00:01:25 03kate-02 07* 0.28-a0-8-g6e044cfa68: Update hints mode 10(2 minutes ago, 8 files, 132+ 220-) 13https://github.com/crawl/crawl/commit/6e044cfa6886 00:03:08 Unstable branch on crawl.kelbi.org updated to: 0.28-a0-7-g84f66890ac (34) 00:04:45 <09gammafunk> man, kind of a shame how hints mode modifications require so many c++ side changes 00:04:55 <09gammafunk> would be great if all of it ran through lua hooks or something 00:12:40 <05kate> yeah, i thought more of it was in the hints.txt file but i guess lots of it has conditional modifiers that would be tricky to implement there 00:24:03 Unstable branch on crawl.kelbi.org updated to: 0.28-a0-8-g6e044cfa68 (34) 01:34:30 Unstable branch on crawl.develz.org updated to: 0.28-a0-8-g6e044cfa68 (34) 01:55:04 Windows builds of master branch on crawl.develz.org updated to: 0.28-a0-8-g6e044cfa68 01:57:23 <10PleasingFungus> @kate nice fixes! i’m confused by the ring of flight giving a ‘resistance’ change? 02:28:37 Unstable branch on cbro.berotato.org updated to: 0.28-a0-8-g6e044cfa68 (34) 02:53:27 Monster database of master branch on crawl.develz.org updated to: 0.28-a0-8-g6e044cfa68 02:57:19 <05kate> the hint about how flight works wasn't showing up because flight doesn't grant an active ability any more, and the hint was specifically checking if you'd gained a new ability by putting on an item 02:58:28 <05kate> and the resistance hint just displays based on a fixed list of items, so that seemed like the easiest way to make the flight hint show up again 02:59:36 <05kate> maybe it should technically be its own separate hint on the code side but this was easier 03:31:28 Fork (bcrawl) on crawl.kelbi.org updated to: 0.23-a0-4217-g7c68dc2372 04:11:21 <08wormsofcan> hmm, looking at the hints change there should be a new entry for ash 08:04:56 -!- mhcerri2 is now known as mhcerri 10:05:57 <10PleasingFungus> maybe just rename the gives resistance function to gives static ability, or something? 10:06:08 <10PleasingFungus> gives passive effect 12:41:27 !lm * cwz x=vlong 12:41:30 12249277. [2021-07-23 16:39:37] [vlong=0.27-a0-1555-g0e7b3fb1d2] KOKOMAN the Cutter (L1 KoMo) began the quest for the Orb on turn 0. (D:1) 12:41:40 is that normal for cwz to be so far behind? 12:42:07 (I noticed because it's before the swamp exit fixup, there's a ton of crashes) 12:54:55 <10PleasingFungus> yes, cwz is updated manually and irregularly 12:54:59 <10PleasingFungus> so it's usually quite far behind 13:13:10 Unstable branch on crawl.akrasiac.org updated to: 0.28-a0-8-g6e044cf (34) 13:18:27 advil: yeah, hong applies some local patches, I think, to help with korean language aspects 13:18:38 probably has to merge them every time, maybe resolve conflicts, I'm not really sure 14:58:39 hi hi 14:58:52 I guess this is the new channel 14:59:12 advil: thanks for looking at that autofight pathing bug, but I think it's not quite fixed yet 15:01:27 I managed to setup a few scenarios that still break pathing with wizmode, here are some screenshots: 15:01:29 https://imgur.com/a/SvlAXE5 15:02:48 notably, one of those setups doesn't rely on unexplored space, and uses a perma-teleport trap instead 15:05:18 I suspect the problem (one of the problems?) is that those LOS stencils in losparam.cc don't actually check pathing connectivity (or at least, the stencil used by view.cell_see_cell doesn't? I'm not even going to begin to pretend I understand all of those potential interactions) 15:05:37 so, using view.cell_see_cell to verify connectivity won't actually work 15:07:02 if it would help, I can upload saves with some setups that still break pathing to that mantis issue 15:11:00 apparently some of those images didn't upload, here's a couple more: 15:11:01 https://imgur.com/pawJqTa 15:11:24 https://imgur.com/w2WSiYt 15:11:48 https://imgur.com/BfU6SAW 15:20:04 <08nicolae> fr: a cyclops unique who worships Ru, just so their description can say that Ru opened their mystical second eye 15:37:28 <10PleasingFungus> lmao 15:49:45 <06advil> Gotta say, those are different autofight bugs 15:50:08 <06advil> I guess it’s ok to use that mantis number as a catchall 15:55:52 yeah I didn't want to open a bunch of issues and they're all technically related, I think? 15:56:32 it's interactions between the losparam stencils, that awful autofight pathing code, and the autofight conditions 15:57:05 I suspect there's probably a buggy position I haven't found yet related to orc priests/orc wizards 15:57:53 I guess the reason you're not just ripping out the pathing code is cpu concerns on cao if you used astar or something? 16:20:23 (A Ru worshipping unique would play almost like fighting mummy death curses but it happens on any damage instead of on kill, for a whole lot of it in a very short span of time but maybe not as bad [cause it's ow-curses instead of death-curses].) 16:29:45 <06advil> I don’t think there are cpu concerns, but I don’t think something sophisticated is called for; it’s also not trivial to write and test something brand new 16:30:35 hm 16:30:37 <09hellmonk> Make polyphemus the ru cyclops 16:30:44 <06advil> I’m not actually sure whether the autofight path code is bad actually, the goal isn’t to find any path 16:31:01 <06advil> It’s to find a visible, intuitive path 16:31:23 well, I mean it's bad in the sense that it wastes cycles checking the same position multiple times 16:31:45 also it's unintuitive from a maintenance perspective to parse that unrolled loop 16:32:25 (otherwise someone would have worked out these pathing bugs years ago, I've been seeing complaints about this for many many versions) 17:11:18 also, it's implicit that finding a visible, intuitive path means finding a visible, intuitive (connected, traversable) path, and the current greedy code doesn't actually do that 17:11:46 admittedly, it's possible that your current approach might be able to work that in 17:31:26 <06advil> I don't mean to say that bugs shouldn't be fixed, but I think it's remarkably effective for what it does, these bugs are very rare I think 17:35:03 <06advil> I mean, if you're putting in the work to find these cases, consider writing a patch yourself? 🙂 17:36:18 <06advil> the case in the original report (https://crawl.develz.org/mantis/file_download.php?file_id=7932&type=bug) was particularly egregious in that the bad path was guaranteed to move out of los of the target 17:39:06 yeah, they are pretty rare I think 17:39:21 I don't mean to get too negative about it, thanks again for having taken a look at it 17:41:03 I might have some residual frustration from doing clua bot stuff and running into autofight bugs much more frequently than real players 17:41:38 on that note, if it would help (?) I think I already had a dijkstra's implementation written in clua that could maybe be adapted to work with autofight 17:59:59 <06advil> one of the tricky things about this code is that it really should ideally have some tests, because changing it for the worse would be immediately noticed 18:01:06 <06advil> I wonder if what this should really do is test all ray paths to the target from the los calculator 18:02:01 <06advil> btw, here's a better example case for why greedy search can't work, I think several of your cases reduce to this 18:02:02 <06advil> https://cdn.discordapp.com/attachments/747522859361894521/868251634155790346/Screen_Shot_2021-07-23_at_6.01.24_PM.png 18:02:36 <06advil> (X there is a test statue, it's much easier to use when setting up examples) 18:03:44 <06advil> well, no the ray thing wouldn't work 18:15:22 so yeah several of those cases reduce to the same thing, just with different transparent kfeats 18:16:23 i wasn't actually looking at the code to set those up, I just saw that you used view.cell_see_cell in the commit and thought "that's not going to work" cause of a tavern conversation i had a while back with snow 18:17:03 might be relevant, might not: https://crawl.develz.org/tavern/viewtopic.php?f=5&t=27049&p=345582#p345582 18:18:06 here's the traversability check i used for my clua dijkstra's implementation: https://github.com/Implojin/ff-rc/blob/master/ff.rc#L192 18:18:11 clua traversability check* 18:18:41 that's probably not even close to being ready to include in mainline cause i was just writing it for myself as a vehicle to learn some lua 18:19:26 re: unit tests 18:19:34 i totally agree? but i don't know how to write unit tests 18:20:11 my level of cs education is, as hellmonk might have once said, "maybe community college dropout level" 18:20:35 if you have any recommended reading about how to write unit tests, i'd be interested in checking it out? 18:21:37 <06advil> for this one I'm not sure a unit test per se is what I'd write, something more like a lua stress test that randomly generates a bunch of layouts with a visible test statue in them and tries to do autofight 18:21:47 <06advil> it wouldn't necessarily be easy 18:22:56 <09hellmonk> I would still say it (about me) 18:23:22 Unstable branch on underhound.eu updated to: 0.28-a0-8-g6e044cfa68 (34) 18:23:36 btw grats on convincing the devil's team of hellcrawl's obvious superiority 18:25:05 <09hellmonk> Thanks 19:11:30 <09gammafunk> it was all a trick 19:11:41 <09gammafunk> now we have access to the hellcrawl repo, we can add upstairs 19:19:16 lol I didn't know that dev team membership was reciprocal 22:14:44 <08nicolae> it works like LOS 23:46:32 03hellmonk02 07[dexterswizlab] * 0.27-a0-1445-g2eaebb3d46: Yara wizlab 10(4 minutes ago, 6 files, 135+ 1-) 13https://github.com/crawl/crawl/commit/2eaebb3d46dd 23:46:32 03hellmonk02 07[dexterswizlab] * 0.27-a0-1446-gf34c3123e1: checkwhite 10(31 seconds ago, 2 files, 5+ 5-) 13https://github.com/crawl/crawl/commit/f34c3123e1d3