code that uses it here would be in Minetest
17:07 erlehmann  there are lots of ways stuff can compile and still crash. i would, for example, advocate to put in each media file that causes a crash or hang as a test case.
17:07 erlehmann  to make sure refactorings do not reintroduce the bug
17:07 sfan5      that's not a bad idea but there are much more relevant things that could be tested
17:07 erlehmann  i have no good answer to “how to verify the frame” btw
17:08 erlehmann  what would you want to test?
17:08 sfan5      take for example spritesheets animations, they are currently broken in Irrlicht 1.9 (presumably they don't have a test case for them either)
17:09 sfan5      as far as I have looked at it the bug is "just" forgetting to load the texture matrix
17:09 sfan5      so you'd want to catch such basic things with a test that sets up an animated scene nodes, renders, steps time and verifies changes
17:11 sfan5      pgimeno: btw you forgot to account for CSM, I'll be fixing this myself https://0x0.st/-3H_.txt
17:12 erlehmann  sfan5, obv you have a better grasp on that than i do
17:12 sfan5      if you think I know lots (or even a medium amount) about OpenGL you're mistaken
17:13 erlehmann  don't worry, i already assume everyone (including me and the original authors of irrlicht) has no idea what they are doing most of the time
17:41 pgimeno    sfan5: maybe csm should have a get_player_by_name()? (obviously not within the scope of the PR, I know)
17:42 pgimeno    I have never used CSM so I don't really know about its limitations, sorry about that
17:42 MTDiscord  <luatic> I actually prefer it if clients can't distinguish players and entities
17:45 erle       luatic they need to for stuff like teamchat
17:49 erlehmann  pgimeno you can look at some of the battle-tested CSMs here: https://repo.or.cz/waspsaliva.git/tree/HEAD:/clientmods
17:51 MTDiscord  <luatic> erlehmann: no they don't need to
17:51 erlehmann  luatic how are you going to do “send this message to players near me” then?
17:51 MTDiscord  <luatic> not as a client mod
17:52 erlehmann  you know what
17:53 erlehmann  if you can't distinguish between players and other entities then one really important use case for CSMs becomes dangerous
17:53 erlehmann  “punch every entity”
17:53 erlehmann  used for a) killaura b) better item pickup on laggy servers
17:53 erlehmann  i have seen the second use case much more
17:54 sfan5      ruining cheat client users' ability to do thing makes implementing suddenly much more appealing
17:54 sfan5      +this
17:54 erlehmann  killaura can easily be detected though
17:55 erlehmann  if someone punches inhumanely fast, cheater
17:55 erlehmann  for an example of how it is done btw: https://repo.or.cz/waspsaliva.git/blob/HEAD:/clientmods/autoaim/init.lua
17:55 cora       hey guys :)
17:55 erlehmann  > for k, v in ipairs(minetest.localplayer.get_nearby_objects(10)) do
17:55 sfan5      the player distinction doesn't matter that much really, no normal entity will have a texture called character.png
17:55 erlehmann  > if (v:is_player() and v:get_name() ~= minetest.localplayer:get_name()) then
17:56 MTDiscord  <luatic> sfan5: but they can
17:56 cora       i cant comment on discord: i'm just honestly interested: for what exactly is vanilla csm supposed to be used ?
17:56 MTDiscord  <luatic> I was planning anticheat as tricking cheat clients into punching fake player entities
17:56 MTDiscord  <luatic> but of course that's not gonna work if players can simply distinguish the two
17:56 sfan5      it wasn't intended for anything really, it just kinda exists
17:56 sfan5      SSCSM was the plan from the beginning but about following plans...
17:57 sfan5      what it can be reasonably used for is chat-enhancing mods
17:57 MTDiscord  <luatic> SSCSM is probably bad for anticheat too...
17:57 cora       oh yeah ok
17:57 cora       that makes some sense i guess
17:58 cora       fair (/re chat enhancements)
17:58 erlehmann  sfan5, would you say fried and enemy lists and autocoloring their nametags client side (red / blue) would be good?
17:58 MTDiscord  <luatic> Well, every time something is offloaded to the client, cheaters benefit a bit.
17:58 Pexin      theres also waypoint/teleport help stuff that only works if the player has appropriate privs
17:58 cora       teamchat prob works 90% on vanilla
17:58 erlehmann  i bet faction servers would LOVE that
17:58 MTDiscord  <luatic> I prefer it if red/blue team is clear.
17:59 MTDiscord  <luatic> Absolute colors instead of relative ones that is.
17:59 sfan5      erlehmann: my general stance is that it's good if the server admin is okay with people doing that
17:59 erlehmann  luatic you are obviously not thinking of cases where there are no official teams
17:59 MTDiscord  <luatic> Also note that with a modded engine, that should already be possible ;)
17:59 cora       i agree
17:59 erlehmann  sfan5, we already have haxnotify for that
17:59 MTDiscord  <luatic> You can just send different nametag colors out to different players
17:59 cora       well the custom version string is prob more helpful tbh
17:59 erlehmann  sfan5 waspsaliva chats “HEY EVERYONE I AM USING A HACKED CLIENT”
17:59 cora       you can just disable the haxnotify csm
18:00 MTDiscord  <luatic> Yeah, waspsaliva might be a well made client with (minor) abuse protection
18:00 MTDiscord  <luatic> But there's other clients
18:00 cora       i was actually thinking of doing sth more sneaky
18:00 MTDiscord  <luatic> Please do
18:00 erlehmann  i think “let's not enhace vanilla bc cheaters might abuse it” is a ridiculous stance, cheaters can always have a better client. the protection needs to be on the server, so the client does not matter.
18:00 cora       but knowing server admins they wouldnt even implement it :P
18:01 erlehmann  cora assuming i know what you mean, it would be defeated the moment server admins knew how to detect it
18:01 cora       yes that would be ideal (doing it on the server)
18:01 MTDiscord  <luatic> I'm not saying "let's not enhance vanilla", I'm saying "enhance vanilla but keep server authority", because that makes implementing serverside "protection" as you call it easier.
18:02 erlehmann  who is harmed by users having friend/enemy lists on servers?
18:02 erlehmann  or team chat
18:02 cora       yeah you might be right
18:02 erlehmann  based on that
18:02 cora       its totally based on obscurity
18:03 cora       uh according to the api.txt get_player_names() exists in vanilla though
18:04 Pexin      I have made debugging CSMs that, for example, query the light level at a specified node
18:04 erlehmann  sfan5 coming from an IT security perspective, the server admin has no business knowing who is on whose friend or enemy list and it represents a liability (if it gets leaked, stuff is weird)
18:04 cora       o yeah for debugging it can be p helpful - even more so if you can interact with entities and items ^^
18:05 Pexin      is it possible for CSM to find entities? I must have missed that
18:05 erlehmann  btw, if vanilla had the capabilities of waspsaliva, one could make csms that speedrun the game basically
18:05 erlehmann  to figure out if it is still working
18:05 erlehmann  on each commit
18:05 cora       Pexin,  no
18:05 cora       thats basically the 2 big things we added to csm api
18:06 erlehmann  Pexin you should obv use hax
18:06 Pexin      there have been several times I wanted to make a CSM that could insert player actions
18:07 Pexin      having a whole optional bot api would be sweeeet. "this server is for botwars"
18:08 cora       https://repo.or.cz/waspsaliva.git/blob/HEAD:/src/script/lua_api/l_clientobject.cpp
18:08 cora       https://repo.or.cz/waspsaliva.git/blob/HEAD:/src/script/lua_api/l_inventoryaction.cpp
18:09 erlehmann  sfan5 look you can just IFDEF the cheaty actions and never make a release with it. will help much more with debugging than it will enable cheating.
18:11 erlehmann  sfan5 also it would make developing anticheat much easier
18:12 sfan5      what are you trying to convince me of
18:13 sfan5      minetest already has "cheat" features hidden behind the debug privilege
18:13 cora       to merge all our cheaty cpp stuff so we can do everything in lua :P
18:13 erlehmann  oh right
18:13 erlehmann  yes i'd definitely use it for CI purposes
18:14 erlehmann  sfan5 i see, hiding the APIs behind the debug priv would be much neater
18:33 erlehmann  sfan5 https://github.com/minetest/minetest/issues/11633
18:35 sfan5      it indeed still happens on 5.5
18:36 erlehmann  sfan5, pls add that to the ticket then
18:37 erlehmann  a good result of the file format massacre: less stuff to break
18:39 MTDiscord  <luatic> What's up with "You can't comment at this time. "
18:39 MTDiscord  <luatic> (on GitHub)
18:39 erlehmann  where?
18:40 MTDiscord  <luatic> Your bad3d isssue
18:40 MTDiscord  <luatic> anyways, I'm just gonna say it here:
18:40 MTDiscord  <luatic> This is interesting. I had to try modlib's B3D reader (which is written in Lua) at it, and it crashed as expected. It seems the file contains a chunk of invalid type. I suppose Irrlicht doesn't handle this case.
18:40 erlehmann  maybe bc sfan5 moved it to irrlichtmt?
18:40 MTDiscord  <luatic> ohh
18:40 erlehmann  i bet github stupid
18:41 MTDiscord  <luatic> apparently yes
18:41 erlehmann  pls comment it there as well
18:41 erlehmann  for future reference
18:43 erlehmann  well i can put it there
18:43 erlehmann  or not
18:43 erlehmann  You can't comment at this time
18:43 erlehmann  lol
18:43 erlehmann  ok now
18:43 MTDiscord  <luatic> lol
18:43 erlehmann  i put both sfan5 and luatic comments there
18:44 erlehmann  appguruea will be sued for plagiarizing luatic
18:44 erlehmann  or are you the same person
18:44 MTDiscord  <luatic> lol
18:45 MTDiscord  <luatic> I am lol
18:45 erlehmann  anyone could say that
18:45 MTDiscord  <luatic> indeed
18:46 appguru    But what about my registered IRC account saying it?
18:47 erlehmann  ok i believe you
18:48 erlehmann  i guess your irc account will not be demonetized today
18:48 Pexin      we are all Kosh
18:51 MTDiscord  <luatic> Irrlicht's B3D reader has a gross length if you ask me: About 1k lines. For reference, the original Blitz loader is a mere 300 something lines of not even C++, but plain C: https://github.com/blitz-research/blitz3d/blob/master/blitz3d/loader_b3d.cpp
18:51 MTDiscord  <luatic> But it seems they do check for invalid chunk types.
18:52 erlehmann  there is only one way to be sure, full recognition before processing
18:52 erlehmann  length does not matter, rather complexity
19:17 rubenwardy <cora> i cant comment on discord: i'm just honestly interested: for what exactly is vanilla csm supposed to be used ?
19:17 rubenwardy https://dev.minetest.net/Client_scripting_plans
19:17 rubenwardy (written pre CSM)
19:17 cora       oh thanks
19:18 erlehmann  > Unlike with Web browsers, Lua code should run in a separate thread. Also, to be completely separate, it shouldn't share the Lua environment with the mainmenu Lua (to forestall intra-environment exploits).
19:18 erlehmann  good luck
19:18 erlehmann  > there will be no way (without changing C++ source code) to execute code outside of that sandbox
19:19 erlehmann  [press x to doubt]
19:20 erlehmann  good that you do not want to send bytecode though
19:20 erlehmann  that would make it much easier to exploit
19:20 rubenwardy bytecode is already blocked by both sandboxes
19:22 rubenwardy there are number of issues discussing various aspects of introducing ~~RCEs~~ SSCSMs
19:22 Krock      sfan5: what would you think about using DISABLE_CLASS_COPY in Buffer, and introducing an explicit copyTo() function?
19:23 Krock      in regard to #11607
19:23 ShadowBot  https://github.com/minetest/minetest/issues/11607 -- Shave off buffer copies in networking code by sfan5
19:24 erlehmann  rubenwardy as i said, if the server starts to execute code on the client, the client should send code back
19:24 erlehmann  just make them equals
19:25 rubenwardy the server-side sandbox isn't designed to be completely secure - it has the insecure environment, for example
19:26 rubenwardy clients are also less trusted in terms of making gameplay decisions, and servers can contain sensitive data like IP addresses and SRP hashes
19:27 erlehmann  oh i don't mean the sandbox
19:28 rubenwardy anyway, SSCSM is notably missing from the roadmap for a reason. It will be hard to get it right, and there's more important stuff with greater benefits to do first
19:28 rubenwardy for example, I should finally finish that Android PR
19:28 rubenwardy as soon as I can be bothered to Java
19:30 Krock      writing patch.....
19:30 cora       cant wait for sscmsm ^^
19:30 erlehmann  Krock, for what?
19:30 Krock      erlehmann: last message
19:32 Krock      found one more class copy
19:39 erlehmann  quick question about this, am i really weird or is “-2147483648 >= 1” a normal thing to expect from C++? https://github.com/minetest/minetest/issues/11620#issuecomment-921175826
19:44 Krock      afaik it's undefined behaviour for signed numbers, but in terms of binary representation it's not surprising at all
19:45 erlehmann  Krock -2147483648 is negative. 1 is positive. how can it be bigger?
19:49 rubenwardy quick maths
19:50 Krock      erlehmann: "i + INT_MAX" standalone evaluates to integer, however within the comparator they seem to use 64-bit registers
19:50 Krock      or it's because the overflow bit is set
19:52 appguru    I'd appreciate it if #11427 was looked at, it's still very much ready for review and IMO mergeable.
19:52 ShadowBot  https://github.com/minetest/minetest/issues/11427 -- Redo serialize.lua by appgurueu
19:55 erlehmann  Krock the actual answer is something else entirely, the optimizer is allowed to do anything and just randomly chooses this branch
19:55 erlehmann  Krock, by which i mean it can change based on moon phase or whatever
19:55 Krock      then it's not a good RNG
19:55 erlehmann  “randomly” as in, you won't be able to guess it
19:56 Krock      then why is it always the same branch? :P
19:56 erlehmann  Krock, the other one is optimized out
19:56 erlehmann  https://blog.regehr.org/archives/140
19:56 erlehmann  > This loop only terminates if it finds a counterexample to a special case of Fermat’s Last Theorem.
19:56 erlehmann  > It turns out that when optimizations are enabled, several compilers (LLVM/Clang 2.7, Open64-x86 4.2.3, Sun CC 5.10, and Intel CC 11.1.072) go ahead and permit this loop to terminate.
19:57 erlehmann  > Faced with this incredible mathematical discovery, I held my breath and added a line of code at the end of the function to print the counterexample: the values of a, b, and c.  Unfortunately, with their bluffs called in this fashion, all of the compilers emitted code that actually performed the requested computation, which of course does not terminate.
19:57 Krock      ah yes. can confirm with Ghidra
19:57 erlehmann  Krock, as soon as you compile my example code with “-fsanitize=undefined”, the compiler chooses the other branch
19:58 erlehmann  i have to go
19:58 Krock      https://i.postimg.cc/vH8BZcGC/grafik.png
19:58 erlehmann  will be back in 1 hour or more
19:59 erlehmann  Krock obv neither of us knows basketball basketball
20:16 Krock      dear lord
20:17 Krock      minetest deep copies all timed out packages
20:24 sfan5      Krock: I think that would break stdlib containers
20:25 MTDiscord  <Jordach> you could run it in browser with the compiler explorer right
20:28 sfan5      appguru: "if a priority queue was used to implement minetest.after" doesn't it do that?
20:28 appguru    No it does not do that
20:28 Krock      sfan5: https://krock-works.uk.to/u/patches/0001-Delete-copy-constructor.patch
20:28 Krock      did a quick test and it seems to work okay so far
20:29 appguru    minetest.after is implemented in Lua in a straightforward way: see https://stackoverflow.com/a/8236407/7185318
20:29 appguru    it simply keeps an unordered list of all callbacks and iterates it in reverse order
20:29 Krock      it's easier to search for copyTo()
20:30 sfan5      ah if it's for searching maybe marking with __attribute__((deprecated)) works
20:30 sfan5      it should throw a warning on every use
20:31 Krock      deprecating an operator overload and constructor? that's an interesting idea
20:32 sfan5      +		return std::move(m_event_queue.pop_front(timeout_ms));
20:32 sfan5      I remember reading that is this pointless (if not counterproductive) because copy elision already does this
20:34 Krock      oh right
20:34 Krock      I wonder why I got errors on that previously... no idea what's changed now
20:34 sfan5      since I just looked for the link anyway here's the thing https://en.cppreference.com/w/cpp/language/copy_elision
20:35 Krock      good to know. thank you!
20:36 sfan5      @luatic I think you linked the wrong SO answer there
20:36 Krock      confirmed that the patch works by connecting and playing on Blocky Survival
20:43 Krock      I updated the patch file accordingly (Ctrl + F5 reload). Feel free to add it to your PR, or I'll open a separate one afterwards
20:45 sfan5      separately is more convenient for me
20:45 sfan5      ;)
21:02 rubenwardy disabling the copy constructor makes more sense than deprecated it
21:04 sfan5      sure, the latter is just a less intrusive way of discovering usage
22:23 erlehmann  Krock, do you have a test that proves/suggests this thing behaves the same afterwards as before or that it fixes some error?