00:18:53 <ebering> does freeze include branch+tag or do we do that later
00:19:17 <Cheibriados> 03ebering02 07* 0.23-a0-900-ga39c873: Changelog through 0.23-a0-899-gbbae0a335a 10(59 seconds ago, 1 file, 29+ 3-) 13https://github.com/crawl/crawl/commit/a39c87378c14
00:22:25 <advil> I think freeze is branch+tag
00:22:37 <advil> ?
00:22:52 <ebering> that's what the release guide in docs/develop says
00:22:57 <ebering> I guess we're a day overdue
00:32:08 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.23-a0-900-ga39c87378c (34)
00:34:01 <Eksell> Unstable branch on CRAWL.XTAHUA.COM updated to: 0.23-a0-899-gbbae0a3 (34)
00:36:09 <ebering> well, I guess I'll go ahead and branch+tag
00:36:16 <advil> sounds good to me
00:36:19 <ebering> unless there are any HANGONS
00:36:44 <advil> I have some bugs to fix and I will add a menu item, but I think those can go later
00:36:54 <advil> so, hm, should `x(y(a), z(b))` have a stable order of execution y and z in c++?
00:37:06 <advil> I would have thought so, but evidence suggests otherwise
00:37:20 <amalloy> advil: in C it certainly doesn't. in C++ i think it also doesn't but i'll have to check
00:37:38 <advil> ok
00:37:43 <advil> do you know if it is just up to the compiler?
00:38:56 <amalloy> advil: you mean, is it implementation-defined, or what?
00:39:04 <advil> yeah, I guess
00:39:17 <advil> well, I suppose it must be implementation defined
00:39:31 <amalloy> no, it doesn't even have to be that
00:39:41 <advil> I wonder if this is a gcc vs clang thing that I just haven't detected
00:39:58 <advil> I think my travis test so far might have been coarse enough to miss this
00:40:18 <advil> the thing that is causing seed variation between windows and unix builds right now is a call that is basically that, `coord_def(random_range(...), random_range(...))`
00:40:23 <amalloy> https://en.cppreference.com/w/cpp/language/eval_order - there is no ordering between the arguments of a function call
00:40:48 <advil> the two random_ranges may run in either order on different builds, which immediately gets the rng out of sync
00:41:21 <advil> thanks
00:41:26 <advil> (this has been an exciting preview of what it is like to debug seeding)
00:42:07 <amalloy> i just tried to paste the relevant section from that page but it right-justified my IRC client somehow
00:42:44 <amalloy> anyway (3) and (15) look most relevant
00:43:50 <advil> well, the fix is easy enough but given this I should audit all random calls to make sure that there aren't more instances around, so I'll fix this tomorrow
00:45:55 <Cheibriados> New branch created: stone_soup-0.23 (0 commits) 13https://github.com/crawl/crawl/tree/stone_soup-0.23
00:58:28 <floraline> is it time to install / push players on to the new branch?
00:58:48 <floraline> also, it's my understanding that the next trunk commit will necessarily contain the next devel tag, is that correct?
01:02:45 <ebering> yes
01:05:11 <alexjurkiewicz> advil: my sympathies. But also exciting you've figured it out
01:05:43 <Cheibriados> 03GenericPseudonym02 {ebering} 07* 0.23-b1-1-g1e53798: Quiet confirm_action in the ability menu (11738) 10(27 hours ago, 1 file, 13+ 10-) 13https://github.com/crawl/crawl/commit/1e5379801fce
01:10:44 <floraline> is that yes to both?
01:13:10 <ebering> floraline: sorry; yes we would greatly appreciate it if you installed the new branch;
01:13:24 <ebering> chei is confused about the description of that commit, I pushed the tag second
01:16:37 <alexjurkiewicz> and mark it as beta? will do
01:17:16 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.24-a0-0-g1e5379801f (34)
01:21:57 <Cheibriados> Riposte + constriction + stairs bug 13https://crawl.develz.org/mantis/view.php?id=11847 by CanOfWorms
01:36:43 <Kelbell> Stable (0.23) branch on crawl.kelbi.org updated to: 0.23-b1-0-ga39c87378c
01:40:44 <bgiannan> huge lags on cxc
01:49:40 <Kelbell> Webtiles server stopped.
01:49:46 <Kelbell> Webtiles server started.
02:09:42 <Cheibriados> Enemy transmuters: corrosion status disappears when they change form 13https://crawl.develz.org/mantis/view.php?id=11848 by GrumpyCat
02:11:22 <Kelbell> Webtiles server stopped.
02:11:30 <Kelbell> Webtiles server started.
02:21:05 <travis-ci> The build passed. (stone_soup-0.23 - a39c873 #10996 : Edgar A. Bering IV): https://travis-ci.org/crawl/crawl/builds/485748211
02:39:01 -!- amalloy is now known as amalloy_
03:07:44 <floraline> i wonder if it's feasible to allow webtiles to reload config.py without restarting
03:07:48 <floraline> with only minimal changes
03:08:18 <floraline> if i get the time i might like to look into that
03:16:44 <travis-ci> The build passed. (0.23-b1 - a39c873 #10997 : Edgar A. Bering IV): https://travis-ci.org/crawl/crawl/builds/485748312
04:15:46 <alexjurkiewicz> how does the scores file work? Is it always sorted?
05:05:31 <travis-ci> The build passed. (0.24-a0 - 1e53798 #10999 : GenericPseudonym): https://travis-ci.org/crawl/crawl/builds/485753088
08:31:41 <Cheibriados> Vaults 3 Floor Extremely Small (Due to Ghost Vault?) 13https://crawl.develz.org/mantis/view.php?id=11849 by Kuprin
09:58:06 <advil> man, barely anything guarantees eval order within expressions in c/c++
09:58:43 <advil> gonna have to rewrite a lot of elegent short bits of code to deal with this
10:15:19 <Napkin> gammafunk: i was down with the flu - now need to catch up at work
10:15:31 <Napkin> gammafunk: how much time would it take?
10:15:32 <gammafunk> oh, glad you're feeling better :)
10:15:50 <gammafunk> Napkin: I don't believe it would take too much time for you to set up? It's mostly just creating the database we can use
10:15:53 <gammafunk> from that crawl user account
10:16:08 <gammafunk> however there is the question as to whether the script would be ok CPU wise?
10:16:18 <gammafunk> it's not very intensive or anything, but I'm not sure if it's a problem
10:16:20 <gammafunk> oh, and disk space
10:16:25 <Napkin> we'll see
10:16:39 <Napkin> disk space is quite low at the moment, less than 100gb
10:17:12 <Napkin> would that be enough?
10:17:14 <gammafunk> looks like we're using about 5.8 GB for all past tournament
10:17:28 <Napkin> ah, that's not much
10:17:29 <gammafunk> yeah, 100gb should be just fine
10:17:38 <Napkin> would you do the work?
10:17:40 <gammafunk> yes
10:17:47 <gammafunk> well myself and maybe another dev with a login
10:37:43 <Henzell> PowerComplex (L1 FoFi) Crash caused by signal #15: Terminated (D:1)
10:54:00 <gammafunk> looks like we're going to try to host this tourney on CDO if the load average doesn't prove to be too much somehow
12:25:52 <Henzell> Unstable branch on crawl.akrasiac.org updated to: 0.24-a0-0-g1e53798 (34)
12:25:53 <gammafunk> ebering: looks like you make 0.23 branch? just need to let me know so I know to contact server admins
12:26:03 <gammafunk> s/make/made/
12:26:09 <gammafunk> %git
12:26:09 <Cheibriados> 07GenericPseudonym02 {ebering} * 0.24-a0: Quiet confirm_action in the ability menu (11738) 10(2 days ago, 1 file, 13+ 10-) 13https://github.com/crawl/crawl/commit/1e5379801fce
12:26:21 <gammafunk> %git stone_soup-0.23
12:26:21 <Cheibriados> 07ebering02 * 0.23-b1: Changelog through 0.23-a0-899-gbbae0a335a 10(12 hours ago, 1 file, 29+ 3-) 13https://github.com/crawl/crawl/commit/a39c87378c14
12:26:55 <gammafunk> Medar: stone_soup-0.23 is available for you to install on CXC when you have a chance
12:38:30 -!- amalloy_ is now known as amalloy
12:44:34 <gammafunk> johnstein: stone_soup-0.23 is available for you to install on CBRO when you have a chance
12:46:43 <gammafunk> |amethyst: you probably know about 0.23, but just in case not, stone_soup-0.23 is available for you to install on CAO
12:46:59 -!- amalloy is now known as amalloy_
12:48:06 -!- amalloy_ is now known as amalloy
13:45:14 <ebering> gammafunk: yes I did
13:46:21 <gammafunk> yeah I've contacted people
13:46:37 <gammafunk> and am setting up tournament (hopefully) on CDO now
13:50:22 <Lantell> Stable (0.23) branch on underhound.eu updated to: 0.23-b1-0-ga39c87378c
14:41:46 <Lantell> Webtiles server stopped.
14:41:53 <Lantell> Webtiles server stopped.
14:41:54 <Lantell> Webtiles server started.
15:15:25 <Cheibriados> 03advil02 07* 0.24-a0-1-g14c3f58: Expand out the seed vs generation test for a few more levels 10(2 minutes ago, 1 file, 17+ 1-) 13https://github.com/crawl/crawl/commit/14c3f5860993
15:15:25 <Cheibriados> 03advil02 07* 0.24-a0-2-gcdddf7c: Sequence many, many random calls whose order was undefined 10(8 minutes ago, 32 files, 412+ 145-) 13https://github.com/crawl/crawl/commit/cdddf7c89e99
15:16:57 <advil> lua eval order is also technically undefined
15:17:58 <advil> it might be that all implementations go left-to-right though, I can't quite tell
15:21:37 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.24-a0-2-gcdddf7c89e (34)
15:26:13 <gammafunk> cool, reading through that guide you wrote
15:27:15 <gammafunk> I wonder if there is an easy way to catch  non-stable code
15:27:30 <advil> I did some googling and couldn't find it
15:28:15 <advil> well, the sei thing did suggest some code analyzers that could find call order issues
15:29:42 <gammafunk> so this is something where you want to do this even for non-dungeon-generation code?
15:30:13 <advil> well, in principle
15:30:20 <advil> it may be a bit quixotic right now
15:34:03 <advil> I hope the lua side of this doesn't matter, there's a ton of relevant cases
15:35:02 <Cheibriados> 03advil02 07* 0.24-a0-3-g5a19998: checkwhite 10(16 seconds ago, 2 files, 1+ 2-) 13https://github.com/crawl/crawl/commit/5a199981cbad
15:47:07 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.24-a0-3-g5a199981cb (34)
15:50:14 <alexjurkiewicz> advil: it seems like gcc and clang evaluate in reversed order from each other, let along windows vs unix
15:50:41 <alexjurkiewicz> (according to this link and the one it links to https://stackoverflow.com/questions/28570758/can-i-change-the-order-of-argument-evaluation-in-clang )
15:52:07 <advil> still getting some diverging seeds on my local build vs travis
15:52:18 <advil> windows matches travis now at least
15:53:09 <alexjurkiewicz> osx?
15:53:19 <advil> yeah
15:53:35 <alexjurkiewicz> at least osx matches itself :D
15:54:26 <alexjurkiewicz> btw, the weekly seeding is working great. All levels generation, one attempt per week, highscores 👌
15:56:38 <ebering> advil: I think the lua side should be ok; lua defines an order of evaluation for arguments
15:56:49 <advil> I found people claiming it doesn't
15:57:04 <advil> it's just that there's basically one implementation, that does
15:57:14 <ebering> hhhhhhhhhm. I thought I'd seen that in the 5.1 manual.
15:57:26 <ebering> Getting perl vs. Perl flashbacks
15:57:53 <ebering> language as the only interpreter interprets it vs. language as specified
15:58:34 <advil> it might be partly clang vs gcc but right now travis clang and travis gcc seem consistent with each other, at least in the ways that I'm testing
15:58:59 <advil> well, gotta go do some real work for a while
15:59:02 <geekosaur> except "as specified" might not bother to specify this, ohai C
15:59:04 <amalloy> ebering: fwiw the first search result i found for this was "Evaluation order and assignment order are both explicitly undefined.
15:59:04 <amalloy> "
15:59:17 <amalloy> geekosaur: no, C explicitly specifies it as unspecified
15:59:35 <amalloy> (which imo is different from not bothering to specify it)
16:04:25 <amalloy> though i don't see in the lua spec where it's "explicitly" undefined. it's left unstated in "All argument expressions are evaluated before the call"
16:11:14 <alexjurkiewicz> my new feature request: jumping spiders can be affected by reflection
16:11:20 <alexjurkiewicz> and bounce off you and web some other monster
16:24:40 <travis-ci> The build was broken. (master - cdddf7c #11000 : advil): https://travis-ci.org/crawl/crawl/builds/486090384
18:22:26 <Lantell> Unstable branch on underhound.eu updated to: 0.24-a0-3-g5a199981cb (34)
19:33:24 <Cheibriados> 03advil02 07* 0.24-a0-4-g4a9ad93: Another batch of randomization sequence fixes 10(32 seconds ago, 6 files, 41+ 17-) 13https://github.com/crawl/crawl/commit/4a9ad9352148
19:38:08 <gammafunk> that faerie armour plus code is really weird
19:38:55 <advil> there is much weird code that shows up in these changes
19:38:55 <floraline> are you sure the compiler is not going to optimize away some of these changes?
19:38:55 <advil> yes
19:38:56 <advil> ; guarantees a sequence point
19:39:36 <advil> well, it ought to optimize them a bit, but if it's obeying the c++11 standard, every compiler ought to optimize them in a way that doesn't change order
19:39:49 <floraline> i see
19:40:22 <advil> one way of thinking about it is that the code I'm removing has more freedom for being optimized, because the compiler could choose either order
19:40:31 <advil> of function call
19:40:41 <floraline> that makes sense
19:41:20 <advil> gammafunk hopefully the ugliness of writing these sequenced random calls will make people avoid it in the future
19:41:39 <gammafunk> yeah, one would hope
19:41:54 <gammafunk> but it shows how there's probably a lot of dubious randomization like that we should audit
19:42:18 <advil> yep
19:42:26 <advil> I have changed very little of it, in fact just one call so far
19:42:39 <gammafunk> for the slaying plus?
19:42:49 <advil> yeah, the cursed slaying minus really
19:42:57 <gammafunk> oh that's what that is
19:43:01 <advil> just because it was easier than adding a sequence point there
19:43:19 <switch__> are you implying that these calls with 4 nested random(8-12) aren't balanced with perfect precision?
19:45:08 <advil> sadly, I'm still not getting things to converge, I guess I'm going to have to finally give in and spin up a linux vm
19:45:36 <advil> windows was matching as far as I've tested, at least
19:45:39 <gammafunk> advil: wouldn't that be better as just -(2 + random2(5)) to preserve min, max, and average?
19:47:28 <advil> oh, right
19:47:46 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.24-a0-4-g4a9ad93521 (34)
19:48:25 <switch__> those enormous nested randoms go straight back to linley's code, don't they
19:51:56 <switch__> yep
19:52:30 <switch__> https://github.com/crawl/crawl/blame/8f474f6a0858eddfa1a6552858f6d79fbf7473cb/crawl-ref/source/dungeon.cc#L1492
19:54:12 <Cheibriados> 03advil02 07* 0.24-a0-5-g797c3c0: Fix accidental cursed slaying nerf (gammafunk) 10(30 seconds ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/797c3c0885db
20:00:57 <alexjurkiewicz> couldn't some of these random2(x) + random2(x) calls be replaced with random2avg(2x, 2)? I might be wrong though
20:02:16 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.24-a0-5-g797c3c0885 (34)
20:05:02 <gammafunk> not the same average
20:05:30 <gammafunk> nor the same max
20:06:15 <alexjurkiewicz> rip
20:11:21 <alexjurkiewicz> where is the code that makes simulacrum attacks be cold?
20:11:37 <alexjurkiewicz> in mon-data.h, all simulacrum entries are just AF_PLAIN
20:12:11 <alexjurkiewicz> !source _downscale_zombie_attack
20:12:12 <Sequell> 1/1. https://github.com/crawl/crawl/blob/master/crawl-ref/source/mon-util.cc#L1817
20:12:16 <alexjurkiewicz> thank you ##crawl-dev
20:14:56 <amalloy> random2(x) + random2(x) should be the same as random2(2x - 1, 2), though
20:15:19 <amalloy> er, random2avg in the last case, of course
20:16:52 <amalloy> poor advil
20:17:53 <amalloy> just catching up on the many randomness changes and they look like a pain. i bet some smart aleck will come along and propose refactoring things back to the way they were in at least one case
20:20:05 <advil> yeah, I didn't anticipate this issue at all
20:23:39 <amalloy> advil: for many of the relatively simple cases (but presumably not some complicated ones), you could define a new function that accepts non-random arguments and calls random2 on them in a guaranteed order
20:24:17 <amalloy> eg, int random_sum(int x, int y) { int ret = random2(x); ret += random2(y); return ret;}
20:24:39 <amalloy> then reuse that at many call sites, instead of having to write this painful multi-step addition over and over
20:24:43 <advil> I thought about that, and that's probably the one that would be used most, but I'm not sure it's that many calls
20:24:59 <amalloy> okay
20:25:07 <advil> might be worth it though
20:25:17 <amalloy> what are you using to discover expressions like this that need fixing?
20:25:37 <advil> I'm basically looking at every random call
20:25:47 <advil> I've used some heuristics to prioritize, I guess
20:26:17 <advil> and I've ignored a lot of non-dungeon code
20:27:38 <amalloy> ah, sure
20:29:47 <amalloy> i wonder if there's a clang warning for something like this
20:30:58 <amalloy> gcc seems to have -Wsequence-point but i'd be surprised it if knew that random2 is a function that needs worrying about
20:31:36 <advil> I did find something in the SEI doc (-Wunsequenced) but it didn't do anything for me
20:31:36 <advil> I think it's equivalent to -Wsequence-point
20:31:36 <advil> https://wiki.sei.cmu.edu/confluence/display/cplusplus/EXP50-CPP.+Do+not+depend+on+the+order+of+evaluation+for+side+effects
20:31:39 <amalloy> yeah
20:31:56 <amalloy> so if instead of random2 you were modifying a local, we'd get a warning
20:36:54 <amalloy> sadly i don't see any way to tag random2 with some attribute or something saying "please consider this function as dangerous as mutating a local"
20:41:03 -!- amalloy is now known as amalloy_
20:42:17 <advil> I don't suppose volatile would do it?
20:42:24 <advil> I realize that's not what volatile means
21:02:15 -!- amalloy_ is now known as amalloy
21:12:13 <ebering> I've got a last set of trunk updates for 0.23 plus tournament announcement in a draft on cdo wordpress
21:12:21 <ebering> advil: you may want to elaborate on the seeding
21:12:31 <ebering> gammafunk: or anyone else I'd appreciate a second set of eyes
21:13:30 <advil> esp since it doesn't work yet
21:17:10 <gammafunk> ebering: trap placement in Tomb got overhauled
21:17:24 <gammafunk> not in changelog proper but good for a trunk update post
21:17:37 <ebering> hm ya
21:17:47 <gammafunk> Don't think it has to say much aside from Tomb:2 traps being toned down and traps placing more throughout all tomb levels
21:18:00 <gammafunk> you could mention your new klown vaults too
21:18:07 <advil> that seems fine, I added the word "experimental" to indicate that it might not work
21:18:13 <ebering> ok
21:18:57 <gammafunk> oh, and singing sword got revamped again; could just mention how it uses its sonic wave spell at all tension levels and generally does more damage
21:19:31 <gammafunk> grate trap removal is a good comedy option, guess you can also mention your klown kar vault
21:19:38 <gammafunk> I think that just barely missed the last update
21:19:54 <gammafunk> literally did!
21:20:01 <gammafunk> the commit after the changelog update for last time
21:20:05 <ebering> ha!
21:20:28 <ebering> I went with High Elves being removed again, datafiles merge caused them to appear in -playable-json
21:20:56 <gammafunk> yeah
21:21:09 <gammafunk> but maybe you could mention how it's gone still
21:21:14 <gammafunk> it's kind of an iconic trap
21:21:19 <gammafunk> infamous rather
21:21:32 <ebering> yeah
21:22:19 <gammafunk> high elf removal is obviously excellent to reference when possible
21:22:40 <gammafunk> before that the great dev meme was "Remove Pan"
21:22:50 <gammafunk> Pan referring to the unique grunt made for Forest branch
21:23:23 <gammafunk> although I feel we have quite a few removal memes going these days...
21:27:10 <ebering> maybe we should remove some
21:51:29 <alexjurkiewicz> +amalloy | gcc seems to have -Wsequence-point but <-- there might be a directive to tell gcc "this function [random2] should be worried about
21:51:41 <alexjurkiewicz> oh you said that already
21:51:50 <alexjurkiewicz> I checked sonarqube and it has no such functionality
21:54:11 <alexjurkiewicz> hm, didn't someone run a static analyser on the codebase a while ago? It was a github issue, and ebering fixed some things I think. Maybe that has functionality...
21:56:33 <ebering> I didn't do the fixing, but the issue is still open
22:02:21 <alexjurkiewicz> checking the pvs warnings, nope
22:07:41 <advil> I think I did the fixing
22:08:00 <advil> alexjurkiewicz: that SEI thing I linked to has a list of analyzers that handle cases like this
22:23:25 <Cheibriados> 03advil02 07* 0.24-a0-6-g964d818: Use a separate rng for ghosts 10(4 minutes ago, 2 files, 9+ 6-) 13https://github.com/crawl/crawl/commit/964d818aa416
22:23:46 <advil> now we're getting somewhere
22:52:18 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.24-a0-6-g964d818aa4 (34)
23:35:52 -!- amalloy is now known as amalloy_
23:40:04 <Cheibriados> 03advil02 07* 0.24-a0-7-g1bb0902: Clear unrand state for relevant tests 10(55 seconds ago, 1 file, 1+ 0-) 13https://github.com/crawl/crawl/commit/1bb09026696b
23:47:10 <Kelbell> Unstable branch on crawl.kelbi.org updated to: 0.24-a0-7-g1bb0902669 (34)