IRC logs for #openttd on OFTC at 2022-09-03
⏴ go to previous day
00:14:41 *** WormnestAndroid has joined #openttd
00:14:54 *** WormnestAndroid has quit IRC (Read error: Connection reset by peer)
00:15:17 *** WormnestAndroid has joined #openttd
00:23:21 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
00:23:42 *** WormnestAndroid has joined #openttd
00:30:35 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
01:19:35 *** Wormnest has joined #openttd
02:11:38 *** ely-the-kitsune has joined #openttd
02:12:16 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
02:17:45 *** debdog has quit IRC (Ping timeout: 480 seconds)
02:31:10 *** Wormnest has joined #openttd
05:54:26 *** WormnestAndroid has quit IRC (Remote host closed the connection)
07:53:57 <andythenorth> eh could include a pylon
08:27:50 *** HerzogDeXtEr has joined #openttd
08:27:54 *** murr4y has quit IRC (Quit: WeeChat 3.6)
08:33:42 <TrueBrain> yup ... UPDATE ALL DEPENDENCIES .. again .. 😛
08:44:22 <TinoDidriksen> That's part of why I generally disagree with committing exact-version lock files. Just let me specify a minimum version that it should be compatible with, and have CI tests confirm that it remains compatible with newer versions - and report to me when it breaks.
08:45:36 <TinoDidriksen> Sure, because it will breaks eventually.
08:46:26 <JGR> There's no intrinsic reason why things need to break all the time
08:46:43 <JGR> It is possible to aim higher
08:47:48 <TrueBrain> TinoDidriksen: you use the exact version in lock files for a final product (mind you, not for libraries :P), as you want to know what is running on your production
08:47:58 <TrueBrain> one of the worst problems to debug is: if I start it now, it works, but it failed yesterday
08:48:05 <TrueBrain> because some library changed, and you didn't know
08:48:16 <TrueBrain> additionally, in this day and age, you really should check the diff of your dependencies
08:48:34 <TrueBrain> as sadly, too many packages are being hijacked 😦
08:49:39 *** Soccerdog has joined #openttd
08:49:39 <Soccerdog> damn bro thats crazy
08:51:18 <TrueBrain> but otherwise, yes, that is exactly what our projects do: we once defined a minimal version, and every new release the CI validates if it is compatible. If so, we accept the update 🙂
09:19:03 *** D-HUND is now known as debdog
10:01:17 <andythenorth> we are so so far behind the referential chain of integrity for components that other industries have 😛
10:01:40 <andythenorth> even car industry has it 😛
10:02:04 <andythenorth> this is how Ford know to recall my van because it has a fuse in it that doesn't work
10:02:19 <andythenorth> or possibly they just guess, wouldn't be surprised 😛
10:54:03 *** WormnestAndroid has joined #openttd
11:26:00 *** OrangeLeaf36 has joined #openttd
11:26:00 <OrangeLeaf36> where would the other end be chatting from
11:28:00 <nielsm> the IRC channel #openttd on OFTC network
11:34:46 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
11:38:14 <peter1138> Hmm, lazy afternoon time, or shall I go out for a wander.
11:43:08 *** WormnestAndroid has joined #openttd
11:52:08 <andythenorth> think it does need a tile with some electricity crap
11:52:12 <andythenorth> and animated sparks
12:03:27 <dP> hm, is there a good way to do `max(map(lambda x: ..., vector))` in C++ without a loop ?
12:10:05 <nielsm> std::accumulate(vector.begin(), vector.end(), 0, std::max<list.value_type>) maybe
12:10:56 <nielsm> actually you need a function wrapping your operation and the max for the one at the end
12:11:52 <nielsm> there's also std::reduce which I'm not entirely sure how differs from std::accumulate
12:11:56 *** ely-the-kitsune has quit IRC (Ping timeout: 480 seconds)
12:12:09 <JGR> std::transform_reduce would probably do it
12:12:30 <JGR> Though arguably at that point a loop is probably better
12:13:05 <nielsm> the theoretical advantage is that you can give an execution policy (automatic parallelism) to the std lib algorithms
12:14:16 <dP> considering I need it for like 30 entries max I doubt it's worth it
12:14:43 <nielsm> and yeah the trouble with min/max operations over collections is that you need to bootstrap them, and might need a zero/minimum/"nothing" value for that
12:16:22 <michi_cc[d]> std::max_element can also take a lambda for the comparison, that could do the map, too.
12:18:05 <dP> converting map to comparator is not very practical
12:19:25 <michi_cc[d]> Very much depends on the lambda I guess. No problem if it just accesses some member of a class/struct, but not very nice if it is a complex calculation.
12:19:58 <dP> you also need to map max_element result
13:25:06 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
13:34:03 *** WormnestAndroid has joined #openttd
14:51:31 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
14:59:55 *** Alkel_U3 has quit IRC (Quit: maintenance)
15:00:23 *** WormnestAndroid has joined #openttd
15:03:16 *** Alkel_U3 has joined #openttd
15:28:03 *** Wormnest has joined #openttd
15:37:49 <frosch> #48 is so weird. so many questions, yet they did not try to run it once
15:41:06 *** WormnestAndroid has quit IRC (Ping timeout: 480 seconds)
15:50:27 *** Flygon_ has joined #openttd
15:56:47 *** Flygon__ has joined #openttd
15:57:55 *** Flygon has quit IRC (Ping timeout: 480 seconds)
15:58:30 *** Flygon_ has quit IRC (Ping timeout: 480 seconds)
16:14:36 <FLHerne> can has option to make all FIRS industries neutral colours? :p
16:14:54 <FLHerne> I do find the pastel ones a bit strange-looking sometimes
16:15:18 <FLHerne> individually they're fine, but when there are half-a-dozen in view in different colours it's weird
16:41:58 <andythenorth> there's already a callback that nerfs some colours
16:42:06 <andythenorth> not sure I'm going to add a parameter for more though 😛
16:42:16 <andythenorth> it's probably one switch and a param though
16:43:18 <glx[d]> it seems you broke github TallTyler 🙂
16:43:55 <glx[d]> ah it's fixed, but I got the unicorn page
16:52:41 <glx[d]> oh I see why it's using this weird construct, it's just to prevent the red colour for negative values
17:00:16 *** ely-the-kitsune has joined #openttd
17:23:04 <andythenorth> 5 layouts will be enough 😛
17:23:18 <andythenorth> shame I couldn't figure out the layout choose dP made 😛
17:23:28 <andythenorth> random is random 😦
17:29:58 <andythenorth> heavy industry with minimal smoke seems wrong though 😛
17:32:09 <peter1138> Oh, cheers x-plane, that's not a great place for a tree.
17:33:58 <andythenorth> middle of a runway?
17:35:08 <andythenorth> smoke at an electric arc furnace!
17:35:16 <andythenorth> might be steam 😛
17:41:46 <peter1138> But yes. Two trees right on a helipad. And the helipad is sloping.
17:55:49 <andythenorth> oof adding the smoke causes clipping on the tile in front
17:56:03 <andythenorth> the molten metal is clipped on the left
17:56:13 <andythenorth> I never did understand bounding boxes 😛
17:59:15 <frosch> don't look at bounding boxes, they do not affect clipping
17:59:29 <andythenorth> normally when I have this problem, I delete one of the animations
17:59:38 <andythenorth> then the problem vanishes 😛
17:59:48 <frosch> the important part is that your sprites do not extend over left/right corner of the tile
18:00:02 <frosch> if they do, you need to split them, and draw them on the other tile
18:00:27 <andythenorth> so maybe I can push the smoke left on x
18:00:41 <frosch> ideally you would draw the whole building on the spritesheet, and cut it up in python/nml
18:01:10 <andythenorth> doesn't TB have one of those?
18:01:20 <frosch> it's on the todo list for truegrf 🙂
18:01:42 <frosch> but now that Truegrf_stations changed nick to WoozyDragon4018 , it's looking grim for trugrf
18:23:17 <TrueBrain> they are just showing their true colours
18:25:36 *** virtualrandomnumber has joined #openttd
18:25:41 *** virtualrandomnumber has quit IRC ()
18:26:03 <michi_cc[d]> Interesting, it is "nml stations" on Discord, but the bridge made it into WoozyDragon4018 🙂
18:26:41 <TrueBrain> the bridge doesn't use the friendly name 🙂
18:26:49 <TrueBrain> as I was pretty sure people would abuse the hell out of that 😄
18:27:56 <frosch> there has been a draft docs since 2012
18:29:46 <andythenorth> how good was it 😛
18:30:50 <TrueBrain> took 10 years to implement; you do the math 😛
18:31:19 <frosch> pretty sure noone involved in the current design/implementaiton looked at that page 🙂
18:31:46 <frosch> we had an issue on old devzone, were we designed another spec some years ago
18:31:56 <frosch> even then i did not know about this even earlier effort
18:32:18 <TrueBrain> how many devs does it take to change a light bulp? 😛
18:33:00 <frosch> 1 doing the work, 1 looking approving, the rest talking in the bike shed
18:33:51 <Eddi|zuHause> can i join in on the bike shed?
18:34:11 <TrueBrain> but yeah, talking about station support .. guess the only way that happens is if I quit my job, again 😛 Stupid free time .. or rather, stupid work-weeks 😛
18:34:38 <TrueBrain> I wish 24h work-weeks becmae the norm ... 😄
18:38:17 <frosch> how much do i have to pay to whom to replace "horst" with "truebrain" or "andy", and add it to the discord stickers?
18:39:01 <TrueBrain> owh please no, aaaaaahhhhh
18:40:43 <frosch> anyway, i'll review that nml-station page now, and delete everything that does not apply
18:41:42 <TrueBrain> so you are going to remove that page?
18:42:53 * andythenorth removes extra smoke due to flickering
18:43:12 <frosch> no, i'll keep the cross-linking between the features
18:43:46 <andythenorth> I am the Marketing Manager
18:47:05 <dP> is there any GS method that takes two strings? need to test something
18:47:50 <andythenorth> NML was made of optimism 🙂
18:48:00 <andythenorth> how old were most of the authors? about 20?
18:50:38 <dP> oh, `ScriptText::GetEncodedText` returns a pointer to the static buffer
19:11:26 <glx[d]> frosch: I know this page, it lists all the variables except random_bits I included with the implementation
19:11:43 <frosch> yes, i added 3 missing vars so far
19:11:58 <frosch> the rest is surprisingly complete
19:20:02 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
19:31:01 <frosch> found a bug in ottd's implementation of newgrf stations 🙂
19:49:05 <frosch> is there any case where you would want a track-type offset?
19:49:27 <frosch> maybe just add the "+0" in all cases
19:52:30 <glx[d]> you want the offset if it's a track tile
19:57:21 <frosch> hmm, oh, ottd does only do the railtype magic itself, if newgrf railtypes with overlays are used
19:57:28 <frosch> i assume it did that always
20:00:27 *** Wormnest has joined #openttd
20:00:31 <frosch> ok, next attempt: you want the "+0" in all cases, except GROUNDSPRITE_RAIL_Y/X
20:08:02 <andythenorth> I've got a 5 frame animation, and I want to offset one set of sprites in it
20:08:10 <andythenorth> so I have frames 0, 1, 2, 3, 4
20:08:24 <andythenorth> and I want to show sprites 2, 3, 4, 1, 0
20:08:37 <andythenorth> I thought I could do it with %, but I can't figure it out 😛
20:09:15 <andythenorth> var is animation_frame
20:09:30 <andythenorth> stacking a bit ternary wouldn't be nice 😛
20:09:37 <Eddi|zuHause> so effectively you want to flip 0 and 1 for some loops, but not all of them?
20:10:06 <andythenorth> specifically there are 5 smoke sprites, and I want to show them twice
20:10:14 <andythenorth> but the second one is offset
20:10:38 <Eddi|zuHause> so what is the complete sequence (repeated infinitely)?
20:11:13 <andythenorth> what's an appropriate notation 😛
20:11:40 <frosch> use a procedure which returns 2,3,4,1,0 for input 0,1,2,3,4?
20:11:48 <Eddi|zuHause> what's wrong with digits?
20:11:53 <andythenorth> [A0, B2], [A1, B3], [A2, B4], [A3, B0], [A4, B1]
20:12:54 <Eddi|zuHause> so each [] ist two sprites somehow encoded as one callback result?
20:13:30 <andythenorth> animation_frame < 3: animation_frame + 2 ? animation_frame - 4
20:13:43 <dP> ```>>> [(x + 2) % 5 ^ x // 3 for x in range(5)]
20:13:55 <michi_cc[d]> 1, 0 or 0, 1? You wrote both right now.
20:13:56 <frosch> glx[d]: disabled_platforms and disabled_length are stupid to document. i wanted to write "bitmask of disallowed lengths", but bits are 0-based and lengths are 1-based. so to disallow length 3, you have to use bitmask(2) :/
20:14:00 <Eddi|zuHause> assuming the "1,0" you showed earlier is a mistake, what you want is "(frame+2)%5"
20:14:20 <andythenorth> yeah it's a mistake
20:14:54 <andythenorth> dP: can't use python in nml 😛
20:14:59 <andythenorth> I am aware of other compilers
20:15:13 <dP> nml should have all the operators :p
20:16:19 <frosch> so andy gave a riddle, which was only interesting, because he made a typo? :p
20:17:49 <frosch> sounds like the math exercises from my studies, which were designed to give nice numbers, except when the author messed up some + - when typing them
20:18:33 <glx[d]> frosch: could be "bitmask of disallow lengths" with 1-7 and a internal left shift inside nmlc
20:20:52 <glx[d]> hmm does bit 7 disable drag/drop in openttd ?
20:21:10 <frosch> bit 7 is for anything longer than 7
20:23:41 <Eddi|zuHause> andythenorth: the ternary version would work, except the second case should be -3, not -4
20:24:36 <andythenorth> and the ? and : are transposed 😛
20:24:51 <frosch> glx[d]: doesn't look like it
20:24:58 <Eddi|zuHause> that's just syntax, that's uninteresting :p
20:26:27 <dP> except it's much better to use % instead of ternary in grf
20:26:55 <frosch> glx[d]: i tested, when you use drag&drop you have to drag an allowed size
20:27:03 <frosch> you have to count the tiles yourself for that 🙂
20:27:40 <frosch> giving the most generic "can't build railway station here" :p
20:28:11 <glx[d]> typical extra explicit error yes
20:28:40 <andythenorth> this fancy smoke also breaks the other animation
20:29:36 <glx[d]> so I can change nmlc to expect bitmask(1-8) and a hidden shift
20:34:03 <glx[d]> but I need to be careful to not apply the shift if it's not a bitmask
20:35:53 <frosch> why would anyone not use a bitmask?
20:36:06 <frosch> "0" is the only meaningful value otherwise
20:36:14 <frosch> and you can shift that as much as you want :p
20:46:27 *** Wolf01 has quit IRC (Quit: Once again the world is quick to bury me.)
20:53:12 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
20:55:17 <andythenorth> not sure that fancy smoke looks good anyway
20:59:17 *** Flygon_ has joined #openttd
21:04:19 <glx[d]> not my best commit message
21:06:55 *** Flygon__ has quit IRC (Ping timeout: 480 seconds)
21:30:51 *** Wormnest has joined #openttd
21:41:52 *** ely-the-kitsune has quit IRC (Remote host closed the connection)
21:43:25 *** ely-the-kitsune has joined #openttd
21:46:53 *** nielsm has quit IRC (Ping timeout: 480 seconds)
21:58:24 <frosch> oh wait, no, chips2 ofc
21:59:09 <andythenorth> depends if I want to learn stations 😛
21:59:18 <andythenorth> also the stations could just go into FIRS
21:59:22 <andythenorth> bit weird but eh
21:59:31 <frosch> longer compile times :p
21:59:39 <andythenorth> FIRS is quite quick...
22:00:02 <andythenorth> also can we allow more than one GS, for the case where they are only allowed to plant objects? 😛
22:00:29 <frosch> there is some text in tb's gist about that
22:01:00 <frosch> maybe even in the wiki
22:01:36 <andythenorth> 'industry fields' in grf is unlikely
22:01:42 <andythenorth> and GS can probably do it
22:27:02 <dP> andythenorth: you can do it with grf-py and let me write half of the code ;)
22:28:05 <andythenorth> I was looking at grf-py again earlier today
22:28:14 <andythenorth> how did the Horse port go?
22:31:01 <dP> well, I kinda achieved what I could with it in terms of developing grf-py
22:31:18 <dP> there is no point on finishing it unless you want to switch, it's just more of the same stuff
22:32:11 <andythenorth> I don't want to switch currently
22:32:19 <andythenorth> but in the long term NML is ....
22:32:38 <andythenorth> ... there are reasons to switch 😛
22:32:52 <andythenorth> hmm, does OpenGFX redefine palettes?
22:33:29 <andythenorth> that pink is transparent in normal testing
22:34:51 <andythenorth> nobody reported it after the release 😛
22:43:17 <dP> well, let me know if you ever want to do anything with grf-py
22:43:32 <dP> it was kind of designed for big projects like yours
22:44:06 <dP> but is currently in a bit of a chicken-egg situation
22:45:46 <dP> it's kinda pointless to develop it without real applications and it's not very attractive with features and docs missing xD
22:47:05 <dP> but, yeah, if you even want to try it we can do smth like you write code however you want and I try to make it work ;)
22:49:42 *** Flygon_ has quit IRC (Remote host closed the connection)
22:50:03 <andythenorth> I wonder what all the FIRS forks will do
22:50:21 <andythenorth> they nearly all take the raw firs.nml and edit it
22:50:34 <andythenorth> not even the separate industry files it generates
22:50:49 <dP> they'll finally have to learn the proper way I guess :p
22:51:03 <andythenorth> well FIRS is in TrueGRF...
22:51:10 <andythenorth> so that might be the end of forks
22:51:28 <dP> or, actually... grf-py can decompile...
22:52:12 <andythenorth> I forget, did you have any timings for compiling with grf-py ?
22:53:10 <dP> I did some but there isn't much to compare with as I didn't do the full horse port
22:53:25 <dP> also didn't do any particular optimizations yet so it's pretty comparable to nml
22:53:33 <andythenorth> TBH, some of the vehicle templating, I never want to touch again
22:53:43 <andythenorth> I'm not sure it can be ported without introducing bugs 😛
22:53:56 <andythenorth> the loops are so nested, and so conditional
22:54:05 <andythenorth> and depend on so much python generation also
22:54:36 <dP> nah, horse temlates are terrible, they should be half-rewritten anyway
22:54:45 <dP> at the very least to get rid of most switches
22:55:44 <andythenorth> didn't you try that?
22:56:00 <dP> yeah, I got rid of some in what I ported
22:56:08 <andythenorth> in theory every realsprite could just be reached as a pre-computed result
22:56:15 <andythenorth> and just a small number of switches
22:56:26 <andythenorth> I just can't work it out
22:56:43 <andythenorth> it would need a lot of bit shifting and masking or something
22:57:35 <dP> yeah, some bit stuffing and large bits->sprite mapping
22:57:41 <dP> ideally automated in python
22:58:01 <andythenorth> 33% of it is literally to deal with vehicles that aren't 8/8
22:58:36 <andythenorth> it's mad that OpenTTD can't fix that
22:58:56 <andythenorth> and that other authors don't even seem to notice how broken it is
22:59:58 <JGR> What exactly is broken about it?
23:00:48 <andythenorth> if you don't transpose views 0-3 and 4-7, vehicles less than 8/8 are considerably out of place when flipped
23:02:34 <JGR> Ctrl-click flipping is rarely used feature so probably others haven't noticed
23:02:56 <dP> you mean vehicles flipped with ctrl-click don't use the right sprites? sounds like a game but, not something you should fix in newgrf
23:03:11 <andythenorth> the amount of drama when it was banned by default, I thought ctrl-click was high usage 😛
23:03:17 <andythenorth> squeaky wheel problems
23:03:40 <andythenorth> dP: fixing in newgrf is currently the conventional solution 😛
23:03:56 <JGR> Nowhere in the UI is it mentioned, so how would users know the feature even exists?
23:03:57 <andythenorth> so all action 2 chains for vehicles < 8/8 are duplicated
23:04:44 <andythenorth> JGR: 'mystery meat' 😛
23:05:54 <andythenorth> ctrl-click is still better than the livery callback 😛
23:06:03 <andythenorth> at least it's not actually broken
23:06:15 <andythenorth> apart from being broken
23:07:19 <dP> too bad debug vehicles can't be reversed
23:09:19 <JGR> Did a quick test with a random GRF and I see what you mean, never knew about that issue
23:09:29 <JGR> Easy solution is "don't do that then"
23:12:24 <andythenorth> it's the major cause of Iron Horse realsprites being nearly 50% duplicates
23:12:46 <andythenorth> they are the same bounding box etc, but transposed 0-3 with 4-7
23:12:55 <JGR> Not setting bit 3 of prop 27 should just make the problem go away?
23:13:13 <andythenorth> it's used in Iron Horse to control liveries etc
23:13:21 <andythenorth> and flip asymmetric vehicles
23:16:03 <dP> andythenorth: that should be possible to deduplicate at compile time
23:16:23 <dP> also iirc that's not the main cause :p
23:16:44 <andythenorth> not sure how we'd de-duplicate the realsprites 🙂
23:17:22 <dP> grf can just write sprite once and refer to it several times in container version 2
23:17:49 <andythenorth> there's no format I'm aware of for re-ordering the views
23:17:55 <andythenorth> they come in blocks of 8
23:18:10 <andythenorth> unless it's hidden docs somewhere
23:18:19 <dP> yeah, I don't think that can be done in nml
23:18:24 <dP> compiler itself has to do that
23:19:01 <andythenorth> are there docs in the nfo spec for this?
23:19:13 <andythenorth> or is implied in OpenTTD codebase only?
23:19:29 <dP> there may be docs on grf format somewhere
23:19:36 <dP> idk, I just read the code xD
23:20:00 <andythenorth> if it's possible, then nmlc should be taught to do it
23:20:18 <dP> but basically in container 2 real sprites are in separate block and pseudo sprites just reference them by id
23:21:52 <dP> haven't actually tried that but I see no reason why same sprite can't be referenced twice
23:23:02 *** Wormnest has quit IRC (Ping timeout: 480 seconds)
23:23:28 <dP> so I plan to teach grf-py to do that eventually
23:23:37 <dP> as a part of a general sprite caching
23:24:40 <andythenorth> 91k Horse realsprites are duplicated twice
23:25:01 <andythenorth> another 10k are duplicated 4 times
23:25:30 <JGR> Deduplication seems like it should be doable in a post-processing step on the output GRF
23:25:40 <andythenorth> 8k are not duplicated
23:26:02 <andythenorth> so of 230k realsprites
23:26:03 <dP> yeah, it's doable as post-processing but I'm also thinking of compile times
23:26:15 <dP> ideally it shouldn't even resolve that sprite
23:26:46 <dP> like, if it's a generator just check the hash and not even run it if it's already cached
23:26:55 <andythenorth> so 43% or so are duplicated
23:26:58 *** Wormnest has joined #openttd
23:27:28 <andythenorth> I don't think it affects compile times much, but I haven't test recently
23:27:39 <andythenorth> I think nml caches the sprites
23:28:24 <andythenorth> Horse is still ridiculously slow to compile though, over 20 seconds
23:28:37 <andythenorth> any improvement is valid 😛
23:29:46 <dP> well, nml doesn't even have sprite generators :p
23:30:54 <JGR> I wish all the projects I worked on compiled in 20 seconds or so...
23:34:59 <glx[d]> most of these 20s is parsing IIRC
23:35:10 <glx[d]> because it's a huge generated NML file
23:37:43 <andythenorth> I think there are ways I could reduce this still with procedures
23:39:03 <andythenorth> but it would be more complicated
23:39:30 <andythenorth> for example there are 80 bulk wagons, each has a full action 2 chain
23:39:35 <andythenorth> but they all use the same rules
23:39:49 <andythenorth> so it could be one chain in a procedure
23:40:14 <andythenorth> returns a numeric result to one switch which chains to the realsprites
23:41:02 <andythenorth> I would guess that about 600 full action 2 chains can be reduced to about 25
23:42:33 <andythenorth> meh, looking, maybe I did a lot of that already
23:42:45 <andythenorth> seems I ran into 255 limit for varact 2 results 😛
23:44:36 <glx[d]> it's possible to split in more varact2 if more than 255 results
23:47:13 <andythenorth> in this case it's "optimising" away one switch, but then replacing it with another to shard the results to handle > 255
23:48:03 <dP> that is if you optimize one switch, not 10 :p
continue to next day ⏵
