Minetest logo

IRC log for #minetest-dev, 2014-10-28

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

All times shown according to UTC.

Time Nick Message
00:31 kahrl joined #minetest-dev
00:41 shadowzone joined #minetest-dev
00:50 Chicken_shadow joined #minetest-dev
00:56 Zeno` joined #minetest-dev
01:02 paramat joined #minetest-dev
01:13 chchjesus joined #minetest-dev
01:20 paramat hmmmm, mgv5 makechunk time is 110ms with 1 eased noise (terrain only), 150ms with all 5 noises eased, 130ms with 3 eased noises (terrain, cave1, cave2) (10ms per easing). any preferences? i think 3 eased is a good balance, being faithfull to the original's caves but not easing dirt/lava blob shape and wetness
01:22 paramat https://cdn.mediacru.sh/eVCExlDvoe29.png
01:25 kahrl hmmmm: did c55 agree with that license change?
01:27 kahrl and I guess PilzAdam would have to agree too as he contributed the faster version of easeCurve
01:29 kahrl proller also changed a tiny bit
01:40 mos_basik joined #minetest-dev
01:41 Zeno` Hi kahrl. I had to manually apply that commit in my refactor_the_game branch (like you warned me lol)
01:42 Zeno` But I didn't know how to make it look like you wrote it :(
01:44 kahrl no worries
01:45 Zeno` I referenced your commit though
01:45 kahrl (it'll remain in the minetest/master commit history regardless, since it's already been pushed)
01:45 Zeno` yep, was just letting you know. I wanted to "credit" you with it, I just couldn't work out how
01:47 kahrl the way to do that would probably be via the --author option of git commit
01:47 kahrl but don't worry about it
01:47 Zeno` ok
02:07 paramat i think the license of c55's mgv5 branch might just have been the result of copy-paste, it seems to need editing with credit also to c55 etc.
02:11 hmmmm kahrl, of course.
02:12 hmmmm with regard easeCurve(), that was my very first contribution to minetest
02:12 hmmmm pilzadam did commit it for me and all
02:12 paramat ah okay
02:12 hmmmm proller removed all of his noise-related modificiations when he rage quit
02:12 hmmmm so it's me and celeron basically.
02:15 OldCoder I had a memory corruption crash just now. Obviously no way to trace it. But are occasional such corruptions known to occur?
02:15 OldCoder I    silvercrab.log (Modified)   Row 10783 Col 24   7:08  Ctrl-K H for help
02:15 OldCoder *** glibc detected *** minetestserver: malloc(): memory corruption (fast):
02:16 OldCoder This sort of error; has it been seen sometimes?
02:54 Zeno` ShadowNinja, I have amended the brace issue, answered your question, and changed all of the sizeof operators to not use () unless necessary
02:54 Zeno` I find sizeof(object) as annoying as return(object)
02:55 Zeno` As for the function like macro expansion... see line note ;)
03:11 zat joined #minetest-dev
03:37 Zeno` Someone should join linuxgaming and watch their RAM usage go up without ever stopping
03:38 Zeno` just walk around doing nothing and watch it rise and rise... I've just hit 5GB usage in.. hmmm about 8 minutes
03:39 Zeno` make that 6GB now
03:42 Zeno` and 7
03:48 hmmmm joined #minetest-dev
04:12 paramat left #minetest-dev
05:03 Miner_48er joined #minetest-dev
05:28 Zeno` I have bad news :(
05:28 Zeno` The mem leak... can't find it
05:29 Zeno` It happens on all servers not just linuxgaming as I thought before; it's just that it seems to happen faster on linuxgaming
05:30 Zeno` bisect time I guess
05:34 hmmmm mem leak?
05:35 hmmmm valgrind no workie?
05:35 sol_invictus joined #minetest-dev
05:35 Zeno` yes, client mem usage just keeps going up and up and up. valgrind shows nothing interesting
05:35 hmmmm i'm willing to bet it's my own shitty code to blame
05:35 Zeno` it probably wouldn't if the mem was finally released anyway
05:35 Zeno` I'll do the bisect soon
05:36 hmmmm did you check to see if it's caused by VBOs
05:37 Zeno` not yet
05:37 Zeno` I wanted to confirm if it was actually happening and not my imagination first
05:38 Zeno` Would VBOs make normal RAM increase?
05:39 Zeno` I suppose they could depending on what opengl does with them
05:39 chchjesus joined #minetest-dev
05:39 hmmmm yes
05:39 hmmmm VBOs cause a dramatic increase in memory usage, the price to pay for an incredible FPS boost
05:39 hmmmm not by their very nature, but because of irrlicht sucking ass
05:40 Zeno` ok, will keep that in mind while searching
05:40 Zeno` it's just funny that I've never noticed this before... i.e. I would have been checking RAM way back...
05:40 Zeno` umm... can't find the commit
05:40 hmmmm having VBOs enabled will make the client look like it's leaking memory, but it isn't, i promise.
05:43 Zeno` I can disable them?
05:48 hmmmm yeah you should be able to in the config, i forget the name of the setting
06:06 Zeno` stupid irrlicht
06:07 Zeno` since the client is using 15GB of RAM I'd really want fps to be 1000 hehehe
06:08 hmmmm crazy!
06:08 hmmmm again, we're not sure if this is the problem or not
06:12 Zeno` hmmmm, you've made a terrible mistake! https://github.com/minetest/minetest/commi​t/bc28ca0636f5bce85683a7114ad99df4e7791847​#diff-5c9fad38a1e2b7a0227fd3f5282dcc09R534    You're missing a tab
06:13 hmmmm what do you mean
06:13 Zeno` needs two tabs! Da style!
06:13 hmmmm for line continuations?
06:13 Zeno` apparently. I think
06:14 hmmmm oh, I never knew that
06:14 hmmmm i always did one tab
06:15 Zeno` Maybe it doesn't count for line continuation but just when breaking conditionals
06:16 hmmmm i don't know man.
06:16 Zeno` neither do I
06:16 Zeno` it's... a conundrum
06:16 hmmmm I just added a lot of flack to a major .h file
06:17 Zeno` Despite this VBO complication I might keep hunting for a bit, just in case
06:19 Zeno` Btw, I was joking (mostly) about the missing tab
06:20 Zeno` My sense of humour if definitely warped
06:20 hmmmm oh
06:21 Zeno` I think many people do not "get" (some) Australian's sense of humour. Gets me in trouble. I will try and refrain
06:21 hmmmm you're australian?
06:23 Zeno` I am
06:36 nore joined #minetest-dev
06:50 darkrose joined #minetest-dev
06:54 hintss joined #minetest-dev
06:59 Zeno` a fellow Australian
07:03 RealBadAngel https://github.com/minetest/minetest/pull/1778
07:03 RealBadAngel hi
07:05 RealBadAngel hmmmm, can you take a look?
07:05 RealBadAngel ShadowNinja, above adds wallmounted for meshnodes as you wanted
07:06 RealBadAngel theres also config option to enable/disable mesh caching at startup
07:16 Zeno` RealBadAngel, https://github.com/minetest/minetest/pull/1777
07:17 Zeno` Also: https://github.com/minetest/minetest/pull/1776
07:17 kilbith joined #minetest-dev
07:19 Zeno` RealBadAngel, also (for testing): https://github.com/minetest/minetest/pull/1756
07:20 RealBadAngel im ok with 1776 and 1777, but why removing brackets in sizeof ?
07:20 Zeno` same reason as you don't do return(0)   (sizeof is an operator not a function)
07:21 Zeno` but I can revert those sizeof changes if it's a big deal
07:21 RealBadAngel i was never thinkin of it this way, brackets seems here quite natural, thats all
07:21 Zeno` I guess
07:21 Zeno` I dunno...
07:22 Zeno` it's probably a pet peeve of mine and if it stops something being merged I should get over it
07:23 RealBadAngel whatever SN decides
07:24 RealBadAngel hes more into code style than me
07:25 RealBadAngel btw, about memory usage, caching meshes can take up to 0.5gb in case of dreambuilder
07:25 RealBadAngel so i decided to add option to disable caching for systems with less ram
07:26 Zeno` I think I'll apply your patch and test
07:27 RealBadAngel caching is disabled by default
07:27 RealBadAngel enable_mesh_cache = true to enable it
07:28 Zeno` hmm, how do I get the raw git diff again?
07:31 Hunterz joined #minetest-dev
07:32 ImQ009 joined #minetest-dev
07:48 Amaz joined #minetest-dev
08:00 chchjesus joined #minetest-dev
08:01 Zeno` RealBadAngel, tested #1778 playing game (minimal RAM usage increase as you mentioned) and I noticed it seemed smoother. Also checked with valgrind with no issues. So without a formal code review it looks fine
08:01 ShadowBot https://github.com/minetest/minetest/issues/1778 -- Add option to enable mesh caching, add wallmounted for meshes. by RealBadAngel
08:20 chchjesus__ joined #minetest-dev
08:23 chchjesus__ joined #minetest-dev
08:27 chchjesus joined #minetest-dev
08:33 Amaz joined #minetest-dev
09:17 FR^2 joined #minetest-dev
09:22 Zeno` Why are there commits happening without pull requests when there are 107 pull requests?
09:27 sol_invictus I wonder too, Zeno
09:38 kilbith because the commits without PR comes from the official devs
09:38 kilbith which are trustable
09:38 GrimKriegor joined #minetest-dev
09:40 Zeno` well you'd hope so
09:41 Zeno` but... don't the core devs need approval from another core dev as well?
09:41 Zeno` That is what I was led to believe
09:41 VargaD_ joined #minetest-dev
09:41 Zeno` And doesn't answer the question why there are 107 pull requests just left to rot
10:02 jin_xi joined #minetest-dev
10:12 ImQ009 joined #minetest-dev
10:13 kilbith Zeno` : ask to proller
10:40 CraigyDavi joined #minetest-dev
10:47 rubenwardy joined #minetest-dev
10:51 CraigyDavi` joined #minetest-dev
10:54 CraigyDavi`` joined #minetest-dev
10:59 jin_xi ShadowNinja: i do not agree with your point regarding lag for https://github.com/minetest/minetest/pull/1737
10:59 jin_xi its not a problem and users do deal with the very same lag for many gui interactions
11:01 jin_xi also it adds a very useful feature with a lot of potential for many mods
11:05 Zeno` yeah but the spacing/formatting is all wrong :(
11:06 jin_xi that is fixable with some helpful comments on the pull ;)
11:07 Zeno` well, it looks good to me *shrug*
11:08 Zeno` I don't think lag is an issue either... it's already an issue in lots of other places as you pointed out
11:09 jin_xi also its not to bad, i made a tetris in a formspec mod and its absolutely playable in single player
11:09 jin_xi some lag, some lag spikes, but no reason not to do such things
12:00 OldCoder *** glibc detected *** minetestserver: malloc(): memory corruption (fast): This sort of error; has it been seen sometimes?
12:26 proller joined #minetest-dev
12:35 shadowzone joined #minetest-dev
12:45 proller still no answer why https://github.com/minetest/minetest/commi​t/d274cbfce6ed39f5b7ad41261ede8c0fad7e980a was commited and who was approsers
12:45 sfan5 proller: what is the problem with that commit?
12:45 sfan5 if you don't like it just revert it in freeminer
12:46 sfan5 it's not like you are doing anything for minetest
12:46 proller but minetest have rules, but not for all
12:54 Zeno` nobody approbed it
12:54 Zeno` just like the last 3 commits were not approved
12:55 Zeno` sfan5, the problem is what I stated earlier
12:55 Zeno` Zeno`> Why are there commits happening without pull requests when there are 107 pull requests?
12:55 Zeno` <sol_invictus> I wonder too, Zeno
12:57 Zeno` there is no problem (that I can find) with those commits, but it's... it's the "method" (or lack of) that annoys me
12:59 Zeno` you can't say "this is the rule" but I am exempt
13:09 rubenwardy joined #minetest-dev
13:22 ImQ009 joined #minetest-dev
13:27 kahrl joined #minetest-dev
13:28 CraigyDavi joined #minetest-dev
13:28 kahrl Zeno`: subsystem maintainers can push commits to their subsystem without approval
13:28 kahrl hmmmm is the mapgen subsystem maintainer
13:29 Zeno` ahh ok, that makes sense
13:30 kahrl http://dev.minetest.net/Organisation <-- list of subsystems
13:36 PenguinDad joined #minetest-dev
13:44 Amaz left #minetest-dev
14:09 kahrl Zeno`: #1644 looks quite good
14:09 ShadowBot https://github.com/minetest/minetest/issues/1644 -- Refactor main by Zeno-
14:09 kahrl I have some time today to discuss the remaining TODOs/FIXMEs, if you want
14:12 Zeno` thanks. Can we catch up tomorrow? It's 12:12AM here and I was just about to sleep
14:12 Zeno` I can be around whenever you might be on next
14:13 kahrl sure, although I'm not sure yet when I'll be online
14:13 Zeno` Discussion is the best option
14:13 Zeno` ok, but I will hang around all day tomorrow (I don't mind doing that)?
14:13 Zeno` or email me! :)
14:13 Zeno` I will PM you my email address
14:14 Zeno` sorry for not wanting to do it now, I'm just very tired :(
14:14 kahrl I understand
14:15 kahrl have a good sleep :)
14:17 Zeno` thanks
14:17 Zeno` I really wish I could stay awake, but I make bad decisions at this time of night anyway heh
14:18 kahrl hehe
14:29 jin_xi joined #minetest-dev
14:58 hmmmm joined #minetest-dev
15:09 zat joined #minetest-dev
15:20 hintss joined #minetest-dev
15:33 Amaz joined #minetest-dev
15:45 shadowzone joined #minetest-dev
15:56 hmmmm for my upcoming unified map generation component interface, one of the goals is to have a single CRUD interface
15:56 hmmmm the update and delete parts are a bit more difficult
15:58 hmmmm not a problem though, thanks to the wonder atomic swaps... which got me thinking, since it's not really that much of an issue to modify registered entities after the world has started, would anybody have a use for changing nodes and item definitions?
16:15 rubenwardy joined #minetest-dev
16:16 CraigyDavi` joined #minetest-dev
16:24 sol_invictus yeah, changing nodes can be useful
16:24 sol_invictus hmmmm: something like more advanced circular saw can get use of that
16:30 CraigyDavi`` joined #minetest-dev
16:40 rubenwardy hmmmm, Wouldn't changing a node definition mean that all nodes that use it will be changed? How would that be used?
16:44 PilzAdam joined #minetest-dev
16:48 sfan5 rubenwardy: changing a nodedef would affect all nodes
16:51 Calinou joined #minetest-dev
17:05 hmmmm yup
17:05 hmmmm so I don't know
17:05 hmmmm only thing is, the server would need to resend nodedefs to the client
17:06 hmmmm this could be useful for season changing or something... imagine all default:dirt_with_grass changing to default:dirt_with_snow instantaneously
17:11 rubenwardy Add notice when only minimal is installed - https://github.com/minetest/minetest/pull/1743
17:15 zat joined #minetest-dev
17:15 nore joined #minetest-dev
17:17 zat joined #minetest-dev
17:23 ImQ009 joined #minetest-dev
17:35 rubenwardy Only set the camera update keymap when using a debug build - https://github.com/minetest/minetest/pull/1744
17:36 rubenwardy Any comments? Would you rather I removed it entirely for non-debug builds?
17:42 hmmmm the problem is that it's useless
17:42 hmmmm any cheater can just build their client for debug and then they can do the same thing
17:42 hmmmm there's no way we're going to fix visibility cheats so don't even bother IMO.. just accept it as it is
17:42 rubenwardy Not useless
17:43 rubenwardy Stops noob confusion if they accidentally hit the key
17:43 rubenwardy that's the #1 reason
17:43 hmmmm well
17:43 hmmmm a lot of superfluous debug information is available to the end user on the client
17:44 hmmmm why not add a config setting for debug mode that allows access to camera update and the more "hardcore" debugging menus
17:45 hmmmm i don't like tieing certain features from working to the build type, that is a source for problems
17:45 rubenwardy I made some code that only allowed camera update to be disabled in debug (F5) mode
17:46 ShadowNinja Um, why is it that every object that get_objects_in_radius returns is a LuaEntity with the name "singleplayer", even though I know they're all supposed to be "worldedit:..."?  Can someone else confirm?
17:48 ShadowNinja This is with a fixed /we-clearobjects that calculates the size correctly (Just sqrt(x^2+y^2+z^2) instead of sqrt(x*2+y*2+z*2) or whatever it was).
17:50 ShadowNinja http://pastebin.ubuntu.com/8722075/
17:52 PenguinDad That would explain why my attempt to fix the player deletion with /we-clearobjects didn't work…
17:52 VanessaE joined #minetest-dev
17:54 ShadowNinja PenguinDad: I fixed the player deletion bug (obj:is_player() still works) but now that the deletion box is calculated properly it includes the selection boxes and I con't exclude those from deletion.
17:57 PenguinDad ShadowNinja: I tried to fix it a few months ago
17:58 ShadowNinja obj:get_entity_name() works though...
17:58 ShadowNinja I might have broken this with my coroutine fixes...
18:05 ShadowNinja minetest.luaentities is broken too.
18:11 shadowzone joined #minetest-dev
18:12 ShadowNinja The metatable in minetest.registered_entities is fine, it's being explicitly overridded in the entity.
18:12 RealBadAngel ShadowNinja, updated https://github.com/minetest/minetest/pull/1778
18:15 ShadowNinja RealBadAngel: Looks good.
18:16 hmmmm HRRMMM.
18:17 hmmmm looking for a C++03 way of statically populating a std::map
18:18 ShadowNinja hmmmm: I don't think there is one.
18:19 ShadowNinja If it'll stay static a const array might work, but if it's just initial values you'll have to add them one by one with operator[] (or equivalent)..
18:20 hmmmm i thought so!  :(
18:21 shadowzone joined #minetest-dev
18:22 Calinou don't all current LTS operating systems support C++11 now?
18:23 hmmmm maybe it does
18:23 hmmmm i'm mostly looking at freebsd 9.x and earlier
18:23 hmmmm there are people who still use that shit, like myself
18:25 PilzAdam Calinou, ubuntu 12.04 doesn't, AFAIK
18:26 Calinou that is not the current LTS
18:26 Calinou 14.04 is
18:26 Calinou if you host game servers I hope you can upgrade every 2 years…
18:26 Calinou :\
18:26 RealBadAngel hmmmm, are you ok with my pull?
18:26 PilzAdam I though by "current" you meant "supported"
18:26 PilzAdam +t
18:30 hmmmm what pull
18:31 RealBadAngel #1778
18:31 ShadowBot https://github.com/minetest/minetest/issues/1778 -- Add option to enable mesh caching, add wallmounted for meshes. by RealBadAngel
18:31 PenguinDad What about #1779 ?
18:31 ShadowBot https://github.com/minetest/minetest/issues/1779 -- Fix indentation issues of fd5eaae by PenguinDad
18:31 Miner_48er joined #minetest-dev
18:33 ShadowNinja PenguinDad: <3 +1
18:33 PenguinDad I just couldn't leave these unfixed :P
18:34 ShadowNinja PenguinDad: I'm working on a bug at the moment, but that's simple enough that anyone can merge it.
18:38 Krock joined #minetest-dev
18:43 rubenwardy joined #minetest-dev
18:55 kaeza joined #minetest-dev
19:10 CraigyDavi` joined #minetest-dev
19:26 hmmmm RealBadAngel, in that pull request you clone and then subsequently drop a mesh inside of mapblock_mesh_generate_special
19:26 hmmmm that can't be too good for performance...
19:28 Calinou wtf
19:29 Calinou you're accepting a commit purely done for styling fixes?
19:29 Calinou you said it was a no-go for comments
19:33 ShadowNinja Calinou: Huh?
19:33 Calinou you're accepting a style commit; which may break PRs
19:34 Calinou why would you refuse comment PR then
19:37 ShadowNinja Calinou: It fixes code that has no active PRs.  Fixing code style over the entire project is still a no-go.
19:37 Calinou ah
19:38 iqualfragile joined #minetest-dev
19:39 ShadowNinja PenguinDad: Fixed that bug, pushed.
20:07 shadowzone joined #minetest-dev
20:09 ImQ009 joined #minetest-dev
20:34 shadowzone joined #minetest-dev
21:04 iqualfragile_ joined #minetest-dev
21:06 proller joined #minetest-dev
21:35 Amaz joined #minetest-dev
22:14 shadowzone joined #minetest-dev
22:23 zat joined #minetest-dev
22:30 Megaf I wish I could compile minetest on OS X as easy I can on GNU/Linux
22:32 RealBadAngel hmmmm, when no longer needed (added to mesh collector) it is dropped
22:33 RealBadAngel thats how it works. anyway this is only case of cache disabled, slower one
22:33 kilbith RBA, is it possible to have collisionbox perfectly adjusted to the models ?
22:33 RealBadAngel no. look at the name
22:33 kilbith boxes yeah
22:33 kilbith just curiosity
22:47 zat joined #minetest-dev
22:52 RealBadAngel kahrl, here?
22:54 kahrl RealBadAngel: just came back here
22:54 kahrl what's up?
22:55 RealBadAngel need a vote, #1778
22:55 ShadowBot https://github.com/minetest/minetest/issues/1778 -- Add option to enable mesh caching, add wallmounted for meshes. by RealBadAngel
22:56 kahrl https://github.com/RealBadAngel/minetest/com​mit/b3b69809ab80032167f456e88609b48d9a9d766b​#diff-c03ca828c6b8a7695f2cd7e52c316a3cR1751
22:56 kahrl that looks like a leak
22:58 kahrl also, is there a way to "not repeat yourself" in that part of the code? (it's almost the same as the branch above)
23:00 RealBadAngel its not a leak. without cache only 0 mesh is stored.
23:01 kahrl how is it not a leak?
23:01 RealBadAngel when you need it rotated you have to make a clone of 0, apply transformations, add it to collector, then wipe
23:01 kahrl in line 1751 you allocate a SMesh and in line 1752 you overwrite the only pointer to it with a different one
23:02 RealBadAngel clone does no overwrite pointers
23:02 RealBadAngel *not
23:02 kahrl but = does
23:02 Megaf Thats a nice error, isnt it? "PANIC: unprotected error in call to Lua API (not enough memory)"
23:03 Megaf caused by a dramatic increased on teh active block range by me
23:03 RealBadAngel clone takes a pointer to empty mesh, and adds to it source mesh (buffer by buffer)
23:03 kahrl huh
23:04 kahrl in line 1752 cloneMesh clearly only takes one parameter, the source mesh
23:04 kahrl and presumably it returns a pointer to the newly allocated cloned mesh
23:04 RealBadAngel hmmz
23:05 kahrl Megaf: so... don't do that then? :)
23:05 RealBadAngel lemme think a while
23:06 RealBadAngel i think youre right, will fix it
23:07 kahrl RealBadAngel: also if you change the type to scene::IMesh* you can remove the cast in line 1752, I think
23:08 RealBadAngel but about "not repeating yourself", using only one append would require another check
23:08 RealBadAngel so code will be slower
23:09 RealBadAngel after first bunch of checks i will have to recheck for existence of pointer
23:11 kahrl well if the cache is disabled the code will be slow anyway
23:12 RealBadAngel no point to make it even slower
23:13 RealBadAngel also such extra check will hit cached one too
23:13 RealBadAngel it is already slowed down by checking for the pointer
23:14 RealBadAngel which shouldnt take place at all
23:14 RealBadAngel but it is here for lazy modders
23:16 RealBadAngel some things have right to crash. modder should know what he is doing
23:16 kahrl does this one extra check really make any difference?
23:16 RealBadAngel yes it does. it is GFX
23:16 kahrl but only few nodes are mesh nodes
23:17 RealBadAngel all nodeboxes are meshes
23:17 RealBadAngel we dont have nodeboxes runtime anymore
23:17 kahrl even if there are many, the branch should always be predicted as "not taken" on modern processors
23:17 kahrl (assuming the cache is active)
23:18 kahrl (or "taken" if the compiler reorders the code)
23:18 RealBadAngel compilers are not magicians
23:18 RealBadAngel if logic require too many steps its faulty
23:19 kahrl have you done any actual measurements?
23:19 RealBadAngel i could but what for? for sake of 3 lines of code being reordered?
23:20 RealBadAngel thats minimal logic to solve the task
23:20 kahrl meh, I don't care enough to continue this discussion
23:21 kahrl fix the missing spaces after } and then push the code already
23:22 RealBadAngel ok, testing now code without creating new mesh pointer
23:23 RealBadAngel lol, i was placing some meshes around, put a sphere under my feet
23:23 RealBadAngel now im stuck inside a sphere :)
23:25 kahrl oh wait, about https://github.com/RealBadAngel/minetest/com​mit/b3b69809ab80032167f456e88609b48d9a9d766b​#diff-c03ca828c6b8a7695f2cd7e52c316a3cR1735
23:25 kahrl wouldn't it be possible to eliminate this branch (yay!) if the conversion was always done
23:25 zat joined #minetest-dev
23:27 RealBadAngel conversion in nodedef is done only when caching is enabled
23:27 RealBadAngel otherwise its converted runtime
23:28 kahrl https://github.com/RealBadAngel/minetest/com​mit/b3b69809ab80032167f456e88609b48d9a9d766b​#diff-70868aa6d6b96c0c1623c761500d23c4R884
23:28 kahrl set k = wm_to_6d[k] and access f->mesh_ptr[k] instead of f->mesh_ptr[j]
23:28 kahrl err k = wm_to_6d[j]
23:29 kahrl also it's kind of ugly to have the wm_to_6d table duplicated in nodedef.cpp and content_mapblock.cpp but meh
23:30 RealBadAngel if wallmounted was compatible with axis dir i would use just j*4
23:30 RealBadAngel but it even upside down for 0
23:31 kahrl uhhhm wait
23:31 RealBadAngel it could be just perfect if we could change wallmounted
23:31 kahrl isn't this also a memory leak when it does f->mesh_ptr[0] = cloneMesh(f->mesh_ptr[0])
23:32 kahrl I mean, if you made the change I suggested
23:32 kahrl ugh
23:33 RealBadAngel mesh_ptr are all NULL before entering here
23:33 kahrl except mesh_ptr[0]
23:33 RealBadAngel yes
23:35 kahrl well just push it as is
23:36 kahrl (with the space after } change that I mentioned, of course)
23:36 RealBadAngel sure
23:36 kahrl and assuming the memleak fix worked
23:36 RealBadAngel yes, but i do need casting there
23:37 kahrl what for?
23:37 RealBadAngel i need clone to work on IMesh
23:37 RealBadAngel it will be needed for animated ones
23:37 shadowzone joined #minetest-dev
23:37 RealBadAngel so in case of static, casting is needed
23:38 kahrl I don't get it
23:38 RealBadAngel it doesnt do automatic conversions
23:39 RealBadAngel i have to cast it
23:39 kahrl in what line do you actually use anything from SMesh as opposed to IMesh?
23:40 RealBadAngel all animated ones will require IMesh
23:40 kahrl so why not use IMesh now?
23:41 RealBadAngel im not sure but it could broke all collecting methods
23:41 kahrl how?
23:42 RealBadAngel wait a sec
23:42 RealBadAngel im already using clone without casting in nodedef.cpp
23:42 kahrl the contract of cloneMesh() states that the return value is an IMesh*. It doesn't guarantee that it's an SMesh*, so callers shouldn't rely on that
23:42 RealBadAngel why it fails in content_mapblock?
23:43 shadowzone joined #minetest-dev
23:43 RealBadAngel /home/realbadangel/minetest-meshes3​/src/content_mapblock.cpp:1751:49: error: invalid conversion from ‘irr::scene::IMesh*’ to ‘irr::scene::SMesh*’ [-fpermissive]
23:43 RealBadAngel thats without casting
23:43 kahrl even when you do scene::IMesh* mesh = cloneMesh(f.mesh_ptr[0])?
23:44 RealBadAngel no, SMesh
23:44 kahrl you have to change the type of the variable to IMesh* of course
23:44 RealBadAngel note that mesh_ptr are of SMesh*
23:44 RealBadAngel and in nodedef it compiles and works
23:45 kahrl nope, they're IMesh*
23:45 RealBadAngel *facepalm*
23:46 RealBadAngel ok, no more questions :)
23:55 RealBadAngel } else if (f.mesh_ptr[0])
23:55 RealBadAngel i think i dont need to check for the pointer here again
23:56 RealBadAngel first if checks for all the possible pointers including 0, both with cache and without
23:56 RealBadAngel so the "without cache" code is executed only for facedir != 0
23:57 kahrl where are you?
23:57 RealBadAngel and pointers surely NULL
23:57 kahrl nodedef.cpp:881?
23:57 RealBadAngel https://github.com/minetest/minetest/pull/1778/f​iles#diff-c03ca828c6b8a7695f2cd7e52c316a3cR1749
23:58 kahrl hmm
23:59 kahrl what if all the mesh_ptrs are NULL?
23:59 kahrl could that theoretically happen?
23:59 RealBadAngel well, yes

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