00:33:29 Unstable branch on CRAWL.XTAHUA.COM updated to: 0.23-a0-164-g7711e57 (34) 00:35:04 Stable (0.22) branch on CRAWL.XTAHUA.COM updated to: 0.22.0-60-ga52c2fc 02:59:36 Monster database of master branch on crawl.develz.org updated to: 0.23-a0-164-g7711e57 03:14:37 Stable (0.22) branch on crawl.beRotato.org updated to: 0.22.0-60-ga52c2fc 03:30:29 Unstable branch on crawl.beRotato.org updated to: 0.23-a0-164-g7711e57 (34) 07:43:29 ^stats 07:46:03 ??servers 07:46:03 servers[1/3]: USA: {cao}, {cbro}; CAN: {cjr}; EU: {cdo}, {cue}, {cxc}; KR: {cwz}; JP: {lld}; AU: {cpo} 07:46:09 ??cbro 07:46:09 cbro[1/4]: Server in Georgia, US: crawl.beRotato.org -- ssh port 22, username crawler, keys at http://crawl.beRotato.org/crawl/keys (no password access). In case of DNS problems, use crawl.boylecraft.net. Webtiles at: http://crawl.beRotato.org. Use glyph/sigil ^ for bots 07:46:19 ^tstats 07:46:25 erm 07:48:23 ^status 07:48:23 8 Crawlers. CBRO disk usage=91% (135GB) | RAM usage=22% (4GB)| uptime/CPU= 07:48:23 up 30 days, 5:29, 2 users, load average: 1.05, 1.14, 1.14 (4 Cores) http://status.berotato.org 07:48:44 ^status 07:48:45 8 Crawlers. CBRO disk usage=91% (135GB) | RAM usage=23% (4GB)| uptime/CPU= 07:48:44 up 30 days, 5:30, 2 users, load average: 1.10, 1.14, 1.14 (4 Cores) http://status.berotato.org 09:11:45 is something up with cbro? 09:13:29 well, it looks like *someone* is running an apocalypse game again 09:14:37 it looks throttled this time though 09:14:49 wonder if it still causes lag even so 09:19:18 ok, yeah, this time he's doing it with an explicit 100ms delay 09:21:59 Stable (0.22) branch on underhound.eu updated to: 0.22.0-60-ga52c2fcf6e 09:31:41 I considered various kinds of flood protection last time this came up, but it wasn't clear that it would be worth the effort to implement something that would actually try to deal with bad actors, since we mostly don't have any 09:32:07 simplest approach I came up with for stuff like this would be to add some throttling above a certain turncount (maybe 500000) 09:34:38 why would an apocalypse run be more intensive than a regular game? 09:35:07 I think because 5 calls world_reacts many times very quickly 09:35:24 and that's all that happens on an apocalypse run 09:35:44 you can easily force high cpu locally just by holding 5 10:11:19 yeah, delay(100) is helping over last time he did this, but with his rc I'm still hitting 50% on a core, I don't know how cbro allocates cores but that probably would have an effect on at least some games 10:11:41 why would anyone even bother to do this 10:26:26 it has enough of an effect on my console game to be worse lag than peak tournament 10:27:27 hm, he has explicitly set the delays to 1 10:27:37 which we have at least an informal policy against doing for bots 10:40:07 I was talking to him about how to use clua and mentioned that 5 goes through a bunch of world_reacts 10:40:20 and he should set the rest_delay to his delay also 10:40:28 so that his game is taking one turn per 100ms 10:42:09 but that would be rest_delay = 100 10:42:25 I also suggested he test *offline* until he can figure out how to do this without pegging a cpu 10:44:36 he can't do it online in a realistic timeframe without impacting the server 10:44:40 it's just not possible 10:44:59 well what do you mean reasonable 10:45:13 !lg apocalypserobin ktyp=quit max(dur) x=dur 10:45:14 1. [dur=1d+12:40:50] apocalypserobin the Farming Insei (L3 VpMo), quit the game in the Temple (circular_temple_6) on 2011-10-25 16:08:51, with 67 points after 199999999 turns and 1d+12:40:50. 10:45:52 he needs 10x the number of turns he has now, and it needs to be run at about 1/100th the speed it's going now (rough estimate, working on refining that) 10:46:27 the best solution is to not 10:46:33 isnt a better approach simply to throttle games that are impacting the server? 10:47:03 yes 10:47:08 which I'm working on 10:47:25 though it might be something that is better done outside of the crawl process 10:48:46 I mean, writing solid flood control that actually deals with intentional bad actors is not an easy thing to do 10:49:30 right; maybe something like cgroups or cpulimit would be most suitable 10:49:39 this isn't intentional griefing I'm pretty sure, though I don't think it involves a lot of thought or care for other people 10:50:51 !lg * turns>500000 -log 10:50:52 203. melllvar, XL27 DsFE, T:609479: http://crawl.akrasiac.org/rawdata/melllvar/morgue-melllvar-20180909-051048.txt 10:51:37 distinguishing between the long tail of megazigging and situations like this is kind of tricky also 10:54:47 !lg * turns>1000000 -log 10:54:48 68. BagonaHo, XL27 DsFE, T:1320154: https://webzook.net/soup/morgue/trunk/BagonaHo/morgue-BagonaHo-20180531-121718.txt 11:03:38 !lg * turns>2000000 -log 11:03:39 34. igooo, XL27 FeFE, T:32486737: https://webzook.net/soup/morgue/trunk/igooo/morgue-igooo-20180527-111432.txt 11:09:30 yea, in TOP the python webserver is taking up 99% cpu. that's 1 out of 4 11:09:44 wonder if it would run faster in console 11:10:48 I talked to him last night. he was manually delaying 11:11:50 I said it was causing lag and he needed to do something else. he said he'd try lua. that sounded reasonable since I know bots like qw and gw use a delay to reduce the impact to the server. I used to have to ask Grunt to increase his delay 11:12:46 so that's what I figured I could ask him to do next. if the server impact is basically nil I'm mostly ok with it. 11:16:54 he's calling a delay in lua already 11:17:28 to have a reasonable cpu impact in my testing he'd need to set rest_delay to something like 200 (rather than 1, as it is now) 11:19:20 qw etc work better with this situation because they call a delay nearly every turn unless they rest (which they don't do sequentially, mostly). But calling delay between instances of `5` puts a small delay every 100 calls to world_reactas 11:19:29 *world_reacts 11:22:10 here's what I'm considering adding, which is a bit wacky: past turn 500k, you need to complete at least a ziggurat every 100k turns to not be throttled 11:22:21 specifically to have rest/etc throttled 11:25:39 how does the throttling work? 11:25:55 err. I mean, what is the current behavior? 11:26:22 I got a message from doesnt in discord saying: malcolm says to tell you he's only running his game for 5 minutes at a time to try and work the lag out 11:26:22 he set up a delay but it was still causing lag 11:26:24 last night 11:26:48 I read that last night as 'running every 5 minutes' which seemed like a really long delay so I was OK with it. 11:27:14 currently, he's delaying 1ms between individual CMD_WAIT instances triggered by 5, and the 100ms after 100 of those 11:27:19 I asked him to make the rest_delay change above 11:27:26 that you recommended, advil 11:27:43 his rc has a thing that waits 5 seconds but isn't every called 11:28:10 a 5 second wait might help too, I haven't tested that 11:28:15 you can see his current rc at http://crawl.berotato.org/crawl/rcfiles/crawl-git/MalcolmRose.rc 11:28:40 *isn't ever called 11:29:15 he would need to explicitly call rest_forever for that to do anything 11:29:18 explore_delay=1 11:29:18 travel_delay=1 11:29:18 rest_delay=1 11:29:35 that's what I see. I'm not following the logic you described 11:29:38 wrong file? 11:29:54 oh, it's been edited 11:29:59 so now that has no delay 11:30:07 aside from the 1ms thing 11:30:26 gotta go do an errand, sorry, bbl 11:31:19 server looks better now too. he must be tweaking it still 13:02:10 Stable (0.22) branch on crawl.akrasiac.org updated to: 0.22.0-60-ga52c2fc 13:29:23 Unstable branch on crawl.akrasiac.org updated to: 0.23-a0-164-g7711e57 (34) 16:00:09 !vault minmay_fire_ice_demons 16:00:09 1/1. https://github.com/crawl/crawl/blob/master/crawl-ref/source/dat/des/variable/mini_monsters.des#L2313 16:00:30 this needs to lose the hellion chance since hellions aren't fire themed anymore 16:27:41 technically neither are brimstone fiends 16:27:51 @?brimstone fiend 16:27:52 Brimstone Fiend (041) | Spd: 10 | HD: 18 | HP: 75-117 | AC/EV: 15/6 | Dam: 25, 15, 15 | 05demonic, 10doors, evil, see invisible, fly | Res: 13magic(immune), 04fire+++, 05damnation, 03poison, 04rot, 13neg+++, 13torm | Vul: 12cold, 12drown, 08holy | XP: 3309 | Sp: hurl damnation (3d20) [06!sil], s.torment [06!sil] | Sz: Large | Int: human. 16:28:14 only thing they have as fire theme is that they have rF+++ and cold vuln 16:28:17 @?hellion 16:28:17 hellion (042) | Spd: 12 | HD: 7 | HP: 30-45 | AC/EV: 5/10 | Dam: 10 | 05demonic, 10doors, evil | Res: 06magic(60), 05damnation, 03poison, 04rot, 13neg+++, 13torm | Vul: 08holy | XP: 384 | Sp: call down damnation (3d15) [06!sil] | Sz: Medium | Int: human. 16:29:05 @?ice fiend 16:29:05 Ice Fiend (001) | Spd: 10 | HD: 18 | HP: 75-117 | AC/EV: 15/6 | Dam: 2509(claw)12(cold:18-53), 2509(claw)12(cold:18-53) | 05demonic, 10doors, evil, see invisible, fly | Res: 13magic(immune), 12cold+++, 03poison, 04rot, 13neg+++, 13torm | Vul: 04fire, 08holy | XP: 2882 | Sp: b.cold (3d27) [06!sil], s.torment [06!sil] | Sz: Large | Int: human. 18:22:35 Unstable branch on underhound.eu updated to: 0.23-a0-164-g7711e572f3 (34) 18:23:27 gammafunk: what commands does the tournament use to fetch rcs? in the docs it says they're in the dobrazupa crontab 18:53:58 ??objstat[https 18:53:58 objstat[2/2]: Spreadsheets for all released versions and trunk: https://drive.google.com/folderview?id=0B7VXhHzhWWb7S282VWhLVWRXbG8&usp=sharing ; See the README for details: https://docs.google.com/document/d/1D5mFqVi8ghz_nzvVmDUc3unx8VanVBWfgvZ8xCHaiJo/edit?usp=sharing 18:55:09 thank you for this readme ^ 18:55:27 do you run 1000 iters just because it's cheap, or do you need that many for numbers to stabilise? 18:56:20 are there any undead creatures with constriction? 19:04:31 DES question: does `MONS: crimson imp / iron imp` cause all 1s I place to be a mix of both monsters, or all one or the other? 19:08:08 ebering: just a simple wget 19:08:08 I can give you the crontab if you want, but yeah it's very low tech 19:08:18 ok cool 19:08:39 didn't know if there was any fanciness 19:09:04 just fetch them into a dir and the python script has code to iterate over the files, read them, match a pattern 19:09:11 which you can see in the repo 19:09:16 yep 19:09:48 alexjurkiewicz: 1k iters isn't even all that stabilized, but it takes like 8 hours already to run all 4 jobs with 4 cpus working in parallel 19:11:13 and re: that MONS question 19:11:29 dang 19:11:42 that defines the monster slot as "50% chance each for crimson imp or iron imp for each instance of that glyph" 19:12:05 if you want it to be all one or the other, define two slots, then use a SUBST like SUBST: 1 : 12 19:12:22 so that the color SUBST picks 1 or 2 and replaces all instances with that choice 19:12:50 gotcha. nope I'm after variety 19:13:00 alexjurkiewicz: anaconda zombie/skeleton can constrict 21:27:37 hi <3 21:27:37 due: You have 2 messages. Use !messages to read them. 21:27:41 !messages 21:27:42 (1/2) nixor said (about 1y 50w 3d 19h 59m 36s ago): I miss you. I hope your pool is okay. Respect!! 21:27:46 !messages 21:27:46 (1/1) minqmay said (about 1y 50w 3d 1h 44m 55s ago): people keep sending you messages in case you come back, so if you do, you might as well know about this https://crawl.develz.org/tavern/viewtopic.php?f=17&t=19305&p=265193#p265193 21:28:30 :-? 21:29:02 Napkin: ping 21:29:57 Napkin: i'm not sure what my cdo account email address is for console/dgamelaunch or what my password is, can you reset it? 21:30:26 !message minmay hi <3 21:30:26 You probably meant !tell, or possibly !messages. 21:31:15 hmmm, the napking is afk for 1 week :( 21:32:44 * alexjurkiewicz waves 21:34:42 yo 22:05:43 just here to say hi or got the dev itch? 22:07:17 New branch created: pull/847 (1 commit) 13https://github.com/crawl/crawl/pull/847 22:07:17 03alexjurkiewicz02 07https://github.com/crawl/crawl/pull/847 * 0.23-a0-144-g67ccab1: Upgrade orb run spawns 10(41 minutes ago, 1 file, 14+ 10-) 13https://github.com/crawl/crawl/commit/67ccab1f512c 22:12:12 He returns in due time 22:14:01 kinda both, not sure to be honest