05:08:21 Unstable branch on crawl.akrasiac.org updated to: 0.33-a0-7-gfa02dad2 (34) 09:38:32 <08nicolae> given the number of new feature decor tiles we've got, i wonder if it might be worthwhile to have an underlying feature for decor that expected to get retiled and renamed by the vault. or possibly even get better feature-redefine handling in the code. 09:38:49 <08nicolae> also there's some sick art in the dcss reddit https://www.reddit.com/r/dcss/comments/1ezkkf9/serpent_of_cocytus/ 10:51:36 <09gammafunk> https://www.reddit.com/r/dcss/comments/1ezkkf9/serpent_of_cocytus/ 10:52:40 <06pleasingfungus> very good. 11:09:26 <08nicolae> splash screen worthy? 11:10:33 <06pleasingfungus> sure, why not? 11:51:58 for a quick splashoverview https://xom.world/splashes/ 11:53:10 i should add text to that dumb page to at least link source , i will 13:52:46 03Medrano8302 07* 0.33-a0-8-g4b97a1a549: Fix Android touchpad scrolling 10(23 hours ago, 1 file, 10+ 1-) 13https://github.com/crawl/crawl/commit/4b97a1a5491a 13:52:46 03Medrano8302 07* 0.33-a0-9-g5a015b80b9: Fix Android mouse detection 10(11 hours ago, 2 files, 2+ 6-) 13https://github.com/crawl/crawl/commit/5a015b80b963 13:53:52 03Medrano8302 07[stone_soup-0.32] * 0.32-b1-9-g3b3b7eff08: Fix Android touchpad scrolling 10(23 hours ago, 1 file, 10+ 1-) 13https://github.com/crawl/crawl/commit/3b3b7eff0895 13:53:52 03Medrano8302 07[stone_soup-0.32] * 0.32-b1-10-g6bbcbaf917: Fix Android mouse detection 10(11 hours ago, 2 files, 2+ 6-) 13https://github.com/crawl/crawl/commit/6bbcbaf91781 14:18:40 kuniqs (L12 SpEn) ASSERT(shadow && shadow->alive()) in 'ability.cc' at line 2589 failed. (D:13) 14:26:19 kuniqs (L12 SpEn) ASSERT(shadow && shadow->alive()) in 'ability.cc' at line 2589 failed. (D:14) 14:29:21 kuniqs (L12 SpEn) ASSERT(shadow && shadow->alive()) in 'ability.cc' at line 2589 failed. (D:14) 14:33:54 kuniqs (L13 SpEn) ASSERT(shadow && shadow->alive()) in 'ability.cc' at line 2589 failed. (D:14) 14:41:20 !crashlog kuniqs xl=13 spen 14:41:21 1. kuniqs, XL13 SpEn, T:14086 (milestone): https://underhound.eu/crawl/morgue/kuniqs/crash-kuniqs-20240824-213354.txt 14:45:26 hm, shadow had just gone away, but they tried to fire shadowslip. perhaps the ASSERT should instead tell the player their shadow isn't there. but _find_shadowslip_affected and its caller would need to be refactored, or its caller needs to verify that the shadow exists 14:45:48 (can't see what its caller is, it's static so the symbol doesn't show up in the stack trace) 14:50:48 nemmind, only called by find_ability_targeter which is in the stack trace. so either _find_shadowslip_affected needs to be a maybe, or the ability needs to verify that you have a shadow before calling it 15:05:26 <04dracoomega> I... I was pretty sure you couldn't use the ability at all without an active shadow 15:05:57 <04dracoomega> (I'll double-check) 15:14:27 make sure that works in the quiver, from some of the bugs I've seen go by I think it bypasses some checks? 15:14:33 <04dracoomega> Okay, you can't. The issue seems to be that... yes, was just about to say that 15:14:58 <04dracoomega> Apparently you can just quivver it and it will crash on trying to aim it, even though it's greyed out to indicate it's not usable 15:15:22 <04dracoomega> Which seems like... some fault on quiver's end, honestly 15:15:43 <04dracoomega> (Like, it 'knows' it's not usable if it's also greyed out, so what is it doing?) 15:18:45 <04dracoomega> ability_action does seem to call check_ability_possible() 15:38:50 Unstable branch on underhound.eu updated to: 0.33-a0-9-g5a015b80b9 (34) 15:54:57 <04dracoomega> As far as I can tell, the gist of it is that using an ability via the quiver constructs the targeter before checking whether an ability is valid to use (like, in general). This is since the method to determine whether an ability is targeted is just 'ask for its targeter'. It just so happens that the constructor for Shadowslip's targeter happened to assume that it would be impossible to request this if the ability was unusable. 15:55:31 <04dracoomega> Probably instead of asserting, it shoul just return an empty target list 20:04:39 <02Monkooky> blue eyes three headed dragon??? 20:39:40 <07acrobat> @dracoomega may I ask you a question for this rmsl change? https://github.com/crawl/crawl/commit/359da3ace002495de0c602a96a3ce3de531338e9 > In player hands, this is a buff if the player is below ~18 EV and a nerf if the player is above that in the commit, was ~18 EV calculated for a single target beam only? 20:43:37 <04dracoomega> I'm afraid I can't specifically remember now (the version of that commit that got put on a branch first did not remove the special-casing for penetrating beams and that part of the message may not have been edited afterward. Or ~18EV may have been an approximation made by some blending of the two cases together in a vaguely-approximating-reality way). I'm afraid that isn't a very satisfying answer >.> 20:44:18 <07acrobat> thanks 21:51:57 03DracoOmega02 07* 0.33-a0-10-gbe950636f5: Fix a crash when trying to use quivvered Shadowslip with no shadow 10(6 hours ago, 1 file, 6+ 1-) 13https://github.com/crawl/crawl/commit/be950636f5ac 21:51:57 03DracoOmega02 07* 0.33-a0-11-gafe045b255: Fix canine familiars being left alive on past floors 10(5 hours ago, 1 file, 0+ 2-) 13https://github.com/crawl/crawl/commit/afe045b2550e 21:51:57 03DracoOmega02 07* 0.33-a0-12-g8ec6a561cd: Fix Shadowslip having a misleading duration when used repeatedly (Acrobat) 10(5 hours ago, 1 file, 7+ 1-) 13https://github.com/crawl/crawl/commit/8ec6a561cd59 21:51:57 03DracoOmega02 07* 0.33-a0-13-gc0fac641d8: Allow casting BBB on deep water (Dilly) 10(5 hours ago, 1 file, 4+ 4-) 13https://github.com/crawl/crawl/commit/c0fac641d874 21:51:57 03DracoOmega02 07* 0.33-a0-14-g9311ea8977: Fix a niche crash with displaying future XL-based mutations 10(5 hours ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/9311ea8977ee 21:51:57 03DracoOmega02 07* 0.33-a0-15-ge8922636c0: Don't leak the presence of monsters in the Refrigeration targeter 10(5 hours ago, 3 files, 7+ 3-) 13https://github.com/crawl/crawl/commit/e8922636c006 21:51:57 03DracoOmega02 07* 0.33-a0-16-g5a265af164: Don't let cosmetic clouds block catoplepas breath (among other things) 10(3 hours ago, 1 file, 0+ 2-) 13https://github.com/crawl/crawl/commit/5a265af164c9 21:51:57 03DracoOmega02 07* 0.33-a0-17-g3094c0ca69: Stop auto-explore rest over deep water with expiring temp flight 10(2 hours ago, 1 file, 1+ 0-) 13https://github.com/crawl/crawl/commit/3094c0ca699b 21:51:57 03DracoOmega02 07* 0.33-a0-18-g2c1f918eca: Remove some old runrest_ignores 10(2 hours ago, 1 file, 0+ 7-) 13https://github.com/crawl/crawl/commit/2c1f918ecac1 21:51:57 03DracoOmega02 07* 0.33-a0-19-g6b671fb534: Display Makhleb's Unleash Destruction hit-chance when targeting 10(27 minutes ago, 1 file, 1+ 0-) 13https://github.com/crawl/crawl/commit/6b671fb5340d 21:51:57 ... and 2 more commits 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-11-g4cf1de86e0: Fix a crash when trying to use quivvered Shadowslip with no shadow 10(6 hours ago, 1 file, 6+ 1-) 13https://github.com/crawl/crawl/commit/4cf1de86e0b7 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-12-g07d888345d: Fix canine familiars being left alive on past floors 10(6 hours ago, 1 file, 0+ 2-) 13https://github.com/crawl/crawl/commit/07d888345d24 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-13-gcfcf4e89b4: Fix Shadowslip having a misleading duration when used repeatedly (Acrobat) 10(5 hours ago, 1 file, 7+ 1-) 13https://github.com/crawl/crawl/commit/cfcf4e89b4cd 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-14-ge82b1b836b: Allow casting BBB on deep water (Dilly) 10(5 hours ago, 1 file, 4+ 4-) 13https://github.com/crawl/crawl/commit/e82b1b836b82 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-15-g18a91a6079: Fix a niche crash with displaying future XL-based mutations 10(5 hours ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/18a91a6079a4 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-16-gd6131a0170: Don't leak the presence of monsters in the Refrigeration targeter 10(5 hours ago, 3 files, 7+ 3-) 13https://github.com/crawl/crawl/commit/d6131a01701e 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-17-gce390522bf: Don't let cosmetic clouds block catoplepas breath (among other things) 10(3 hours ago, 1 file, 0+ 2-) 13https://github.com/crawl/crawl/commit/ce390522bfb2 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-18-g81615af6d8: Stop auto-explore rest over deep water with expiring temp flight 10(2 hours ago, 1 file, 1+ 0-) 13https://github.com/crawl/crawl/commit/81615af6d886 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-19-g44997244a2: Remove some old runrest_ignores 10(2 hours ago, 1 file, 0+ 7-) 13https://github.com/crawl/crawl/commit/44997244a2ff 21:53:09 03DracoOmega02 07[stone_soup-0.32] * 0.32-b1-20-g5f5b096286: Display Makhleb's Unleash Destruction hit-chance when targeting 10(28 minutes ago, 1 file, 1+ 0-) 13https://github.com/crawl/crawl/commit/5f5b096286ff 21:53:09 ... and 2 more commits 22:06:44 Stable (0.32) branch on underhound.eu updated to: 0.32-b1-10-g6bbcbaf917 22:10:31 <07acrobat> Continuing the conversation, since the rmsl change, I've been feeling like it became much better than before. So I did an experiment today and the results were pretty surprising. monster purple draconian fired quicksilver bolt at 51.0 EV player 500 times in each version. in 0.31, it hit 38 times and missed 462 times(repelled 31 times) in 0.32, it hit 31 times and missed 469 times(repelled 121 times) From these results, it appears that 22:10:31 even at ~50 EV, it was a buff for penetrating bolts. According to the commit, your intent was to make itemization decisions a little more diverse by making rmsl better for low EV characters. But I suspect that flat +15 EV actually made itemization decisions narrower by making rmsl just op. Imo it needs some nerf. (I don't have a strong opinion on monster rmsl, but it probably doesn't need a nerf.) 22:21:05 <04dracoomega> Hmmm... (If I recall right, almost all of my original number crunching was under the assumption that it would maintain the same behavior it had regarding piercing/non-piercing attacks, and later on elliptic and PF argued to remove that entirely. I am not sure much was adjusted afterward, and perhaps it could stand to be. I don't suppose you have numbers on hand for any other scenarios?) (If anything, I think monster repel missiles 22:21:05 might be a little undertuned in its present state, though, given how many characters I've had that still comfortably power through it without really adjusting their behavior in the face of it.) 22:34:13 <04dracoomega> Wait, wait. Why would it say that it repels 4 times as often if it still only hit approximately the same number of times in total? 22:35:12 <04dracoomega> (I... I was never entirely confident that my math for determining if a miss was a result of the added EV from Rmsl - in order to print those messages - was actually right) 22:35:36 Unstable branch on crawl.develz.org updated to: 0.33-a0-21-g2325b17bb0 (34) 22:35:48 <04dracoomega> Since it more or less couldn't reuse the old code for doing so directly. So I wonder if there's additionally a message bug there? 22:35:55 <04dracoomega> Unless I misunderstood what you said 22:36:43 <07acrobat> I don't know under what conditions the 'repelled' message comes up, so I don't know why at all. but still, I'm pretty sure I didn't mess up the experiment! 22:37:48 <07acrobat> I think it was a good choice to remove the piercing/non-piercing distinction (for simplification). I don't have anything else besides this experiment result (I actually don't know how to read code or use proper calculation tools, sorry) 22:37:51 <04dracoomega> But to be clear, when you say missed 469 times(repelled 121 times) you mean that of the 469 times that it didn't hit, it claimed 121 of those times it was repelled? 22:38:05 <07acrobat> yes 22:39:05 <04dracoomega> So, what that reads to me is that it's actually of very similar power on a 51 EV character to what it was in 0.31 against piercing beams, but the messaging is different. After all, all that matters is how often it hits 22:39:35 <04dracoomega> (The messaging might have an actual bug) 22:40:17 <07acrobat> yes (38 vs 31 isn't very similar though) 22:41:34 <07acrobat> maybe similar on ~55 ev char? 22:41:39 <04dracoomega> I'm not sure how much larger that is than the margin of error here? (At least, results of fsimming against the same monster feel like they can vary almost as much from test to test, though I admit I'm not sure the sample size it uses) 22:41:58 <04dracoomega> (I'm curious now, and will check) 22:43:44 <04dracoomega> Apparently the default for fsim is 4000 rounds 22:44:53 <04dracoomega> (...you did all that miss counting by hand, didn't you? =/) 22:45:04 <07acrobat> I did 22:45:08 <04dracoomega> That sounds miserable T.T 22:45:13 <07acrobat> wasn't that hard lol 22:45:20 <07acrobat> I colored it 22:45:22 <07acrobat> with rc 22:46:11 <07acrobat> I mean I counted 500 quicksilver bolt firing (with force more rc) and counted 'hit' and 'repelled' 22:46:25 <04dracoomega> It's perhaps worth noting that when I was doing numbers for examining the repel missiles change originally - and constriction and such that were all done together - I was just trusting that what the game said my to-hit and to-be-hit chance was were correct. (Though this doesn't account for spells, of course) 22:46:46 <04dracoomega> Whether, of course, they may or may not have been in one direction or the other 22:48:50 <04dracoomega> (Like, my statement of 'approximately equivalent at X' was a simplification, since the breakpoint of equivalence varies with every single monster in the game. I just tried to be like 'something broadly average across large chunks of the game') 22:54:29 <04dracoomega> Though after a certain point, exactly how it compares to a past balance state is more academic than anything. While a useful yardstick sometimes, it's less important exactly how equivalent something was to the past than whether the present state is good or not on its own merits. Regarding repel missiles making itemizations decisions less nuanced by just overpowering other options, I find it an interesting statement since I've seen 22:54:29 basically no evidence that other people treat it that way or actively prefer it to other cloak options. Do you think they're incorrect not to do so? 22:59:02 Windows builds of master branch on crawl.develz.org updated to: 0.33-a0-21-g2325b17bb0 23:32:13 Unstable branch on cbro.berotato.org updated to: 0.33-a0-21-g2325b17bb0 (34) 23:35:19 <07acrobat> That's right, it's not important how its performance is compared to past. Whether the current performance is in a good balance is important. The reason I compared it to the past was to make following logic. rmsl was already widely accepted as being very good before the rework (but not for low ev char). And (according to the commit) you were trying to change it to be better for low EV char and worse for high EV char. This was because if 23:35:19 you purely buffed it, it would be too good for high EV char, right? And that baseline was ~18 EV, which I think reasonable number. However, looking my experiment, this is still a buff for char with EV much higher than 18 for penetrating bolts, so I think current rmsl is overall just better than the old one which was already very good. > I find it an interesting statement since I've seen basically no evidence that other people treat it that way or 23:35:20 actively prefer it to other cloak options. Yeah it is just one person's opinion that current rmsl is op and can be very wrong. But imo, something like rmsl is not quite easy to quickly recognize its performance (I also just assumed my feeling was probably wrong until I experimented today), or someone reading the commit might just think, 'I have EV higher than 18, so my rmsl must be worse than it used to be.'. 23:40:26 <04dracoomega> It probably is worse than it used to be at high EV for non-penetrating beams (or at least I hope so, because that would indicate an even bigger number problem). Exactly what percentage of incoming damage is those versus penetrating ones across the total of the game, I'm not at all sure. (I'd felt like old opinions on repel missiles were a little more mixed, in that while people thought it was 'good' for high EV characters, they 23:40:26 didn't think it was clearly superior to a +2 cloak. ...I kind of think they were wrong, but also you aren't nearly as likely to find a scarf of repulsion than a +2 cloak in the first place.) In light of this conversation, I wouldn't mind running numbers on it again myself in a few contexts. (Though probably not in the very short-term). Certainly this isn't something I am making any changes to before 0.33, but there are a couple defense-related things 23:40:27 I was planning to take a closer look at then, so I'd be curious to double-check a few things on this front. 23:41:56 <07acrobat> I thought rmsl was clearly much better than +2 cloak for high ev char from mid game 23:42:25 <04dracoomega> I know I preferred to it, but other people seemed more ambivalent 23:43:48 <04dracoomega> (Like, obviously repel missiles is very strong against what it's good against, but how much percentage of meaningful damage are dodgeable projectiles versus 'everything else' that 4 AC would apply to? I really don't have any clear answer on that front specifically, and honestly it's a very hard thing to measure objectively) 23:44:59 <04dracoomega> Not exactly the same as, but somewhat similar to, weighing the value of a resist versus AC, I think 23:45:22 <04dracoomega> "Much stronger against a few things" versus "Less strong but against almost everything" 23:45:39 <04dracoomega> (Repel missiles is less narrow, but I think there's parallels in how it can be tricky to evaluate) 23:45:41 <07acrobat> Yes it's probably impossible to calculate it. feeling of many skilled players is probably a better answer 23:55:35 Monster database of master branch on crawl.develz.org updated to: 0.33-a0-21-g2325b17bb0