00:00:53 <Eksell> Unstable branch on CRAWL.XTAHUA.COM updated to: 0.23-a0-837-ga3e6480 (34)
00:04:36 <gammafunk> ebering: post looks good, all the links are working
00:06:32 <ebering> great!
00:06:41 <ebering> I'll post it and crosspost it a few places
01:51:03 -!- amalloy_ is now known as amalloy
02:11:31 <switch__> wait alexjurkiewicz what exactly does that change do
02:14:32 <bgiannan> huh a little help
02:14:48 <bgiannan> i had the great idea to copy paste HIlariousDeathArthis rc file with virus on CXC
02:15:17 <bgiannan> but now it's taking all the input as actual commands and it's spamming messages in dgamelaunch
02:15:20 <bgiannan> i can't stop it
02:16:03 <bgiannan> ok it stopped
02:16:27 <bgiannan> isn't it a better way of putting a large rc file on a server ?
02:16:48 <gammafunk> bgiannan: best way is pasting via webtiles
02:16:56 <bgiannan> oh great idea
02:17:01 <gammafunk> click edit RC and just paste into the provided window, then save
02:17:16 <gammafunk> I use a script to update mine across servers
02:17:34 <gammafunk> but that's only really helpful if you have a bunch of RCs to manage
02:18:12 <bgiannan> btw can i split my rc in different files on a server ?
02:18:23 <gammafunk> you can, but you have to use accounts to do that
02:18:34 <bgiannan> i see
02:18:37 <gammafunk> I think hda does that for some of his RC stuff
02:18:50 <gammafunk> but I do that too, just using like two accounts; one is for 'speedrun' stuff
02:19:15 <gammafunk> just include = ACCOUNTNAME.rc
02:19:24 <bgiannan> yes
02:19:34 <bgiannan> but hda is not on cxc
02:20:08 <bgiannan> is dgamelaunch still maintained ?
02:21:11 <gammafunk> how do you mean maintained?
02:21:23 <gammafunk> it's running on cxc and console access works fine
02:21:44 <gammafunk> it's true that pasting a large amount of RC text into e.g. virus is not a great idea
02:21:45 <bgiannan> i meant is someone adding new features to it
02:22:08 <gammafunk> no
02:35:01 <amalloy> bgiannan: you can also just enter insert mode (by pressing i) before pasting into virus
02:35:32 <bgiannan> yes i know i just messed up
04:02:46 -!- amalloy is now known as amalloy_
05:49:04 <alexjurkiewicz> switch__: reduces the priority of gdb spawned when the game crashes
06:06:58 -!- muffindrake1 is now known as muffindrake
09:10:16 <advil> bug # 24, what is up with initial positioning in formatted scrollers right now
09:10:31 <advil> is there a decision on when feature freeze is?
09:32:26 <gammafunk> advil: I was going to ask you about that since I figured you had maybe changed that recently!
09:32:39 <gammafunk> but yeah, ctrl-p now takes you to the top of the log, which is pretty bad
09:32:40 <advil> not intentionally
09:32:53 <advil> well, in my experience (which isn't systematic testing) it only sometimes does that
09:33:01 <gammafunk> oh really, interesting
09:33:08 <advil> also &? has been taking me to the bottom of the help screen
09:33:18 <advil> which is pretty new I think?
09:33:35 <gammafunk> feature freeze is I think this Sunday unless ebering feels only Friday will do
09:33:43 <advil> ok
09:33:53 <advil> wondering if my secret project can squeeze in ;-)
09:33:59 <gammafunk> ooh, what is it?
09:34:05 <gammafunk> you can trust me
09:34:18 <gammafunk> my project is to make runed clear doors the default runed door
09:34:58 <advil> https://github.com/rawlins/crawl/tree/seeded
09:35:54 <gammafunk> *gasp*
09:36:13 <gammafunk> is that what I think it is
09:36:28 <gammafunk> so people could actually really play the same seed?
09:36:31 <advil> probably, or at least it's a step forward, I still haven't solved all the problems
09:36:53 <gammafunk> how did you deal with the "alternate choices for level visitation leads to different vault selection" issue?
09:37:12 <gammafunk> oh or is this full pregeneration
09:37:14 <advil> I have a version of bel's pregeneration
09:37:29 <advil> the rng improvements in the first commit still make a huge difference for deterministic randomness though
09:38:12 <advil> still a bunch of things on the checklist before I can decide whether to merge it
09:38:33 <advil> it's mainly an offline feature right now though so I realized this morning that it would benefit from being part of a release
09:39:30 <gammafunk> I'm guessing that things like the abyss and pan are sort of fundamentally incompatible with this?
09:40:23 <advil> pan is more or less ok actually, though I haven't tested it -- each branch has independent rng state, so every new pan level just advances the state deterministically
09:41:00 <advil> as long as the rest of the dungeon is pregenerated, there's no interaction with vault choice
09:41:31 <gammafunk> that's cool
09:42:59 <advil> abyss, I'm not really sure - though it has its own rng too
09:43:15 <advil> there's perhaps a case to be made that abyss shouldn't be determined by the seed
09:43:34 <gammafunk> yeah; player choice literally dictates when new areas are generated
09:44:13 <gammafunk> well, not so much "when" but "in what way", I guess
09:44:16 <advil> right teleportation would break any dungeon generation determinism
09:44:58 <advil> maybe?
09:45:30 <gammafunk> I could see teleports being fine in the same sense your approach works fine with pan
09:45:47 <gammafunk> but walking towards the edge of the level generates new sections as well
09:45:53 <advil> well, I think it would because of the way terrain gets recycled which may interact with player position
09:46:16 <advil> but the abyss randomization code is so messy I more or less decided to ignore it for the first version
09:46:48 <advil> I'm not sure that allowing seeded abyss-scumming is a good lifestyle choice
09:47:10 <gammafunk> yeah, having the abyss be a non-deterministic area does seem fine for the idea of people playing the same seed
09:47:22 <gammafunk> although I guess it'd be nice if it were as deterministic as possible?
09:47:29 <gammafunk> so that people could play seeded AK runs
09:47:33 <advil> heh
09:47:47 <advil> yeah the AK start is on my list of things to test
09:47:59 <advil> the very first bit will definitely be deterministic from the seed
09:48:04 <gammafunk> that's a very cool project though
09:49:37 <advil> I've kept it 95% offline for now since I think there may be a culture issue with online games and the "permanent record"...I suspect we'd at least want to keep seeded games out of sequell, or under a special designator
09:50:41 <advil> also, I'm worried that pregeneration will be a lot harder on server cpu. I have some ideas for mitigating that but I want to get the main pieces in place before optimizing
10:00:52 <gammafunk> for the cpu generation issue, would that involve  not actually generating the level until the player entered, but somehow ensuring that the result ended up the same way regardless of order of visitation?
10:01:08 <gammafunk> s/cpu generation/cpu utilization/
10:05:22 <advil> yes, one idea is to have a fixed order of generation, and if the player skips around relative to that, pregenerate some of the missing levels
10:13:45 <|amethyst> advil:  you could also pregenerate just a seed for each level, and generate that level based on the seed when it is visited
10:14:20 <|amethyst> advil:  oh, right, that doesn't help with vaults
10:14:46 <advil> yeah, it would be nice if that works, but it doesn't
10:14:48 <advil> uniques also
10:15:13 <|amethyst> hm, and I guess taking the portal from Lair:6 to Hell probably involves skipping a lot of levels
10:15:27 <|amethyst> (for example)
10:15:31 <advil> yes, or even just entering slime:1 depending on the canonical order
10:15:53 <advil> we'll see, I haven't tested a lot of these cases in practice to know how bad they'd be
11:21:40 <switch__> oh wow, sweet.
11:21:53 <switch__> one step closer to replays in tiles :)
12:10:53 <Henzell> Unstable branch on crawl.akrasiac.org updated to: 0.23-a0-837-ga3e6480 (34)
12:47:53 <advil> does the overall dungeon have an official name besides "the dungeon"?
12:51:25 <switch__> what does "nice" mean in that command that alex PR'd
12:53:04 <gammafunk> advil: it's just "Dungeon", yeah
12:53:21 <gammafunk> although I guess somewhere in the manual or something it might be call the Dungeon of Zot
12:54:49 <advil> switch__ nice is a unix command, https://linux.die.net/man/1/nice
12:54:58 <advil> that said, I'm pretty sure the change in that PR doesn't work
12:55:02 <switch__> thanks, i had a lot of trouble finding that
12:55:17 <switch__> looking through linley's readme, it's just "the Dungeon"
12:58:58 <switch__> would someone really PR something without testing it?
13:01:28 <switch__> (speaking of PRs, am I the only one that's a little excited about merging 948? i'm starting to get a little insecure about it)
13:02:18 -!- amalloy_ is now known as amalloy
13:13:57 <advil> switch__: fixing cloud generators is important and I know you keep saying it's "easy", but I've reviewed a lot of PRs and reviewing that is going to take substantial time for whoever does it, it's not easy
13:16:32 <switch__> i think you're overestimating how complicated it is
13:16:42 <switch__> but fair enough, alright
13:18:47 <switch__> from my perspective, i've put in all of the hard work of understanding how to get the code working, and submitting a feature-complete PR, I just need someone to say "yes, this works"
13:22:08 <advil> maybe I'm just speaking for myself, but I don't think any of us is willing to merge code we don't understand ourselves
13:22:44 <switch__> totally; the cloud C++ code isn't complicated at all
13:22:58 <floraline> be prepared to wait months for someone to sign off on it
13:23:11 <switch__> that's ridiculous
13:23:33 <switch__> there's barely anything there, i just moved the framework for calling the place cloud functions from lua to c++
13:24:57 <floraline> even the most trivial PRs can take a long time to review, that's just the nature of volunteer time
13:26:19 <switch__> i think it's different if the PR is actually finished
13:26:58 <floraline> you may be right, all i'm saying is a long timeframe is possible
13:32:39 <ebering> switch__: I said this before, but I'll say it again: I presently don't have time to review that PR before 0.23; even after, I share advil's trepidation at not wanting to merge code I dont understand myself, even if it is "just moving a framework"; your own comments as you've developed it have not been the most confidence inspiring, it seems theres still some mystery about how desolation clouds work?
13:33:20 <ebering> I also indicated that it would greatly help in the reviewing process if you could use the github commenting features to annotate the sticky parts of the PR to aid understanding.
13:33:34 <switch__> there's no sticky parts, it's done, i finished it
13:33:50 <floraline> that's something i want to start using more, the github commenting thing is very cool
13:35:15 <switch__> and yes, i don't know exactly how desolation clouds ran at delay = 0, but that's not something you want anyway, because that's an infinite loop. i don't know how they broke out of the infinite loop, but it also doesn't matter. desolation looks basically the same, and that can be iterated on if need be.
13:36:01 <switch__> (and since nobody knows how to debug the lua in a managable way, nobody will ever find out)
13:37:36 <switch__> you said you didn't have time to review the PR when I was in the middle of working on it and asking questions/for help; you didn't say "if you finish this PR and give me a feature-complete solution, which will noticably impact the performance of all servers running crawl, i still won't make time for it"
13:44:56 <amalloy> it sounds like ebering said he didn't have time to review the PR, and is now confirming he does not have time to review the PR. i think you are underestimating the amount of work and time it takes to review a contribution. nothing nontrivial just gets a quick rubber stamp
13:47:08 <amalloy> i looked through the PR and there are definitely some superficial things that could be improved, but i don't really want to start the formal "review" process because i don't know that i have the time to fully understand the meat of the change
13:50:02 <advil> yes, I noticed some surface things that are easy enough to point out, code style stuff, and that constructor isn't gonna fly
13:50:54 <switch__> sure, the surface stuff isn't important, and is easy to fix, just point it out
13:52:36 <switch__> if someone on the dev team fully understood how it worked, they would have fixed it. it's been an open problem in crawl for 3 years now, and past devs have tried and given up. i don't understand this idea that you have to have the understanding to solve the problem yourself to sign off on completed code.
13:53:14 <amalloy> then it's a good thing we have people who do understand that idea
13:56:48 <floraline> when someone merges someone else's code, they have to be willing to take responsibility for that code. that takes a certain level of understanding, despite the fact that it's already written
13:57:58 <switch__> totally, but that isn't what i was responding to. ebering made it sound like there were problems that i hadn't provided a solution for, or that there was a gap in understanding that could lead to a problem in the PR
14:04:04 <ebering> there's certainly a gap in *my* understanding, and until I have time to fill it I won't be able to merge it. Knowing my schedule, I won't have time to fill that gap in the 0.23 cycle. The time it takes me to bridge that gap could be reduced with some commentary on the PR. I'm in particular leery about this "desolation running at delay 0 should be causing an infinite loop in current crawl but somehow
14:04:10 <ebering> it doesn't"
14:04:46 <ebering> it's great that you've charted the course and I appreciate that. It's always easier to follow someone else's solution than come up with it yourself
14:05:03 <switch__> he didn't provide a solution
14:05:22 <switch__> he just wrote the constructor for the class, which did help
14:06:08 <ebering> I'm not saying that. I'm saying that your PR is what solves the problem, but someone else needs to follow it in order to merge it
14:06:08 <amalloy> switch__: i think ebering is saying it will be easier for him (or some other reviewer) to review your PR than to implement it themselves, and thanking you for your effort
14:06:26 <switch__> oh, that wasn't a jab. sorry, i'm getting impatient with this conversation. i would appreciate it very much if you would look at the code that i spent a lot of time on, and finished.
14:06:28 <switch__> the current state of desolation doesn't matter because you don't want what you currently have. if you set a breakpoint in the middle of it you'll see that you have thousands of clouds that you can't see
14:06:45 <switch__> i'm going to go get coffee and do something else.
14:07:14 <ebering> switch__: we will; no one has time to do so *right now*. I will not until the 0.24 cycle; if someone does not beat me to it I will look at it
14:07:28 <ebering> but, I've got the rest of my life plus other crawl tasks going on
14:38:11 <Cheibriados> New branch created: pull/957 (1 commit) 13https://github.com/crawl/crawl/pull/957
14:38:11 <Cheibriados> 03stenella02 07https://github.com/crawl/crawl/pull/957 * 0.23-a0-834-gb583eb9: Update gen-all.cmd 10(20 hours ago, 1 file, 4+ 0-) 13https://github.com/crawl/crawl/commit/b583eb9c2bd0
14:38:49 <switch__> ^ makes it possible to build with MSVC again
14:41:03 <Cheibriados> 03stenella02 {amalloy} 07* 0.23-a0-838-g3073675: Update gen-all.cmd 10(20 hours ago, 1 file, 4+ 0-) 13https://github.com/crawl/crawl/commit/307367515f00
14:47:28 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.23-a0-838-g307367515f (34)
14:51:49 <Cheibriados> 03stenella02 07https://github.com/crawl/crawl/pull/948 * 0.23-a0-716-g2833ae2: C++ Cloud Generators 10(8 days ago, 9 files, 328+ 94-) 13https://github.com/crawl/crawl/commit/2833ae2ea2ef
15:02:25 <Cheibriados> 03stenella02 07https://github.com/crawl/crawl/pull/948 * 0.23-a0-716-g347e296: C++ Cloud Generators 10(8 days ago, 7 files, 325+ 91-) 13https://github.com/crawl/crawl/commit/347e296d1b84
15:05:29 <switch__> (please merge this msvc pr as well, since it won't build without it; removed the changes from the cloud pr while i was at it)
15:07:42 <Cheibriados> New branch created: pull/958 (1 commit) 13https://github.com/crawl/crawl/pull/958
15:07:42 <Cheibriados> 03stenella02 07https://github.com/crawl/crawl/pull/958 * 0.23-a0-839-g2b2255a: Move a MSVC property to a property file 10(5 minutes ago, 2 files, 3+ 3-) 13https://github.com/crawl/crawl/commit/2b2255a3933d
16:08:31 <travis-ci> The build has errored. (master - 3073675 #10951 : stenella): https://travis-ci.org/crawl/crawl/builds/484050626
16:19:52 <advil> ebering: do you have a fixed qw I can quickly download? I see you have some PRs in that repository
16:39:50 <ebering> advil: one sec, I can put a branch in my copy that has them all applied
16:43:32 <ebering> advil: https://github.com/ebering/qw/tree/patched
16:43:47 <ebering> qw sometimes tries to use abilities while Engulfed, which I haven't fixed
16:49:52 <advil> thanks!
16:50:16 <advil> need to run it a bunch with dungeon pregeneration on
16:54:54 <advil> hm, wonder if qw will reliably win on a seed where it wins
17:00:10 <advil> in fact its behavior could be deterministic given a seed
17:21:03 <advil> well, it's possible that the previous seeding was enough that that was already true
17:22:59 <alexjurkiewicz> advil:seeded is amazing
17:23:34 <advil> alexjurkiewicz did you try it? or just looking at the code?
17:23:37 <advil> hm, not deterministic, wonder why
17:24:07 <advil> lmk if you come up with some issue I haven't thought of, the main missing piece I know about is Boris
17:25:29 <alexjurkiewicz> just looking at the code. If you want an experimental branch let me know
17:25:35 <advil> it doesn't do much online
17:25:44 <alexjurkiewicz> I am pretty tempted to set one up with a static seed for all players
17:25:51 <alexjurkiewicz> and just change the seed weekly
17:26:23 <advil> that could definitely be fun, my worry about online seeding is that we need to think about how it works with sequell and cao scoring
17:26:37 <advil> also pregeneration is very heavy right now
17:27:04 <advil> though I suppose some/most explbrs just don't report to sequell
17:30:19 <advil> I might add a daily seed main menu option to offline builds
17:30:24 <alexjurkiewicz> yep. it wouldn't report to sequell
17:30:51 <alexjurkiewicz> first idea for online scoring would be to add a seed=x key to milestones & logfiles
17:31:03 <floraline> would it be possible to save the pregenerated dungeon to disk somehow? so new games on that seed wouldn't have to do the work every time
17:31:24 <floraline> like a save file that still lets you choose your character
17:32:25 <advil> that's possible in principle and would make sense for alexjurkiewicz's idea, I'd have to work through the practical issues
17:32:48 <advil> but pregeneration basically generates savefile chunks
17:34:16 <advil> I have some ideas for not frontloading pregeneration, but I don't think I'll get to them right away if I'm aiming to merge this for 0.23
17:35:38 <advil> alexjurkiewicz also fyi the progress bar ui element doesn't work on webtiles, would probably be best if I implement that first
17:36:54 <alexjurkiewicz> advil: oh a commit message says pregeneration is disabled on webtiles, is that true?
17:37:02 <advil> yeah
17:37:06 <advil> because of the cpu issues
17:37:16 <alexjurkiewicz> advil: i would appreciate if it was enabled just with bel's old UI. We could start weekly challenges immediately
17:37:22 <advil> I wasn't really thinking of a setup like you had in mind
17:37:32 <alexjurkiewicz> build some #hype
17:40:53 <alexjurkiewicz> actually, i can probably do that myself. I'll have a god
17:44:09 <advil> alexjurkiewicz if you are building it, can you do me a favor and tell me what `&ctrl-e` gives for seed 1 D:1? that is a sanity check I haven't done
17:47:10 <alexjurkiewicz> in webtiles?
17:47:16 <advil> shouldn't matter
17:47:16 <alexjurkiewicz> i'm building console locally
17:47:20 <alexjurkiewicz> alright
17:47:36 <advil> D:1 should also be the same for pregeneration or no pregeneration
17:50:18 <alexjurkiewicz> Level build method = random_map_for_place, level layout type  = rooms, absdepth0 = 0
17:50:21 <alexjurkiewicz> Level vaults:
17:50:23 <alexjurkiewicz> _    dpeg_arrival_round_c;     layout_basic;
17:50:37 <alexjurkiewicz> that's from `./crawl -seed 1` built with debug
17:51:58 <advil> well, that is depressingly different
17:52:06 <advil> are you on linux?
17:54:49 <alexjurkiewicz> osx, built with `nice make -j8 debug`
17:54:58 <advil> oh, even weirder
17:55:16 <alexjurkiewicz> i just rebased your branch against master fwiw
17:55:22 <alexjurkiewicz> i mean, after this test
17:55:23 <advil> ohh
17:55:27 <advil> that could change it
17:55:39 <advil> seeds are very sensitive to version
17:56:26 <alexjurkiewicz> This is Dungeon Crawl Stone Soup 0.23-a0-830-g80cebd10a9
17:56:46 <alexjurkiewicz> i get the same start with the same seed on multiple new games at least
17:57:18 <advil> yeah, that part is very reliable in my testing
17:57:35 <advil> but replicating across computers is a real test I guess
17:57:42 <alexjurkiewicz> does character affect it? I chose hofi sbl
17:57:47 <advil> it shouldn't
17:59:42 <alexjurkiewicz> rebuilding gives the same output. Hey, I don't have to turn on pregen-dungeon as well do I?
17:59:56 <advil> nope, D:1 should always be the same
18:00:18 <advil> it's only transitioning levels out of order that can depart from a deterministic dungeon
18:00:52 <advil> I'm rebasing my branch on master and seeing if that changes anything
18:01:04 <advil> this tells me I really need to do some kind of testing on multiple devices though
18:04:39 <alexjurkiewicz> ok
18:04:56 <alexjurkiewicz> can you push that up so I can be testing the same ref as you
18:06:11 <advil> ok, force pushed
18:08:19 <advil> hm, updating to current master didn't change D:1
18:11:21 <advil> (for reference I see lightli_arrival_statue_moat as the entry vault plus a bunch of other stuff)
18:11:52 <alexjurkiewicz> Level build method = random_map_for_place, level layout type  = rooms, absdepth0 = 0
18:11:54 <alexjurkiewicz> Level vaults:
18:11:56 <alexjurkiewicz> _    dpeg_arrival_round_c;     layout_basic;
18:11:58 <alexjurkiewicz> ditto
18:12:04 <alexjurkiewicz> just a different ditto :D
18:20:32 <amalloy> advil: listening to this discussion about the order of visiting levels requiring pregeneration made me think of splittable RNGs. is that something you've looked into already? it seems like a good fit at first thought
18:24:23 <amalloy> the general outline being, instead of actually pregenerating levels in a fixed order, we could: when generating stairs to a different level, split the current level-gen RNG. one side of the split keeps generating the rest of the current floor, and the other is saved to disk as the seed for the other floor
18:25:07 <advil> amalloy: the problem is that generation choices for vaults, uniques, and randarts is not independent
18:26:20 <amalloy> more specifically, those things depend on things the player has done?
18:28:28 <advil> yes...toy example: say you go to D:3 and it generates vault X which can only generate once. However, the same seed might have tried to generate X on D:2. Now when you go up to D:2, you'll necessary get a different layout (even with the same seed) than if you had used stairs to get there
18:28:43 <advil> er, was supposed to be, say you fall down a shaft to D:3
18:29:04 <amalloy> this has gotta be like the fourth time i've suggested this and someone has reminded me of this same basic problem with my suggestion
18:29:08 <advil> heh
18:29:30 <advil> the most extreme cases are when an entire layout gets vetoed for reasons along these lines, which just leads to the level builder trying again from scratch
18:29:41 <advil> though tbh I have *no* idea how much this comes up in practice
18:30:03 <advil> alexjurkiewicz I just tried my same binary on a different computer, and it's getting the same results. So maybe something's different in the build process? are you using clang?
18:36:00 <amalloy> you could still defer pregeneration in a lot of cases, i think? start with the approach i gave above, but also impose some linear order on levels. when you attempt to generate a level X that comes "after" some other level Y, you have to first generate Y
18:36:35 <amalloy> this fixes the theoretical problems but has a pretty small gain: you now have to support both pregenerated levels and saved seeds, so the code is more complicated, and all you've gained is less up-front worldgen time
18:41:31 <alexjurkiewicz> advil: yes
18:41:38 <alexjurkiewicz> $ clang --version
18:41:40 <alexjurkiewicz> Apple LLVM version 10.0.0 (clang-1000.11.45.5)
18:42:58 <alexjurkiewicz> amalloy: it would save time in the early game which could be important
18:43:17 <amalloy> yeah, i guess people die a lot early
18:43:20 <alexjurkiewicz> if you have to later generate "the rest of the world" when players decide if they go to zot or hell, that would be pretty rare
18:44:09 <alexjurkiewicz> but yeah, generate D levels on demand would be nice for server admins
18:44:18 <amalloy> well, and for players
18:44:58 <amalloy> if you're doing a seeded run and wait 8 minutes for the whole universe to be generated, then die to a D:2 gnoll, that kinda sucks
18:44:58 <alexjurkiewicz> pfft, players
18:45:11 <amalloy> (where 8 minutes is just some time i made up; i have no idea really)
18:53:25 <alexjurkiewicz> advil: hahahahahaha your progress bar is tops
18:56:55 <advil> :-)
18:57:02 <advil> it's more like 30s on my local machine
18:59:21 <alexjurkiewicz> ok, I think to host this on CPO I need two things: 1) works in webtiles (just mpr'ing progress messages would be fine initially), generates one branch at a time
18:59:29 <alexjurkiewicz> *2) generates one branch at a time
18:59:56 <alexjurkiewicz> by which I mean, when you start the game only dungeon is generated, then if you enter another branch, it will generate that branch and any previous unvisited branches
19:00:24 <alexjurkiewicz> actually, would that work? Or would you end up with diverging layouts
19:02:34 <alexjurkiewicz> maybe what I want to do is pre-generate a single save file, then when a player starts a "weekly challenge" game, copy that save file to their account and somehow edit it to change the name
19:07:49 <alexjurkiewicz> hm, the name isn't in there in plain text. annoying
19:10:48 <switch__> why is the build segfaulting o_O
19:13:41 <alexjurkiewicz> did I dream it, or did crawl once have a command-line parameter to edit save files
19:14:55 <advil> alexjurkiewicz that would work, but it'll take a bunch more effort to implement
19:14:59 <advil> -edit-save
19:22:50 <alexjurkiewicz> hm cool, so I can extract the you chunk. But I still need to parse it, apparently
19:26:02 <alexjurkiewicz> is the save file format documented anywhere?
19:26:27 <switch__> i remember reading about it a little bit when looking through level builder docs
19:26:37 <switch__> might be something there
19:29:32 <alexjurkiewicz> $ dd conv=swab < you.cs 2>/dev/null | strings | grep -- -
19:29:35 <alexjurkiewicz> ti0:2.-30a8-15g-9536f32e1b
19:29:37 <alexjurkiewicz> :)
19:32:42 <geekosaur> tags.cc has some, but mostly as code
20:59:43 <alexjurkiewicz> pregen is cool to see what levels take a long time to generate
20:59:46 <alexjurkiewicz> the answer is: depths
21:02:59 <alexjurkiewicz> hm, i wish -playable-json showed recommended combos
21:03:37 <Cheibriados> 03advil02 07* 0.23-a0-839-g6d2b917: Don't build for checkwhite/unbrace travis tests 10(39 seconds ago, 2 files, 6+ 0-) 13https://github.com/crawl/crawl/commit/6d2b9173c93b
21:05:32 <advil> alexjurkiewicz I plan on using this at some point to take an optimization pass through levelgen
21:06:40 -!- amalloy is now known as amalloy_
21:11:23 <advil> also, I implemented a travis test that outputs vault names for D:1 for a bunch of different seeds, and early evidence suggests that there's definitely something leading to instability across devices or build configurations, so I won't be merging this until I figure that out :=/
21:11:36 <advil> although so far two distinct travis builds have been consistent with each other!
21:16:04 <alexjurkiewicz> I remember reading an article about someone trying to do this for nethack
21:16:18 <alexjurkiewicz> and finding every third run would fail because something was doing gettimeofday() % 3
21:16:23 <advil> heh
21:16:32 <alexjurkiewicz> so uh. good luck!
21:16:53 <advil> well, it's super stable within machines as far as I can tell
21:17:07 <advil> one thing I've been wondering is if some of the floating point random calls interact with fp weirdness
21:17:19 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.23-a0-839-g6d2b9173c9 (34)
21:20:27 <advil> also, qw had non-deterministic behavior which was a bit surprising
21:20:41 <advil> since it's using the same rngs
21:24:28 <alexjurkiewicz> how are strings stored in save file chunks? Are they null terminated or is there a length field?
21:25:01 <advil> I'm happy with my state of not needing to know that
21:25:10 <advil> what is it that you're doing?
21:25:28 <alexjurkiewicz> Creating a "weekly challenge" game mode
21:25:37 <advil> if it's copying dungeons around, I almost want to say you'd be better off copying the dungeon than editing the player
21:25:52 <alexjurkiewicz> with pregen dungeon, I can generate a save file with a static dungeon and specific character
21:26:00 <advil> though I don't think I'd try to do this with --edit-save
21:26:16 <alexjurkiewicz> the idea is when someone picks "weekly challenge" game mode, I'll copy the master save to their account, update the save file name, and then load it
21:26:37 <advil> or you could take a save file from the player, a save file with the dungeon, and swap out the you chunk from the former to the latter
21:27:29 <alexjurkiewicz> it seems like the name is in the chr chunk, but i see what you're saying. let me try that
21:32:58 <alexjurkiewicz> advil: you're a genius
21:42:07 <advil> did that work?
21:43:54 <alexjurkiewicz> yes, writing a script to do it automatically now
21:45:21 -!- amalloy_ is now known as amalloy
21:47:55 <floraline> i'm excited to hear how this turns out, that's such a cool idea
21:53:04 <alexjurkiewicz> ah, one problem is I need to match the species/background between the games as well
21:58:09 <advil> I was wondering what you had in mind for that
22:08:10 <alexjurkiewicz> i'm hacking it. I'll generate both saves with the same data
22:10:08 <alexjurkiewicz> BACKGROUND=$(strings chunk | tail -1) :)
22:12:42 <travis-ci> The build passed. (master - 6d2b917 #10955 : advil): https://travis-ci.org/crawl/crawl/builds/484175124
22:18:51 <alexjurkiewicz> weird. Trying to edit the save file too quickly after quitting crawl causes errors
22:22:03 <alexjurkiewicz> god damn, it works. I can generate a master save file, and then give it to any other user
22:22:13 <alexjurkiewicz> now to put it on CPO
22:23:33 <advil> keep in mind that the seeding is likely to change once I figure out what's going on across systems
22:24:14 <advil> not that that matters for exactly what you're doing, but the seed numbers my branch is currently printing may be a bit meaningless
22:24:40 <alexjurkiewicz> advil:does your branch allow webtiles to play with -seed ?
22:24:52 <advil> that should work
22:24:58 <alexjurkiewicz> nice
22:25:16 <advil> the rc file option won't work
22:25:24 <advil> but the clo one should
22:27:29 <alexjurkiewicz> kool
22:31:40 <advil> alexjurkiewicz oh, it's maybe worth knowing that this branch increments a minor version saves rng state, so mixing and matching with trunk would be a bad idea
22:31:55 <advil> there are other things that would go wrong too I think, some map marker stuff
22:32:40 <alexjurkiewicz> should be fine, won't have to do anything like that
22:50:38 <alexjurkiewicz> "# apt install expect" [...] "After this operation, 213 MB of additional disk space will be used."
22:52:33 <alexjurkiewicz> hm, dang. you can't pregenerate levels with a webtiles build even if you play with non-webtiles output
22:53:08 <alexjurkiewicz> advil: could you implement webtiles support just with no output? That would be sufficient for my needs right now
22:53:49 <advil> pregeneration worked for me in webtiles, but it shows nothing while it is working
22:54:24 <advil> oh, you may be running into that it is disabled with DGAMELAUNCH
22:54:37 <advil> so there are some #ifdefs you would need to remove
22:56:25 <alexjurkiewicz> yeah
23:02:27 <advil> look in startup.cc (for the pregen code itself) and in initfile.cc and options.h for the option
23:02:48 <advil> or just look for DGAMELAUNCH in https://github.com/rawlins/crawl/commit/a198bc671b77fda172ca570942270833e435aa1b#diff-41de3124e6c7d06e1e50d68f96e6990c
23:15:22 <alexjurkiewicz> it's done!
23:15:31 <alexjurkiewicz> you can now start a weekly challenge game on CPO :)
23:15:56 <alexjurkiewicz> caveats: pregen doesn't work yet, so the only thing that's the same is d:1 and the seed
23:59:47 <Eksell> Unstable branch on CRAWL.XTAHUA.COM updated to: 0.23-a0-839-g6d2b917 (34)