03:33:17 Experimental (bcrawl) branch on underhound.eu updated to: 0.23-a0-5208-geafff8c3b6 05:07:05 Unstable branch on crawl.akrasiac.org updated to: 0.32-a0-1535-gb4869f8 (34) 05:44:41 <06mumra> Didn't get a chance to look at this yet ... i wonder if alpha is somehow ending up at a negative value and underflowing to some very high value ? Definitely didn't think there was a way for this to happen, must be some strange conflagration of circumstances ... 08:00:06 <04dracoomega> I stared at the codepath a bunch, but couldn't see an obvious angle for it to be bugged. It may be worth noting again that it happened intermittently to the people it was affecting. Or rather, it would happen every time one person cast OTR for some period of time, and then went back to normal again at some point of time - despite there being multiple unrelated flashes during that period. For the person with opaque berserk, saving and 08:00:06 reloading seemed to fix it. I wondered a little at the code that only updates the flash alpha of a cell if it's changed, and if it were possible for junk data to somehow not be properly refreshed? But I have only a poor grasp of what is going on behind the scenes with the json here. 08:02:00 <06pleasingfungus> if this happens often enough, it’d be nice to try to bisect it to a specific commit 08:02:10 <06pleasingfungus> unless we’re sure about which one of the batch is at fault? 08:02:56 <04dracoomega> It's from the recently-merged blindness batch, which is what added variable transparency to flashes in the first place 08:03:13 <04dracoomega> So it's clear what code is handling this. It's just not clear why it sometimes doesn't work. 08:03:20 <04dracoomega> As I said earlier, I wasn't ever able to reproduce locally 08:04:01 <04dracoomega> (I disabled variable alpha as a stopgap measure, but this does make the blindness effect look worse - especially for glitter pies, which are really quite noisy at default transparency) 08:04:08 <06pleasingfungus> is this the same issue as ‘berserk red is los only’? 08:05:11 <04dracoomega> Well, the 'LoS only' part was intentional. That's how things like that have always worked locally (I didn't even realize webtiles was different) 08:05:30 <04dracoomega> But apparently webtiles could (until recently) only flash the entire screen at once for anything 08:06:02 <04dracoomega> (Which makes a bunch of things like Refrigeration a lot louder) 08:06:36 <04dracoomega> And mumra made it able to take the same hitfuncs that specify exactly what part of the screen to flash 08:08:04 <04dracoomega> But in a related commit, he also made it so that flashes could have manually specified alpha (so that translucent overlays were always ~40% opaque and could be less in a graduated way). And that seems to confusingly sometimes make them 100% opaque instead, inconsistently. 08:08:44 <06pleasingfungus> gotcha! 08:09:38 <04dracoomega> And that is the problem I was unable to figure out (I worry it relates to some browser weirdness, but it's a little out of my depth) 08:11:19 <06pleasingfungus> catching up on other discord’s backlog: various people were arguing for removing autotraining 08:11:31 <06pleasingfungus> i think that would be a mistake, for similar reasons to those you alluded to 08:12:14 <06pleasingfungus> dcss has a ton going on; it is very good that new players can avoid worrying about a major piece of complexity for a bit 08:13:01 <06pleasingfungus> they will die because of tactical choices, not because of skilling, so no sense bothering them with it yet 08:13:13 <06pleasingfungus> i do think we could simplify autotraining quite a bit 08:14:46 <06pleasingfungus> eg removing focus options 08:14:49 <04dracoomega> There's merit in "Don't make new players have to worry about something", but it seems more or less universally agreed that the current system doesn't even make great 'simple' decisions. 08:15:04 <06pleasingfungus> i agree with that, but it doesn’t matter 08:15:11 <06pleasingfungus> it doesn’t need to make good decisions 08:15:31 <06pleasingfungus> it just needs to make ones that are vaguely passable in d:1-8 08:15:31 <04dracoomega> Wait, do you mean removing the ability to focus a skill to give it more weight when multiple skills are enabled? I use that regularly. 08:15:33 <06pleasingfungus> which it does 08:15:42 <06pleasingfungus> with autotraining? 08:15:49 <04dracoomega> No 08:15:52 <04dracoomega> Does it work differently there? 08:16:38 <04dracoomega> Or is this just "If you're not micromanaging things in general, you don't need to be able to micromanage them at all?" 08:17:00 <06pleasingfungus> yes 08:17:02 <06pleasingfungus> exactly 08:17:08 <04dracoomega> I mean, that's fair 08:19:13 <06pleasingfungus> the code is also laughably baroque and i’m sure there’s some simplifications that could still provide reasonable results 08:19:18 <04dracoomega> Anyway, I didn't feel strongly about removing autoskilling. More that I was unconvinced at how much better a result it achieves for players not interacting with the system than something much simpler. 08:20:28 <04dracoomega> Like "Yes, manually you never interact with is 'worse', but will it be worse enough in a way that someone who never looks at the skill menu will even be capable of noticing, between all the other parts of the game they don't know?" 08:21:36 <06pleasingfungus> it’s a reasonable question 08:22:07 <06pleasingfungus> it’d be good to dig up the old writeups from the tutorial design folks 08:22:24 <06pleasingfungus> the ones from the university 08:23:39 <04dracoomega> I don't believe I was aware of these 08:57:47 <06pleasingfungus> found http://crawl.develz.org/wordpress/wp-content/uploads/2014/02/DCSS_Survey_summary.pdf, which isn’t it 09:02:04 <06pleasingfungus> ah, https://www.reddit.com/r/dcss/comments/13fln90/dcss_tutorial_usability_testing_report/ links to https://crawl.develz.org/wiki/doku.php?id=dcss:usability_project:start- 09:02:09 <06pleasingfungus> latter is probably what i want 09:05:03 <06pleasingfungus> nothing at all about skilling in the old link… 09:10:22 <04dracoomega> Neither one of them. (But honestly, reading the newer one, if we're talking about people still having difficulties navigating the interface at all or what keys to press or noticing that an enemy they are fighting died instead of vanished, then we're quite a few steps away from anything involving skills mattering) 09:11:08 <04dracoomega> Not meaning to sound dismissive, but 'physical play the game' has comes before any consideration of playing it effectively 09:12:03 <04dracoomega> (In a vacuum, I don't disagree with the sentiment that sound effects and more obvious animation cues would help with onboarding, but that alas seems a very tall order for the framework we're working in at thsi point) 09:12:28 <06pleasingfungus> indeed 09:12:43 <06pleasingfungus> sound is sort of supported… 09:43:57 <06mumra> Some kind of procedurally generated synth noise kind of thing would be great 09:44:42 <06mumra> I have some look at the flash code and I think you are most likely right that the fla value is getting initialised to some larger number and then never sent because the code believes it's already been sent as 0 09:46:13 03dolorous02 07https://github.com/crawl/crawl/pull/3882 * 0.32-a0-1536-g1ba4b7552e: Reflavour Terence. 10(3 days ago, 3 files, 38+ 20-) 13https://github.com/crawl/crawl/commit/1ba4b7552e40 09:47:54 <06dolorous_84348> Ported the Erolcha stuff to Terence. No more Abyss references (death/afterlife references instead), and no more references to eating players, since that seems to be more of an ogre thing. The branch still has the same name, but I can't change that. 09:50:30 <06dolorous_84348> Unless I open a new PR, but that seemed like overkill. 09:51:21 <06dolorous_84348> Like what, specifically? 12:53:27 promet (L11 TrMo) ASSERT(missile.base_type == OBJ_MISSILES) in 'throw.cc' at line 856 failed. (D:10) 14:11:09 <06mumra> Ah kind of like very retro adlib-type beeps and boops 14:12:44 <06mumra> So you have lots of knobs to tweak to vary noises procedurally so you can have a lot of different noises for different monsters, weapons etc 14:17:00 <06mumra> Actually webaudioAPI would be pretty nice for doing this in webtiles, but c++ i would have no idea 15:18:37 <06dolorous_84348> Ah. That does sound interesting. 15:36:50 Unstable branch on underhound.eu updated to: 0.32-a0-1535-gb4869f84db (34) 16:26:29 <04dracoomega> So, I ended up going down a bit of a rabbit hole today, investigating if there was any specific reasoning for why mindless monsters (which will otherwise always enter any clouds without concern) will always refuse to enter miasma specifically. I'd gotten the impression over the years that miasma was a lot worse for monsters than it actually appears to be. It applies the exact same amount of poison as poison clouds, and does a whole 16:26:29 1 more direct damage per turn. (It's even affected by rPois in the way most poison damage is - while less resistible than poison clouds, it still only does 1/3 damage to rPois monsters). This means that in a lot of cases, it's almost literally just a poison cloud with a 1/3 chance of applying Slow. Which is relevant, of course, but for some reason I'd thought it was like... megadamage cloud. But going back to when this specific check was added, out 16:26:30 of curiosity, it was to replace miasma having no hazard check at all. And what strikes me as especially hilarious is that at this point of time when monsters considered miasma harmless, you know what they didn't consider harmless? Grey smoke. 16:26:32 <04dracoomega> Monsters of animal intelligence had a 50% chance to refuse to enter a cloud of grey smoke specifically (black was fine) but only if they also had no fire resist. Because, presumably, they weren't smart enough to realize there wasn't a fire there. >.> 16:30:24 Did 0.32 corrosion changes take Dis hell effect into account? 16:48:48 <04dracoomega> It should have 16:48:57 <04dracoomega> (And it seems to, when I double-check) 16:49:14 they didn't stick around for an answer 22:34:58 Unstable branch on crawl.develz.org updated to: 0.32-a0-1535-gb4869f84db (34) 22:56:18 Windows builds of master branch on crawl.develz.org updated to: 0.32-a0-1535-gb4869f84db 23:13:12 Unstable branch on cbro.berotato.org updated to: 0.32-a0-1535-gb4869f84db (34) 23:35:34 TheMeInTeam (L20 FoAl) ERROR in 'target.cc' at line 1279: ASSERT failed: r of 0 out of range 1 + 1 (2) .. you.current_vision + 1 (1) (Elf:3) 23:39:08 <04dracoomega> Huh. I am going to guess that is trying to use a lightning rod with LoS 0 23:39:29 <04dracoomega> Which I guess nobody has tried before now 23:48:10 03DracoOmega02 07* 0.32-a0-1536-g4ceacca870: Fix a possible crash with Sign of Ruin (Dhd) 10(4 minutes ago, 1 file, 7+ 3-) 13https://github.com/crawl/crawl/commit/4ceacca8704c 23:54:35 Monster database of master branch on crawl.develz.org updated to: 0.32-a0-1535-gb4869f84db 23:59:55 03Isaac Clancy02 07https://github.com/crawl/crawl/pull/3877 * 0.32-a0-1484-g08c69744be: Allow manifold assault to be cast when it would warn 10(5 days ago, 4 files, 94+ 38-) 13https://github.com/crawl/crawl/commit/08c69744be59