Time Nick Message 01:17 Warr1024 http://i.imgur.com/lKHlzQB.png 01:20 VanessaE loking good, Warr1024 01:20 Warr1024 btw, the inverting isn't based on the LSB, it's the 2nd LSB 01:20 VanessaE +o 01:20 Warr1024 so if I expand the image by 2px, it inverts 01:20 VanessaE strange 01:20 Warr1024 my guess is it has something to do with the way inventory cubes are created 01:20 Warr1024 they're rendered to a virtual scene then resized to 64x64 01:21 VanessaE what scaling algo are you using now? 01:21 Warr1024 resizing the image by 2px would be what would change the screen half-size between even/odd. 01:21 VanessaE (I ask because that dry shrub looks a bit blurry) 01:21 Warr1024 my algo is basically a linear filter. 01:21 Warr1024 blurry is sort of a consequence of resizing a 16x16 image to 49x49 01:21 VanessaE *nod* 01:22 Warr1024 the lanczos you were playing with compensates by sharpening, but it's a bit beyond what I'm working with here. 01:22 Warr1024 also, dunno how lanczos handles alpha channels, which is where the blur is really noticeable 01:22 VanessaE that one is the result of your in-world filter? 01:23 Warr1024 you mean the texture filter? 01:23 Warr1024 that's not in this branch 01:23 VanessaE oh ok 01:23 VanessaE and yeah 01:23 Warr1024 there's not a hell of a lot I can really do; any non-integer scale risks blur 01:23 VanessaE I meant the one you had in that other pull for the in-world filtering-with-threshold 01:24 Warr1024 oh, yeah, that one 01:24 VanessaE no worries then 01:24 Warr1024 the one where I fake the RGB values of transparent pixels 01:24 Warr1024 I had to pull in an adaptation of that here. 01:24 VanessaE in any case it's a vast improvement over what was the case before 01:24 Warr1024 I wonder if maybe I should merge this with my other pullreq, since they share some common logic. 01:24 Warr1024 it's an improvement if I can squash that upside-down texture bug. 01:27 Warr1024 yeah, the bug must definitely be in the inventoryimage generator 01:28 Warr1024 even if I do nothing but convert the texture to an image and then back, it's still flipped. 01:28 Warr1024 there must be some internal bit on the texture on the image to control its flippedness that's not propagated to the image. 01:42 Warr1024 holy crap, all that code I was manipulating made no diff because it was all in an #ifdef __ANDROID__ block I didn't notice 01:42 Warr1024 which I guess might explain why android nodes are upside-down... 01:47 Zeno` heh 01:50 Warr1024 Zeno`: http://i.imgur.com/lKHlzQB.png 01:51 Warr1024 made them look better, but now I've got a weird issue where they flip upside down if the target rectangle is an odd number of px height/width. 01:53 Zeno` err, that does sound very weird 01:54 Warr1024 https://github.com/Warr1024/minetest/commits/guiscaling 01:54 Warr1024 it's definitely not a problem with the drawing 01:54 Warr1024 er, teh scaling filter 01:54 Warr1024 I think the problem is we render to texture and then I convert texture back to image for the scaling filter 01:55 Warr1024 and the texture->image->texture round trip has some isse 01:55 Warr1024 like the ITexture I'm looking at has some hidden isFlipped bit that's not accessible on ITexture. 01:55 Warr1024 or maybe it's a bug in the driver trying to pull it from video to main memory 01:56 Warr1024 the filter is basically an box filter, i.e. anti-aliased nearest-neighbor. 01:57 Warr1024 should make non-exact-integer-scaled images look like non-crap. 01:57 Warr1024 works nicely, aside from seriously messy code and that one damn caveat. 01:58 Warr1024 ah, good, I thought I found a 2nd bug, but apparently it's an upstream one. 01:58 VanessaE Warr1024: something must be wrong. 01:58 VanessaE Warr1024: because I ...think I understand this code. O_O 01:59 Warr1024 resize your screen to very narrow like a portrait phone, and the invbar and hud get kinda messed up 01:59 Warr1024 VanessaE: :-) 02:00 Warr1024 I thought that iterating over images using getPixel in nested loops would be horribly slow, until I read a bit of the Irrlicht source code and saw that that's exactly what they do in many places. 02:00 Warr1024 I figured it'd have some kind of fancy 2D accel for things like that, but I guess it just leans on the compiler to provide decent optimizations. 02:02 Warr1024 it's almost tempting to write a very lightweight software renderer just to parse the node meshes and render those inventory cubes 02:04 Zeno` sy <= maxs32sy ? 02:04 Warr1024 Zeno`: I don't know a shitload about optimizing compilers 02:04 Warr1024 but I didn't really want to do a cast inside a loop 02:04 * kahrl is suddenly reminded of https://github.com/kahrl/minetest/commits/optimize_invtex 02:04 Warr1024 where the result value would always be the same. 02:05 VanessaE Warr1024: there's already a software renderer in irrlicht....... 02:05 Warr1024 yeah, I played with it 02:05 Warr1024 there are 2, actually 02:05 VanessaE yeah 02:05 Warr1024 one that works poorly and one that's pretty much useless 02:05 VanessaE probably not good enough to do what you have in mind 02:05 Warr1024 I tried playing in fullscreen in BurningsVideo, and it feels like I'm playing Minecraft again, only still with better drawdist. 02:06 Warr1024 well, my idea was a fake renderer that'd actually do all the work in 2d 02:06 Warr1024 since it's an isometric projection, it's just a simple transformation matrix. 02:06 Zeno` kahrl, interesting :) 02:06 Zeno` Maybe you should look at kahrl's link, Warr1024 02:06 Warr1024 kahrl: why did work stop on this, and/or why not make a pullreq? 02:07 Warr1024 for one thing, I know android will probably benefit 02:07 Warr1024 render-to-texture is very finnicky and not very portable. 02:07 Warr1024 all the android images right now are upside-down, except for glass, which is just trash for me. 02:07 kahrl it was buggy and I didn't figure out a way to fix it 02:07 Warr1024 maybe I'll have to pull it 02:08 kahrl can't remember what exactly the problems were 02:08 Warr1024 see if I can beat it into good enough shape to at least compete with the turd running on android right now 02:08 kahrl in any case, it was orders of magnitude faster than render-to-texture, on my machine at least 02:08 Warr1024 nice 02:09 Warr1024 ok, here's a weird thing 02:09 Warr1024 images are either always upside-down or right-side-up for a given target size 02:09 Warr1024 depending on odd or even 02:09 Warr1024 but apparently on subsequent runs, whether odd or even are the upsidedown ones changes...? 02:10 Warr1024 the first time I open my inventory, it's always upside-down 02:10 Warr1024 then if I resize the window, it inverts depending on the LSB of the size. 02:11 Warr1024 so if I maximize the window, open my inventory, then restore, the images are fine whenever I'm not maximized 02:11 Warr1024 but if I open my inventory the maximize, they're fine when I AM maximized. 02:12 Warr1024 kahrl: I kind of wouldn't care if your renderer were actually faster, since there's the potential to make it less buggy, but for RTT, I don't think there's much we can do other than layer on workarounds. 02:13 Warr1024 just about the only thing I think I can do right now for this pull request will be either (1) some awful heuristic hack to detect upside-down shit and flip them, or (2) just exclude inventory elements from filtering for now. 02:13 * Zeno` compiles it 02:14 Zeno` I did something similar to this a *long* time ago but I used a modified breshenhams algo (well, I used the error accumulator idea from that) 02:15 Warr1024 whoa, and glass is NOT upside down for me 02:15 Zeno` back when FPUs were not common heh 02:15 kahrl Zeno`: yeah, that could potentially be even faster 02:16 kahrl I didn't bother with that because I wanted to fix the bugs first :P 02:16 Warr1024 holy shit, glass is una-f***ing-ffected. 02:16 Zeno` kahrl, of course 02:16 Zeno` get it to work, then optimise... the standard rule 02:17 Warr1024 Zeno`: my filter is pretty much a completely unoptimized naive box filter; it works well for me mainly because I cache the generated images permanently, and the source images are generally not huge to begin with. 02:17 Warr1024 though I wouldn't mind seeing how some of my stuff behaves on against a modpack with insane texture sizes. 02:17 Zeno` Warr1024, if they're cached then optimisation may not be worth the hassle (it probably wouldn't be) 02:17 Warr1024 you can still almost always process images faster than they're loaded from disk. 02:18 Zeno` grr I have to look at the src code again 02:18 Zeno` save me the time... anything special I need to do to enable this? 02:18 Warr1024 yeah, config 02:18 Warr1024 gui_scale_filter = true 02:19 Zeno` k 02:20 Zeno` lol at my config 02:20 Zeno` I just noticed this: 02:20 Zeno` new_style_leaves = true 02:20 Zeno` pooh = 21 02:20 Zeno` I wonder what I was doing... :3 02:22 Warr1024 combine this patch (assuming we can work out the bugs) with my other one for the 3d textures, and we can get this game looking damn smooth. 02:22 Warr1024 I remember playing minecraft and wondering why it didn't let you enable texture filtering. 02:23 Warr1024 but I still got like 15fps with a draw dist of 32 in it 02:23 Warr1024 probably because they recalculated all the geometry every frame 02:23 Warr1024 now I'm starting to get used to a fully-filtered MT at 15fps and a 256 draw dist on my old-ass laptop 02:24 Warr1024 15 is my wanted_fps; I don't get involved in a lot of action, and I like to be ready to take HQ screenshots :-) 02:28 Warr1024 btw, sorry about my abuse of errorstream; I couldn't figure out where the messages went when I was using infostream... 02:34 Warr1024 http://i.imgur.com/doE76WJ.png 02:35 Warr1024 world and GUI filters both. 02:36 Warr1024 It's still quite playable on an IGP from about 5 years ago; seems to run about as well as with all filters off 02:36 VanessaE the 3d/in-world seems a bit blurry compared to the screenshots you provided the other day? 02:38 VanessaE (particularly that near cactus and the wielded cobble stairs) 02:38 Warr1024 just noticed that glass actually has an [inventorycube{ image, while the others are using RTT, so that explains THAT. 02:38 Warr1024 VanessaE: I'm a lot closer to that cactus that I was before 02:39 VanessaE oh ok 02:39 Warr1024 I run my textures at 64x, and it doesn't bother me; if it bothers you, you could go up to 128x or higher. 02:39 VanessaE Warr1024: no worries. just taking note of it is all. 02:39 VanessaE makes it look like a focus-limit filter :) 02:40 Warr1024 :-) 02:42 Zeno` it's a large aperture filter 02:42 Zeno` heh 02:43 Zeno` not all images are upside down 02:44 Warr1024 yeah 02:44 Warr1024 which ones are you seeing? 02:45 Warr1024 I think it's only true RTT's that are upside down 02:45 Zeno` http://i.imgur.com/31AuLqI.jpg 02:45 Warr1024 some nodes use an [inventorcube{ image, which are done in 2D, and are thus not affected 02:45 Zeno` so jungle leaves are ok... normal leaves, not ok 02:45 Warr1024 weird that 02:46 Warr1024 them hearts are antialiased too 02:47 Zeno` grrr, cavegen is using unitialised values again 02:47 Warr1024 I've mostly been playing MP on a 0.4.12 server, so testing out in SP today was the first I ran into the "HUD/inventory not updating when I eat an apple" bug. 02:47 Warr1024 it's somewhere near the top of the issues on github... 02:48 Zeno` memory leak in Server::ProcessData() ... interesting 02:48 Warr1024 you gotta admit, it may be wrong, but at least it's pretty :-) 02:48 Warr1024 it's also quite different from the android issues 02:49 Warr1024 for one, on android, the mesh generation is all messed up; here, it uses the correct mesh, and rasterizes it fine, just ends up flipping the finished image 02:49 Warr1024 then I guess what that means is maybe it's [inventorycube{ that's actually messed up on android...? 02:50 Warr1024 since that's what glass uses...? 02:50 Warr1024 and my glass-materialed nodeboxes are fine 02:50 VanessaE glass doesn't need to use that feature anymore, kahrl fixed the normals in that drawtype a while ago 02:51 VanessaE btw, your minetest_game is out of date. 02:52 VanessaE WAY out of date :) 02:52 VanessaE zeno ^^ 02:53 Warr1024 ha, I just assumed it was a different gametype that "borrowed" some of the old textures. 02:53 Warr1024 mt has the very best fjords 02:54 Zeno` VanessaE, i kind of know that :) 02:54 Zeno` it's the commit before minetest next was merged 02:55 VanessaE oh right, forgot 02:55 Warr1024 I hope mgv7 still has nice fjords... 02:55 Warr1024 minetest next? 02:55 Zeno` hdx-128 looks nice with this scaling/smoothing business 02:55 VanessaE Warr1024: https://github.com/BlockMen/minetest_next 02:56 Warr1024 had controversial artwork? 02:56 VanessaE BlockMen was maintaining minetest_game in a so-named separate branch for a while, eventually that merged into minetest_game mainline. a few problems sprung up at the time but I think they've all been ironed out 02:57 VanessaE s/branch/fork/ 02:57 Warr1024 VanessaE: they're practically the same thing, the way git is designed... 02:57 Warr1024 Zeno`: if you're using 128x textures to begin with, they probably won't be upscaled unless you set your texture_min_size above 128 02:58 Warr1024 Zeno`: BUT if you're on a world with a mix of texture sizes the texture_min_size thing will scale them up 02:58 Warr1024 GUI scaling doesn't need any fine-tuning 02:58 Warr1024 nor does the transparency fix 03:00 Warr1024 hm, one drawback to the transparency thing is that it does weird things to non-fancy trees 03:00 Warr1024 because non-fancy trees are a weird alpha hack themselves 03:00 Warr1024 they DO blend nicely though in the distance 03:02 Warr1024 hm, it was actually better when I had the filters applied in the other order 03:02 Warr1024 though doing that had its own cost in terms of load time 03:07 Zeno` they're upscaled in the world though aren't they? 03:08 Warr1024 well, first I apply the transparency filter 03:08 Warr1024 then I upscale them 03:08 Warr1024 so I save the cost of having to transparency filter the extra pixels, since, for normal transparent textures, the end-result is the same 03:09 Warr1024 except for trees, where transparent gets baked into black, I guess it would have been better to do it the other way around? 03:09 Warr1024 but that's such a weird corner-case... 03:09 Warr1024 frankly, the way the fanciness of leaves works is a bit bizarre to me 03:10 Warr1024 the fact that that one optimization is so highly positioned as a setting to warrant a dedicated checkbox for it is dependent on a lot of assumptions about the game/mods/mapgen the user will be using.. 03:10 Zeno` yuck, I am getting nasty spikes in drawtime 03:11 Warr1024 related to my patches? 03:11 Warr1024 I also gotta see how the inventory/gui filter looks on droid. 03:11 Zeno` not related 03:12 Zeno` it's nasty 03:12 Zeno` my drawtime spikes from 4 to ~60 03:12 Zeno` fortunately I can repeat it 03:13 Zeno` wtf would be causing that?! 03:13 Warr1024 I've actually seen a lot less spikege than I used to 03:13 Warr1024 IIRC mesh recalcs can be bad 03:13 Zeno` I've never seen this before 03:14 VanessaE Zeno`: got anything causing periodic mesh regens? like a mesecons blinky plant? 03:14 Zeno` VanessaE, no 03:14 Zeno` this is an area of my test world I use in testing a lot 03:14 Zeno` I've never seen it before 03:14 Zeno` it's a complete freeze then the scene redraws 03:14 VanessaE minimal game, no mods, I assume? 03:15 Zeno` bizarre 03:15 Zeno` yeah 03:15 Zeno` it's a jungle tree area where drawtime has always been longer but not with spikes/freezes like this 03:16 VanessaE that sounds like something "spooling up" and then dumping out after a while 03:17 VanessaE meshes being sent to the GPU getting hung up somewhere in a cache maybe? 03:19 Zeno` I would have noticed it before now; it's one of my controlled test areas I used for optimising the previous 2 releases 03:19 Warr1024 that one black pixel on the wall-torch animation is driving me nuts 03:20 * VanessaE hides from Warr1024 03:20 Zeno` *sigh* I'm going to have to bisect I suppose 03:21 VanessaE that black pixel was supposed to be coal tar or something similar 03:21 VanessaE the torches are actually downscaled and tweaked from the ones I made for HDX 03:21 Zeno` lol I can force packets to be lost 03:22 Zeno` walk into the jungle leave, lose 4 packets 03:22 VanessaE how? 03:22 Warr1024 a lot must have been lost in the downscale 03:22 Zeno` rinse and repeat 03:22 Warr1024 seems like the coal tar should be always visible or never... 03:22 VanessaE Warr1024: indeed so: https://github.com/VanessaE/hdx-512/blob/master/default_torch.png 03:23 Warr1024 actually, one network bug that's been bothering me is that when my master-branch client quits my 0.4.12 server, sometimes the server never seems to get the quit packet and thinks I'm still logged in 03:23 Warr1024 I'm on a wired network, so I REALLY shouldn't be seeing dropped packets. 03:23 Warr1024 VanessaE: wow, holy shit that's a downscale if I ever saw one 03:24 VanessaE yep. looks better than what we had before though :) 03:24 Zeno` I'm not even on a network... well unless you call 127.0.0.1 a network, which it is, but it's neither wired nor wireless 03:25 Zeno` what troubles me is it always drops packets when I walk into these leaves 03:25 VanessaE Zeno`: well technically it would be a wired network, albeit an extremely fast one, wouldn't it? ;) 03:25 Zeno` and always 4 03:26 Zeno` VanessaE, technically I suppose you're right 03:26 VanessaE you're right, notwithstanding 03:26 * Zeno` tries to work out how to trace this 03:26 VanessaE but, always 4 packets just from walking into some leaves?? 03:26 Zeno` a big bunch of leaves yes 03:26 VanessaE you sure it isn't because you cross some arbitrary mapblock boundary or something? 03:27 Zeno` I walk into them, lose 4 packets, back out, wait and walk into them again (lose 4 packets again) 03:27 Zeno` no I'm not sure 03:27 Zeno` checking 03:27 VanessaE what happens if you manually force some of those leaves to snap into view by manipulating your min view range, without walking? 03:27 VanessaE some of the leaves out PAST the ones you walked into I mean 03:27 Zeno` they are in view. I have to be inside them 03:28 Zeno` s/be/go inside 03:28 VanessaE inside? as in non-walkable? 03:28 Zeno` as in no clip into them 03:28 VanessaE interesting 03:28 VanessaE (same difference) 03:29 VanessaE I just wondered if there was stuff coming into view out past those leaves, that might be triggering the packet loss somehow 03:36 ShadowNinja Zeno`: Nocliping disables occlusion culling, could that be it? 03:37 Zeno` ShadowNinja, maybe 03:37 ShadowNinja Zeno`: Do the spikes repeat regularly? 03:37 Zeno` http://dpaste.com/0C723ZG 03:37 Zeno` nah once I'm "in there" they stop 03:38 Zeno` I'll continue looking at my logs 03:39 Zeno` not that it helps much 03:39 Zeno` I think I'll run it under callgrind 03:42 WarrTab Huh. So adding a cpp file to cmake doesn't seem to add it to the android build... 03:42 Zeno` WarrTab, correct 03:43 Zeno` add it to minetest/build/android/jni/Android.mk 03:45 WarrTab That's kind of awkward... 03:45 Zeno` I can think of a better word, but that's the way it is right now :) 03:45 WarrTab I managed tonfigure it out with grep anyway; was jist remarking... 03:46 WarrTab Sounds like we need a meta-meta-build process :-) 03:46 Zeno` weta-meta! 03:47 WarrTab Technically metameta == pata 03:47 Zeno` well, I'll leave you to write the pata-build process (haha) 03:48 WarrTab Makes me think it's time to uprgade to a sata-build process 03:49 WarrTab Most of the stuff I build is basically gcc -o $(OUT) `find src -type f -name '*.c'` 03:55 Zeno` gcc *.c 03:55 Zeno` :D 03:55 Zeno` works 03:55 Zeno` the build when you're umm not really building 03:55 Zeno` I have lots of those 03:55 Zeno` haha 03:55 WarrTab Cheapo one-step-compile-and-link 03:56 Zeno` well not everything needs a makefile 03:56 WarrTab np, everything does, but not every makefile needs to be more than 2 lines long... 03:57 WarrTab Time to try this apk... 03:59 WarrTab Total fail 04:00 WarrTab Apparently I can't retrieve images back from the gpu on android 04:00 WarrTab I figured portability could be a problem... 04:00 Zeno` well that was an epic waste of time 04:01 WarrTab Oh? 04:01 Zeno` I forgot to callgrind_control -i on lol 04:01 WarrTab Ha 04:01 Zeno` I have a 15 minute call profile of... nothing 04:01 WarrTab Well, I have a possible workaround, if I can intercept where images become textures 04:01 WarrTab RTT's will still be tricky 04:03 WarrTab We need some way to keep android from quitting if the activity is disposed 04:04 WarrTab Sucks to tab out to respond to irc, load a link, and have mt get killed 04:07 WarrTab Touchscreen controls are also counterintuitive 04:08 WarrTab bot axes should benreversed, thougn not exactly either... 04:09 WarrTab It's more like the thing you first touch should stay locked under your finger, and your view should rotate as necessary to keep that true 04:10 WarrTab Tap-to-look would be nice too. 06:16 neoascetic why do we have all of this CUSTOM_bla-bla-bla variables? Why not simply redefine original variables with -DLOCALDIR, for example? 08:32 nrzkt hi Zeno` 08:32 Zeno` howdy 08:33 Zeno` Only here for a moment (but will be back properly later on) 08:33 nrzkt interesting conversation this night. Are you trying to improve rendering of the interface ? 08:33 Zeno` interface? 08:33 nrzkt gui 08:33 Zeno` oh, not me. I was going to but WarrTab began before I got a chance 08:34 Zeno` Since he's started I'll let him continue :) 08:34 nrzkt that's good, more hands = precise tasks for other devs :p 08:34 Zeno` Yeah. I've been testing his patch 08:34 nrzkt good ? 08:34 Zeno` it looks pretty good... well, apart from the upside down images heh 08:34 nrzkt i see, what is the real improvement in fact ? 08:35 Zeno` just better scaling 08:35 Zeno` of images 08:36 nrzkt okay :) 08:42 Zeno` I'm going to add memset(this->heightmap, 0, csize.X * csize.Z * sizeof(*this->heightmap)); to the constructor of MapgenV6 because it's using unititialised values again 08:43 Zeno` will merge in a second 08:53 nrzkt okay 08:57 Zeno` for reference, https://gist.github.com/Zeno-/37566659c0e442605771 08:58 nrzkt good :) 08:59 Zeno` I sent paramat a memo as well 08:59 Zeno` ok, will be back later :) 09:43 neoascetic Hi, could someone help me with testing of #2471 on windows? 09:43 ShadowBot https://github.com/minetest/minetest/issues/2471 -- Implement dependencies copying to target directory on OSX by neoascetic 13:06 Zeno` who would have thought tooltips would be such a controversy :/ 13:06 Zeno` bugger adding an option to turn them off. Either add them or not 13:06 Zeno` Personally I think add them 13:10 Zeno` But if the decision is made to not add them it probably should be a very good reason :) 13:11 est where? 13:11 Zeno` #2438 13:12 ShadowBot https://github.com/minetest/minetest/issues/2438 -- Add tooltips for many elements of the main menu by Wuzzy2 13:13 Zeno` I'm not sure how tooltips in the main menu can distract anybody 13:13 Zeno` heck, I hardly even look at the main menu 13:14 nrzkt right, or maybe wait 1 sec before showing it ? :) 13:14 est yup 13:14 est that should be done 13:14 est like normal tooltip 13:14 est tooltips* 13:14 est 1 sec or .5 sec or whatever 13:15 est instantaneous is not good 13:15 est I mean no program does instantaneous toltips do they? 13:15 Zeno` it's not instant 13:16 Zeno` maybe the tooltips detract from the submersion into the main menu 13:16 nrzkt main menu submersion... it's a strange thing :) 13:16 Zeno` but to be honest I don't want to be submersed into the main menu and I've never met anyone who would 13:17 est tooltips are no manual 13:17 est manuals are manual 13:17 est manuals* 13:17 Zeno` yes, but tooltips are not strange nor unexpected 13:17 Zeno` not it that situation anyway 13:18 est yes thats right, but not more than one sentence 13:18 est 80 chars not more 13:18 Zeno` yep. It needs improvements 13:18 est perhaps even less 13:18 est (just said random number :D) 13:48 neoascetic once again :) Could someone help me with testing of #2471 on windows? 13:48 ShadowBot https://github.com/minetest/minetest/issues/2471 -- Implement dependencies copying to target directory on OSX by neoascetic 13:53 Zeno` sfan5 ^ 13:55 neoascetic This is sadly that travis cannot into pure-windows testing 13:57 est31 sadly that windows still gets used 13:57 neoascetic well, that's also true :) 15:15 sfan5 neoascetic: how would fixup_bundle know that it need to copy the DLL files? 15:16 est31 whats all this narrow_to_wide stuff about? 15:17 est31 and back again 15:17 neoascetic sfan5 it uses otool (something else on windows, some black cmake's magic) 15:17 est31 why does it need to be typed the right way 15:17 est31 its data 15:19 sfan5 neoascetic: it links to the .dll.a or .a files, so it can't know where the DLLs are 15:19 sfan5 -> this can't work 15:19 sfan5 also bundles are an osx only thing 15:20 sfan5 i would be surprised if the bundle code handled windows and/or linux 15:21 neoascetic I sure it not only osx-specific, it's just strangely named 15:21 neoascetic Let me search 15:22 neoascetic here one example: http://www.cmake.org/pipermail/cmake/2012-January/048487.html 15:22 neoascetic have you tried it? 15:22 sfan5 apparently it handles other OSs aswell 15:24 sfan5 I'll try this later 15:24 neoascetic Ok, thanks, I hope it will work 15:24 sfan5 neoascetic: does the bundle code print something when doing it's work? 15:25 neoascetic yes, on OSX is something like 'fixing_bla bla bla 15:25 sfan5 it didn't print something 15:25 sfan5 https://travis-ci.org/minetest/minetest/jobs/53765089#L9239 15:25 neoascetic I guess this is because it is not true windows env 15:26 sfan5 it also still *can't* even work because it doesn't know where the DLLs are 15:26 sfan5 you mean it will only work when you compile on windows? 15:26 sfan5 I'm against that in that case 15:27 sfan5 everything should work as expected when cross-compiling 15:27 neoascetic I don't know if CMake's WIN32 flag defined when building from linux 15:27 sfan5 and if you are going to add the code back for cross-compiling you could just not do this bundle thing for windows 15:28 sfan5 it is 15:28 sfan5 otherwise it wouldn't install the DLLs and get many other things wrong 15:30 neoascetic I. e. we never build minetest on windows but always use cross-compiling? 15:31 exio4 it may be built on windows, or it may be crosscompiled 15:31 exio4 both things should work 15:32 neoascetic Ok, I didn't knew that. I'll investigate then. I just didn't want to setup Windows env for build. Now I'll setup debian box for cross-compiling tests 15:34 est31 Zeno`, sfan5: #2472 15:34 ShadowBot https://github.com/minetest/minetest/issues/2472 -- Unescape tooltip texts by est31 15:54 neoascetic sfan5 I guess safest way for now is to implement that part for OSX only 15:56 neoascetic I hope one day all these CMakeList files will be refactored and cleaned 15:57 est31 neoascetic, there is currently need for a build system maintainer 15:58 neoascetic Yeah 16:45 nrzkt est31: you are right 16:46 est31 lol with what 16:46 nrzkt we can rename some functions on this pr 16:47 est31 ok 16:47 est31 ah that 16:47 est31 thx 16:53 * ShadowNinja points neoascetic to #2402 16:53 ShadowBot https://github.com/minetest/minetest/issues/2402 -- Clean up and tweak build system by ShadowNinja 16:54 VanessaE ~tell nrzkt seems the problems I've been bitching about predate your major network changes. meh. 16:54 ShadowBot VanessaE: O.K. 16:54 ShadowNinja And, while we're at it, PilzAdam: You O.K. with that ^ ? 16:55 est31 ShadowNinja, before merge, can I test whether it works on android? 16:55 neoascetic ShadowNinja oh, great, I'll take a look 16:56 est31 could have had idea earlier I admit 16:57 PilzAdam ShadowNinja, looks good, if it works 16:59 neoascetic ShadowNinja why do we have all of these CUSTOM_bla-bla-bla variables? Couldn't we simply redefine original variables instead, could we? 17:02 neoascetic https://github.com/minetest/minetest/pull/2402/files#diff-af3b638bc2a3e6c650974192a53c7291R5 17:02 neoascetic this message is also unnecessary because we do not build deb packages? 17:03 est31 wasn't there an option to build deb packages 17:03 sfan5 no 17:05 est31 but some people may want to run CPack? 17:05 est31 its far more better than make install 17:05 est31 dunno if we support that 17:06 est31 lots of dunnos 17:06 est31 waiting for android build of SN's patch 17:06 sfan5 ofc we support cpack 17:06 sfan5 make package is handled by cpack 17:06 neoascetic But it creates only zips 17:06 neoascetic as far as I know 17:07 sfan5 it could theoretically create other things 17:07 sfan5 packages 17:07 neoascetic but anyway our CMAkeList should be modified 17:07 neoascetic to do this 17:08 ShadowNinja neoascetic: It might be possible to do that. I'll check. 17:08 ShadowNinja est31: Sure, try it now. 17:09 est31 ShadowNinja, doing it some reason my downstream is 3 kbps 17:09 est31 oh 5 17:12 est31 ShadowNinja, btw you could clean up your branches :) 17:14 ShadowNinja neoascetic: It's doable, but I'd have a bunch or DEFAULT_* vars instead. 17:15 neoascetic I also do not like this stuff: https://github.com/minetest/minetest/pull/2402/files#diff-af3b638bc2a3e6c650974192a53c7291R39 17:15 neoascetic can it be replaced with something shorter? :) 17:16 neoascetic ShadowNinja but why all these DEFAULT_* vars? Why simply not let to redefine variables? 17:19 ShadowNinja neoascetic: option()/set() needs a default. 17:20 ShadowNinja And changing it later will be even uglier. 17:21 ShadowNinja As for the RUN_IN_PLACE thing, I can use #cmakedefine01 RUN_IN_PLACE instead. 17:30 neoascetic https://github.com/minetest/minetest/pull/2471 implemented this stuff for OSX only. I guess this may be merged now? 17:36 est31 ShadowNinja, sfan5 building on android works, I've also added a github comment 17:37 est31 for* 18:46 Calinou so, nrzkt, you could make a forum post to announce your Jenkins once it is ready. 18:46 Calinou so that users can download daily builds 18:47 nrzkt then, we continue to talk about daily builds here. ATM i can provide windows-x86 builds. but i need to talk with celeron55 because he maintain the page, and we need to talk together to interact :) 18:47 Calinou rubenwardy & others can edit minetest.net if needed 18:47 rubenwardy Yes. What needs to be done? 18:48 nrzkt nothing atm :p 18:48 neoascetic What purpose of daily builds? I think fast releases are more better for end users, no? 18:49 nrzkt no 18:49 est31 no agree 18:49 nrzkt quality process 18:49 nrzkt we are working on great work and it need a longer release cycle, because we need to break some huge things and have feedback from community tests before release :) 18:49 est31 because with daily builds, you can say "its a daily build, dont expect to it be working" 18:50 Calinou neoascetic, people can test them for bugs 18:50 Calinou without having to compile 18:50 est31 and fast releases require testing each time 18:50 est31 or not 18:50 neoascetic But if cross-compiling works as expected, we may use travis to make daily builds (and even every-push-to-master) build with no problem 18:50 est31 bullshit what I said 18:51 est31 travis already is building 18:51 est31 its building every PR 18:51 est31 we only need to record it 18:51 neoascetic Yes, but result of build is not published 18:51 est31 yes 18:51 est31 it can be turned on 18:51 nrzkt est 18:51 est31 you need to specify it as "release" 18:52 est31 then turn on release publishing with travis 18:52 est31 yes nrzkt? 18:52 nrzkt est31: right :) travais already build we don't need to override him atm i think. But for daily builds it's not the correct tool :) (sorry i type enter too fast :p) 18:52 neoascetic Why not? 18:52 nrzkt because it's not our instance ? :p 18:52 neoascetic I don't see the problem 18:53 neoascetic Not our, and?.. 18:53 nrzkt if it's not our server, how to get the builds ? :p 18:53 nrzkt and how to schedule releases because it's only a PR build system ? 18:53 est31 there is a feature that even does download publishing 18:53 nrzkt oh, i didn't know about taht 18:53 neoascetic http://blog.biicode.com/automatically-build-publish-via-travis-ci-github/ 18:53 neoascetic for example 18:54 est31 yup 18:54 neoascetic I want do same thing for OSX - daily builds 18:56 nrzkt i will look at this 18:56 nrzkt it only permit to build push and pull requests 18:56 est31 no 18:56 neoascetic It may be easy than you think to setup daily builds :) 18:56 nrzkt i'm on https://travis-ci.org/minetest/minetest/settings 18:56 neoascetic you may trigger the build via API 18:56 est31 you can set it up to test every commit 18:57 nrzkt yes push and pull :) 18:57 neoascetic Nope 18:57 nrzkt there are Build only if .travis.yml is present / Build pushes / Build pull requests / Concurrent jobs 18:57 neoascetic for example, here a little utility I've wrote https://github.com/neoascetic/travis-cron 18:57 est31 push does mean every commit 18:57 neoascetic you may trigger latest build on specific branch (for example, master) 18:58 nrzkt i know est31 18:58 nrzkt i mean trigger a build every day for example 18:58 nrzkt not every commit or every pull :) we don't have this option on the interface 18:58 est31 ah yes thats not possible 18:58 est31 also publishing requires to mark every commit as release 18:58 nrzkt it's there where jenkins can come. 18:59 neoascetic Ohh... give me a time and I'll show how to do this with travis :D 18:59 neoascetic I need this be merged, however: https://github.com/minetest/minetest/pull/2471 19:00 neoascetic Btw, how is long the review process? 19:00 est31 2 core devs need to say yes 19:00 est31 officially 19:00 est31 if its small thing then I think one is enough 19:00 est31 I'm not sure 19:00 neoascetic core devs, cmon, say yes please, this won't break anything :D 19:00 nrzkt right est31 19:01 nrzkt i look at your commit, please wait 19:01 est31 nrzkt, also consider reviewing #2472 19:01 ShadowBot https://github.com/minetest/minetest/issues/2472 -- Unescape tooltip texts by est31 19:02 nrzkt i don't understand the improvement here, because i'm not very familiar with formspec 19:03 rubenwardy nrzkt, it builds current master as well. 19:04 nrzkt rubenwardy: on which PR are you talking about ? 19:05 rubenwardy It does here: https://github.com/rubenwardy/NodeBoxEditor 19:05 rubenwardy So Minetest would too 19:06 rubenwardy https://travis-ci.org/minetest/minetest 19:06 rubenwardy See 19:06 nrzkt building != working :) 19:06 est31 but building == you can compile and bisect 19:07 est31 which is most important thing 19:07 neoascetic Yeah, cpp is not Haskell ;) 19:07 nrzkt but it seems good, as far i can remember the modificated options are good 19:07 rubenwardy I may have misunderstood 19:07 nrzkt you have my +1 if a mac os fanboy (burn them) core-dev agrees 19:08 exio4 neoascetic: you said the magic word 19:08 est31 nrzkt is a devil he is burning master 19:09 exio4 there are no os x core devs, as far as I know? 19:10 neoascetic yeah, that's why my PR sometimes stuck :) 19:12 nrzkt i need to test it on a real BSD in fact 19:15 nrzkt oh lol 19:15 nrzkt we can use travis & jenkins together 19:15 neoascetic but why? 19:15 nrzkt build on FreeBSD :) 19:16 est31 because as it seems travis doesnt have enough ram for android 19:16 nrzkt erf jenkins became crazy 19:16 neoascetic We may build on OSX with travis - that's better IMO 19:16 nrzkt i launch the PR process and he will try to build every opened PR 19:17 nrzkt lohttp://jenkins.unix-experience.fr/ 19:17 nrzkt http://jenkins.unix-experience.fr/ 19:17 neoascetic Oh android 19:17 nrzkt and also android yes, i will add the 3 jobs 19:17 nrzkt 2* 19:18 neoascetic We may use travis for android also... 19:18 nrzkt yes but not enough ram then killed 19:18 nrzkt my server has 8-12GB free, it's sufficient for this build 19:19 nrzkt when the crazy jenkins will have finished i will look at android compilation integration, i also need a NDK for other android projects 19:20 est31 nrzkt, you can start with the one I have created for travis 19:20 est31 #2436 19:20 ShadowBot https://github.com/minetest/minetest/issues/2436 -- Add Travis checks for Android 19:21 est31 its a bit complicated as version numbers are important 19:21 neoascetic I hope you guys will not turn travis off because I have a dream of builds and for OSX too 19:21 est31 I would be for having both enabled too 19:21 nrzkt travis can build for OSX ? 19:21 neoascetic yes 19:22 nrzkt i think i also can, because i run FreeBSD and OSX is a FreeBSD :p 19:22 neoascetic https://travis-ci.org/neoascetic/minetest/branches I track master branch for every push and build it on OSX with travis 19:23 neoascetic oh well FreeBSD is unix and linux is unix and if it builds on unix it should also work on OSX 19:23 neoascetic but that's not true ;) 19:23 nrzkt linux is not unix 19:23 nrzkt :) 19:23 neoascetic unix-ispired, ok 19:23 neoascetic inspired 19:24 est31 linux isn't even unix-inspired 19:24 est31 linux is a kernel, unix is an operating system 19:24 est31 thats two things 19:25 Warr1024 unix isn't really an operating system;m more of a specification for one. 19:25 Warr1024 Unix is an OS like ARM is a processor. 19:25 neoascetic let's holywar begin 19:26 Warr1024 nah, let's not. 19:28 neoascetic as per http://docs.travis-ci.com/user/ci-environment/#The-Build-Machine, On the Linux platform, builds have 3 GB of memory available. Disk space is limited, but there's no fixed limit per build. Builds have up to two cores available (bursted). 19:28 neoascetic 3 GB is not enough for android build? 19:30 est31 as it seems 19:30 est31 perhaps you can point to something else that makes build failing 19:30 neoascetic something wrong with this world 19:32 neoascetic swapon: /swap: swapon failed: Operation not permitted 19:33 neoascetic that's why I guess? 19:33 est31 no 19:33 est31 that was test error message 19:33 est31 I tried to add swap to test whether its ram issue 19:33 est31 travis doesnt allow swap 19:33 est31 thats why 19:33 neoascetic Ok, where is previous? 19:36 est31 neoascetic, https://travis-ci.org/minetest/minetest/jobs/53444462#L7045 19:36 est31 I also linked to it in history 19:53 nrzkt hmmm sorry for the jenkins PR build on github, i let him compile all issues and disable the task after. clang is not very happy on freebsd with our build :p 22:42 nrzkt Ok core-devs, my jenkins has built all PR, sorry for the inconvenience. I will let him build PR for freebsd builds (using Clang) natively, which permit to point other bugs. The builder is called jenkins-ci/freebsd-nrz 23:02 VanessaE so what's the story with the ghost entities? 23:02 VanessaE 00:02:35: ERROR[ServerThread]: ServerEnv: Trying to store id=30875 statically but block (-43,1,-28) already contains 554 objects. Forcing delete. 23:02 VanessaE etc. 23:03 VanessaE players being deleted while the server tries to delete these non-existent entities. 23:05 VanessaE it just keeps repeating the same message over and over for a given mapblock, but with the count gradually increasing (that block is at 558 now)