00:33:30 Unstable branch on CRAWL.XTAHUA.COM updated to: 0.23-a0-290-g290e36d (34) 02:25:31 did Sequell web api endpoint changed ? I get a 503 on https://loom.shalott.org/api/sequell 02:37:55 is cpo online? 02:38:05 heh 02:38:12 testing sequell behaviour 02:38:20 oops wrong channel 02:38:56 Time limit of 60 exceeded 02:40:16 lol 03:00:06 Monster database of master branch on crawl.develz.org updated to: 0.23-a0-290-g290e36d 03:24:37 Unstable branch on crawl.beRotato.org updated to: 0.23-a0-290-g290e36d (34) 04:35:40 -!- amalloy is now known as amalloy_ 05:40:52 how to update chrome 05:44:57 Erm, sorry, wrong keyboard. 09:47:08 ^status 09:47:09 16 Crawlers. CBRO disk usage=94% (135GB) | RAM usage=18% (4GB)| uptime/CPU= 09:47:09 up 54 days, 7:28, 2 users, load average: 0.51, 0.77, 0.71 (4 Cores) http://status.berotato.org 09:48:20 !versions 09:50:29 &versions 09:50:58 CAO: 0.23-a0-277-ga4718cd, CBRO: 0.23-a0-275-g0cf4261, CDO: 0.23-a0-178-g440b721, CJR: 0.23-a0-290-g290e36dc6f, CPO: 0.23-a0-290-g290e36d, CUE: 0.23-a0-286-gcae3914892, CWZ: 0.23-a0-136-g722f094, CXC: 0.23-a0-290-g290e36d, LLD: 0.23-a0-279-g95cc57954e 09:56:56 johnstein: CBRO trunk seems to be dead, both in webtiles and console. Not sure why, there's no error message, just exits back to dgl/webtiles menu, and it seems to be only CBRO 09:58:59 greensna1k, is https://loom.shalott.org/api/sequell still the endpoint for Sequell's json api ? 10:00:37 a bunch of sequell/shalott web stuff has been getting errors for a couple weeks, so I suspect he needs to fix something 10:11:02 03advil02 07* 0.23-a0-291-ge5a9903: Fix two weapon select bugs 10(2 minutes ago, 1 file, 2+ 1-) 13https://github.com/crawl/crawl/commit/e5a990390b45 10:14:57 something weird happened, it prompted me eventually to transfer my trunk game, then I clicked yes and it was successful, then back to lobby 10:15:01 (on CBRO) 10:15:31 yeah, same thing for me...I'm not sure its about the upgrade though because any of trunk, sprint, tutorial just bounces me back to the lobby 10:15:38 going to try a rebuild after that push 10:16:02 good idea 10:16:55 didn't work before because it claimed to be on 290 (despite &versions) 10:19:19 I logged into the server. did a manual git fetch. no complaints 10:19:38 I'm running a rebuild right now fyi 10:19:48 nothing else on the server has changed. but I did get a note from someone recently that I needed to do some upgrade 10:19:51 but yeah the git part was ok 10:19:53 I haven’t done that yet 10:20:45 that's probably pyyaml, that PR hasn't been merged yet 10:21:59 yeah, that wouldn't cause a problem 10:22:55 other versions work ok? 10:23:03 &versions 10:23:22 CAO: 0.23-a0-277-ga4718cd, CBRO: 0.23-a0-275-g0cf4261, CDO: 0.23-a0-178-g440b721, CJR: 0.23-a0-290-g290e36dc6f, CPO: 0.23-a0-290-g290e36d, CUE: 0.23-a0-286-gcae3914892, CWZ: 0.23-a0-136-g722f094, CXC: 0.23-a0-290-g290e36d, LLD: 0.23-a0-279-g95cc57954e 10:23:34 the &versions output is apparently wrong for cbro 10:24:11 when I first tried to rebuild, it wouldn't let me because it was up to date with 0.23-a0-290-g290e36dc6f 10:25:31 that sounds suspicious 10:26:06 let's see what the rebuild does 10:27:33 it's now on s 10:30:56 which s though?! 10:30:58 Unstable branch on crawl.beRotato.org updated to: 0.23-a0-291-ge5a9903 (34) 10:31:05 neve rmind 10:31:24 hrm 10:31:26 it transfered me 10:31:30 then took me back to lobby 10:31:40 yeah, still broken 10:31:52 johnstein: what happens if you run the latest crawl binary from the command line? 10:34:32 I did an su to the crawler user to play console 10:34:39 maybe something about the new password reset upgrades? 10:34:53 logged into cbrotest. played. it asked me to transfer my game. I did 10:35:01 then it brought me back to the lobby 10:35:16 webtiles log have anything as to what's going on? 10:35:37 every time I press P it takes me right back to the lobby. if I press P a bunch I eventually see flickers for the refresh 10:35:39 is that still over ssh? I'm wondering what happens if you run the binary directly (which should get the regular menu) 10:35:40 I’ll check 10:35:43 yes 10:35:57 I always forget how to run that. I have to run it from 10:36:02 these symptoms would most likely be triggered by the binary not running at all for some reason 10:36:06 within the chroot right? 10:36:31 yeah maybe, rebuild said "Installing game binary (crawl-git-e5a990390b) in /home/crawl/DGL/usr/games" 10:36:39 right 10:36:44 so that's the current binary name I think? 10:36:52 you could copy it elsewhere to test 10:37:00 but I do think the webtiles log is a good first place to check 10:37:07 if there's an error it should show up there 10:37:37 yes, true 10:37:48 I guess if you want to run it like advil says...hrm, can you get a chroot and chuser 10:37:58 guess just running as root from chroot is fine though 10:39:10 I don't see how the password reset stuff could cause this, it should be breaking the whole server (and probably not impacting console), but that's the only server changes lately 10:40:07 yeah, just a shot in the dark 10:40:23 what was that version it was stuck at 10:40:26 275? 10:40:39 0cf4261 10:40:45 %git 0cf4261 10:40:45 07ebering02 * 0.23-a0-275-g0cf4261: Revert "Check bounds for player_can_hear (#689)" 10(35 hours ago, 1 file, 1+ 2-) 13https://github.com/crawl/crawl/commit/0cf42614084f 10:41:58 2018-10-03 14:41:14,932 INFO: #135825 P114421 ERR: /usr/games/crawl-git-e5a990390b: /usr/lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.17' not found (required by /usr/games/crawl-git-e5a990390b) 10:42:53 huh 10:42:56 this might be a good time for me to do some system updates. I’m probly behind on some stuff 10:43:11 interesting how that would just break? 10:43:55 that’s not a new dependency ? 10:44:43 I don't know how the specific version of libstdc++ is getting required 10:45:12 did we use some new standard library feature that bumped the version required? 10:45:24 no, I would have thought it would set that correctly at build? 10:45:26 not that I know of 10:46:38 it comes from the .so at link time, to ensure the .so at runtime matches 10:47:13 so you may have a *.so that isn't the same as the *.so.N it loads at runtime 10:47:45 yeah, there must be some conflict between the library paths at build and run? no idea how this would spontaneously happen though 10:47:48 libstdc++.so.6 doesn't match libstdc++.so in other words 10:49:11 my best guess is that I must have done something weird or nonstandard to get things to work In The Beginning, and for all these years, no one made any updates that would perturb this fragile setup 10:49:15 Till Now 10:49:28 (highly untechnical speculation) 10:49:47 nothing else runs on the cbro server anymore cept my ZNC bouncer 10:49:54 and crawl 10:50:11 can you check the version of libstdc++6 installed in the chroot? 10:51:18 what's the magic spell to do that? I've only messed around with .so stuff when I set up the server and that was like 2014 10:52:24 from chroot, dpkg -s libstdc++6 10:52:29 root@www:/# /sbin/ldconfig -p | grep stdc++ 10:52:29 libstdc++.so.6 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 10:53:27 Package: libstdc++6 10:53:27 Status: install ok installed 10:53:27 Multi-Arch: same 10:53:36 strings /usr/lib/x86_64-linux-gnu/libstdc++.so.6 | grep LIBCXX 10:54:10 GLIBCXX_3.4 10:54:10 GLIBCXX_3.4.1 10:54:10 GLIBCXX_3.4.2 10:54:10 GLIBCXX_3.4.3 10:54:10 GLIBCXX_3.4.4 10:54:11 GLIBCXX_3.4.5 10:54:13 GLIBCXX_3.4.6 10:54:15 GLIBCXX_3.4.7 10:54:17 GLIBCXX_3.4.8 10:54:19 GLIBCXX_3.4.9 10:54:21 GLIBCXX_3.4.10 10:54:23 GLIBCXX_3.4.11 10:54:25 GLIBCXX_3.4.12 10:54:27 GLIBCXX_3.4.13 10:54:29 GLIBCXX_3.4.14 10:54:29 welp 10:54:31 GLIBCXX_3.4.15 10:54:33 GLIBCXX_3.4.16 10:54:37 GLIBCXX_DEBUG_MESSAGE_LENGTH 10:54:40 oh 10:54:42 it stops 10:54:45 right before 3.4.17 10:55:13 johnstein: in that dpkg output, what package version was it? field we wanted was: Version: 10:55:37 Version: 4.6.3-1ubuntu5 10:55:58 oh, I thought cbro was debian, didn't realize it was ubuntu 10:57:07 (side note, does anyone who knows anything about debian have insight about https://crawl.develz.org/mantis/view.php?id=11699&nbn=3#bugnotes? I suspect it's a linux mint problem but I don't really know) 10:58:02 I've got to run 10:58:10 I have to go to work now. I'll check back in later. this could be good motivation for me to do some house cleaning or set up cbro2, rElec boogaboo 10:59:07 yeah it seems that crawl wants a more recent libstdc++6 than you have installed 10:59:21 probably just a chroot upgrade or something?" 10:59:29 I guess you're running 12.04 ubuntu? 10:59:36 hrm, right 10:59:43 maybe you've updated the host but not the chroot 10:59:59 the build happens on the host 11:00:08 so the mismatch is the likely problem 11:00:11 I can't remember what version I'm on. but it's old 11:00:15 I think you need to update packages in the chroot 11:00:19 I'm really really really late on basic maintenance 11:00:47 * johnstein just plays a sysadmin on tv 11:01:31 you could do a dist upgrade on both maybe, but possibly just 'apt-get update' within the chroot would fix this; I'm not sure how you update cbro outside the chroot 11:02:54 a dist upgrade for both is more work, but just updating chroot packages would likely fix the issue, since I think cbro and chroot have gotten out of date 11:03:04 s/date/sync/ 11:10:42 Unstable branch on crawl.jorgrun.rocks updated to: 0.23-a0-291-ge5a990390b (34) 11:12:12 tried to do an apt-get update then upgrade in the chroot, but says nothing updated. which seems really suspicious. really have to head to work now though. definitely warming up to the idea of starting from scratch and just migrating everything over to a newer updated setup 11:12:57 sounds good, thanks for taking a look johnstein 11:14:13 no problem. I’d say the main issue here is that I’ve deferred maintenance for a while and this will end up being a balloon interest payment on some long standing technical debt 11:35:02 in the meantime I'm guessing that somewhere after 275 there's some change that obscurely requires a c++11 feature not implemented in 4.6? I won't have time to look until much later though 11:44:29 that would seem to be the case, yeah 11:45:21 it's the absolute most high priority possible to fix for the sake of the project 11:45:36 because otherwise I can't stream my award-winning OpGl of Nem 11:45:56 the loss of roguelike marketshare from any delay for a fix would be staggering 11:46:15 haha 11:47:11 well the extreme approach would be to do bisecting reverts while rebuilding cbro 11:47:20 hrm 11:47:30 everything since then is a bunch of special-case bugfixes, so not super crucial 11:47:46 advil: could we also just do bisects/rebuilds locally and look for changes in those defines? 11:48:03 like presumably we'd see the GNUCXX value lower 11:48:50 yeah probably, if you use gcc 11:48:55 well, I'm not sure how to inspect crawl to find that info 11:48:56 which I guess you do 11:49:14 yeah, I do have gcc 11:49:24 just not sure how to see what version of that symbol crawl wants 11:49:40 oh it's GLIBC 11:49:42 not GNU 11:50:14 hrm, seems that crawl has a few of them 11:50:30 GLIBCXX_3.4 11:50:30 GLIBCXX_3.4.21 11:50:30 GLIBCXX_3.4.19 11:50:30 GLIBCXX_3.4.15 11:50:30 GLIBCXX_3.4.18 11:50:33 GLIBCXX_3.4.20 11:50:35 GLIBC_2.14 11:50:38 GLIBC_2.4 11:50:40 GLIBCXX_3.4.14 11:50:43 the relevant GLIBCXX ones in my build of crawl 11:51:06 not sure to what extent the ones required vary by system installation 11:51:31 I also wonder...can you set up 12.04 on aws 11:52:22 14.04 is available 11:52:24 not 12.04 11:56:13 Is there any desire for a modified kobold proposal? 11:57:04 like something really goofy like a chaos-branded bite aux 12:00:12 I don't think there's any strong adversity to the notion of changing kobolds; MarvinPA was the last person to rework them, and for any ideas it might be good to get his feedback 12:00:28 ah, I see 12:00:34 thanks :D 12:00:37 I don’t think you can build old versions of trunk via rebuild 12:00:51 and if you bypass that locally it screws up the versions db 12:01:12 hrm, let me try an older checkout 12:01:28 since I think something in the System is hardcoded to assume the last entry is the latest trunk version 12:01:38 and see if my glibcxx even changes; I would guess that this may be tied to which libstdc++6 you actually have 12:01:43 which is how I screwed things up when I was force updating things. 12:02:04 yeah, I vaguely recall what you're talking about 12:02:04 oh but actually that might be the behavior you want. hmmm 12:02:15 hrm, what we might be able to do 12:02:17 since I’m my case I was annoyed the latest got changed 12:02:18 is put up a branch 12:02:59 well that's probably not quite going to work if we can't go backwards 12:03:08 maybe we could start with an old enough version of trunk in a branch 12:03:10 and just go forward 12:03:21 but that would still require setting up a branch for builds on cbro 12:03:52 if I had a generic experimental branch that always did a forced hard reset to the upstream you could use that for arbitrary testing and just force push a version you want to test. 12:03:59 yeah 12:04:07 wasn't sure if you could set up such a branch or not 12:04:30 before you bother doing that, maybe we should hear back from advil though 12:04:50 and I should see if I can mess around with compiling on my system or on an older ec2 instance 12:04:55 14.04 might work 12:07:21 johnstein: but just to be clear, this is to get cbro trunk working before you have time to upgrade cbro itself, since I presume that will take you some time 12:07:31 just so we can make a temporary fix not requiring such a recent GLIBCXX 12:08:20 yup. trying to work through my emails at work to see if I can try to get a branch set up via my phone 12:11:45 is this the right channel to inquire about design stuff or do y'all prefer that in ##crawl? 12:14:34 it's a channel to talk about anything related to development, including design 12:14:50 well, my GLIBCXX strings do change if I compile that last working CBRO version 12:15:35 it's still a mix of versions but the highest version and lowest versions listed for commit 175 are both lower 12:16:13 3.4.20 instead of 3.4.21 and 3.4.9 instead of 3.4.14 12:17:06 i was curious why the inventory cap is 52 12:17:54 52 letters for lowercase and uppercase alphabet character, i believe? 12:18:13 and each one corresponds to an inventory slot, so it becomes 52 items 12:18:37 ohhhhh 12:18:58 <|amethyst> yes, the alphabet. Any solution involving dynamically assigning, or reassigning, letters means the you'd have to go through the item menu (and scroll to the item) whenever you interact with an item 12:19:31 yeah that makes sense 12:19:51 and numbers and etc aren't bound by default so that people can do some remapping? 12:20:54 <|amethyst> digits are used for inscriptions (e.g. if you inscribe an scroll with @r1 you can do r1 to read that item regardless of which slot it's in) 12:21:08 mostestballerwiz: I think you over-presume the degree of design. Linley's Crawl had 52 item slots because so does NetHack and anyway encumbrance sorted it out. Now Crawl has 52 item slots because it always did. 12:21:14 <|amethyst> and other symbols have special meanings, like ! for select all potions, * for show all items, etc 13:07:18 Unstable branch on crawl.akrasiac.org updated to: 0.23-a0-291-ge5a9903 (34) 13:21:51 ??rebuild 13:21:51 rebuild[1/2]: http://crawl.akrasiac.org/rebuild/ http://underhound.eu:81/rebuild/ http://crawl.berotato.org/crawl/rebuild/ http://crawl.xtahua.com/rebuild/ https://crawl.jorgrun.rocks/rebuild/ Bug gammafunk, |amethyst, or Nap.Kin for CDO. Use your powers wisely. 13:21:52 ??rebuild 13:21:52 rebuild[1/2]: http://crawl.akrasiac.org/rebuild/ http://underhound.eu:81/rebuild/ http://crawl.berotato.org/crawl/rebuild/ http://crawl.xtahua.com/rebuild/ https://crawl.jorgrun.rocks/rebuild/ Bug gammafunk, |amethyst, or Nap.Kin for CDO. Use your powers wisely. 13:21:53 :/ 13:22:23 oh lag 13:22:46 gammafunk/advil. I added an experimental “devtest” 13:23:09 it’s only console right now. I can add webtiles later. 13:23:31 all you need is a branch called devtest and a rebuild should do the rest 13:23:43 in console it’s game “x” 13:26:12 johnstein: oh, thanks 13:37:05 it's very hard to see what among those changes could possibly do something this weird 13:37:43 no obvious interactions with >4.6 stuff in this table: https://www.gnu.org/software/gcc/projects/cxx-status.html 13:38:35 I guess the good news is that we can try builds and just see where things change 13:38:48 yeah, have you already created a branch? 13:39:03 the most suspicious thing (and that's not saying much) is the bitwise math change 13:39:14 %git 95cc57954e 13:39:15 07advil02 * 0.23-a0-279-g95cc579: Fix a bad bitwise calculation for ban_glyph in arena 10(24 hours ago, 1 file, 7+ 9-) 13https://github.com/crawl/crawl/commit/95cc57954e75 13:39:19 so I'd probably start with that 13:39:27 I have not, and I'll be away for a couple hours, so feel free 13:39:33 ok 13:39:53 I don't have too much time but I do have a few minutes 13:45:32 gammafunk/advil: FYI if you do a force push, I’ll probably have to do a hard reset in the repo to manually fix. pretty simple to do. just let me know if you get into that situation 13:45:55 ok, shouldn't need to do it that way, since the history on this branch won't matter 13:45:59 (I didn’t tweak the build script to do this automatically but I think it would be easy to do this when I get home) 13:46:01 cool 13:46:18 worried it might be an issue if you do bisect 13:47:47 New branch created: devtest (1 commit) 13https://github.com/crawl/crawl/tree/devtest 13:47:47 03advil02 07[devtest] * 0.23-a0-292-g25fd8f8: Revert to 0cf42614084f78e 10(3 minutes ago, 26 files, 66+ 65-) 13https://github.com/crawl/crawl/commit/25fd8f82c6f2 13:48:07 I'm going to bisect with git revert + cherry-pick 13:48:28 so that just undid the last n commits 13:48:38 but did it as a new commit 13:49:35 rebuild didn't work though 13:49:59 error: pathspec 'devtest' did not match any file(s) known to git 13:50:23 oh got it 13:50:38 I needed to do trunk rebuild so it would. pull and get the branch 14:00:07 I think what happens is, if you rebuild a new branch the first thing it does is a checkout of that branch 14:00:23 which doesn’t exist locally. I’ve been meaning to add a git fetch before that 14:00:54 so sometimes I’ve had to manually git fetch 14:01:01 let me know if it doesn’t work 14:01:39 it's building now 14:04:39 -!- amalloy_ is now known as amalloy 14:04:53 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-292-g25fd8f8 14:17:09 looks like that is working 14:17:18 my game started 14:46:09 Mission Accomplished 14:46:57 shortly to be followed by mashin' accomplished 14:59:02 how did you get into that branch? experimentals still seem to be in the "disabled for tournament" state 15:00:08 well, I can see some of them in dgl, but not that one 15:06:47 advil: webtiles not hooked up yet. in console it’s X but there’s no menu option 15:07:11 on the experimentals screen 15:08:05 webtiles requires a reboot and I like shutting down Rotatell before I do that and editing the json file and doing server stuff is fiddly via my phone 15:08:25 so I literally just phoned the kludge in 15:09:08 The build passed. (devtest - 25fd8f8 #10348 : advil): https://travis-ci.org/crawl/crawl/builds/436758917 15:09:35 advil: yeah, as johnstein said, you go X)perimentals then just hit 'x' to start it up (not listed in the menu) 15:45:00 ah got it 15:47:53 03advil02 07[devtest] * 0.23-a0-293-g173927b: Fix disaster area adjacency check 10(28 hours ago, 1 file, 8+ 7-) 13https://github.com/crawl/crawl/commit/173927b0a9c7 15:47:53 03advil02 07[devtest] * 0.23-a0-294-g38b7d48: Exclude bones from checkwhite and run it 10(28 hours ago, 8 files, 17+ 16-) 13https://github.com/crawl/crawl/commit/38b7d4820179 15:47:53 03ICC02 {advil} 07[devtest] * 0.23-a0-295-gc8c327b: Change some dead links to relative links in new_dev_guidelines.md 10(35 hours ago, 1 file, 2+ 2-) 13https://github.com/crawl/crawl/commit/c8c327b7e5f9 15:47:53 03advil02 07[devtest] * 0.23-a0-296-gae91e53: Fix a bad bitwise calculation for ban_glyph in arena 10(26 hours ago, 1 file, 7+ 9-) 13https://github.com/crawl/crawl/commit/ae91e5378f54 16:03:02 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-296-gae91e53 16:06:00 03advil02 07[devtest] * 0.23-a0-297-ga4409f7: Fix two iterators becoming invalidated during iteration 10(24 hours ago, 1 file, 7+ 4-) 13https://github.com/crawl/crawl/commit/a4409f724be8 16:06:00 03advil02 07[devtest] * 0.23-a0-298-g9762f52: Fix an uninitialized variable 10(24 hours ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/9762f52d1b26 16:06:00 03advil02 07[devtest] * 0.23-a0-299-g44b9e23: Ensure fsim file always gets closed 10(24 hours ago, 1 file, 1+ 0-) 13https://github.com/crawl/crawl/commit/44b9e23f1534 16:06:00 03advil02 07[devtest] * 0.23-a0-300-gd6dcbd1: Remove some vacuous ASSERTs 10(24 hours ago, 2 files, 0+ 4-) 13https://github.com/crawl/crawl/commit/d6dcbd18f93b 16:06:00 03advil02 07[devtest] * 0.23-a0-301-gd095339: Change a + to a += 10(23 hours ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/d09533958561 16:06:49 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-301-gd095339 16:08:36 03advil02 07[devtest] * 0.23-a0-302-g62cfe62: Revert "Change a + to a +=" 10(39 seconds ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/62cfe62989be 16:08:36 03advil02 07[devtest] * 0.23-a0-303-g823101b: Revert "Remove some vacuous ASSERTs" 10(23 seconds ago, 2 files, 4+ 0-) 13https://github.com/crawl/crawl/commit/823101b89b92 16:08:36 03advil02 07[devtest] * 0.23-a0-304-g8f60b78: Revert "Ensure fsim file always gets closed" 10(11 seconds ago, 1 file, 0+ 1-) 13https://github.com/crawl/crawl/commit/8f60b783b4d4 16:09:12 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-304-g8f60b78 16:09:28 !versions 16:09:45 &versions 16:10:00 oh nm I could just do 16:10:04 #git 16:10:08 CAO: 0.23-a0-291-ge5a9903, CBRO: 0.23-a0-275-g0cf4261, CDO: 0.23-a0-178-g440b721, CJR: 0.23-a0-291-ge5a990390b, CPO: 0.23-a0-291-ge5a9903, CUE: 0.23-a0-286-gcae3914892, CWZ: 0.23-a0-136-g722f094, CXC: 0.23-a0-290-g290e36d, LLD: 0.23-a0-291-ge5a990390b 16:10:13 %git 16:10:13 07advil02 * 0.23-a0-291-ge5a9903: Fix two weapon select bugs 10(6 hours ago, 1 file, 2+ 1-) 13https://github.com/crawl/crawl/commit/e5a990390b45 16:10:23 03advil02 07[devtest] * 0.23-a0-305-g5719d83: Revert "Fix an uninitialized variable" 10(15 seconds ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/5719d833a498 16:10:58 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-305-g5719d83 16:11:47 what 16:13:47 if that's really the cause it must be something internal and extremely indirect 16:14:24 03advil02 07[devtest] * 0.23-a0-306-gfe95a3d: Revert "Revert "Fix an uninitialized variable"" 10(15 seconds ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/fe95a3dc8940 16:14:58 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-306-gfe95a3d 16:22:33 looks like it’s still broken. sounds like we have a winner? 16:24:42 hopefully it's not some weird interaction 16:24:50 I have *no idea* why that commit would have this effect 16:25:11 I'm going to check it in a different way on that branch 16:25:19 03advil02 07[devtest] * 0.23-a0-307-g1636827: Revert to e5a990390b45 10(2 minutes ago, 13 files, 23+ 27-) 13https://github.com/crawl/crawl/commit/163682702f08 16:25:19 03advil02 07[devtest] * 0.23-a0-308-gb83c6e6: Revert "Fix an uninitialized variable" 10(72 seconds ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/b83c6e6ccd3b 16:26:47 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-308-gb83c6e6 16:26:57 also, the code without that commit is objectively wrong 16:27:33 wtf, it really is that commit 16:28:13 are there any compile logs I could search for a warning message? 16:29:05 this has got to be some weird internal compiler thing 16:32:32 03advil02 07[devtest] * 0.23-a0-309-g366e295: Try moving a const value out of a function 10(37 seconds ago, 1 file, 18+ 17-) 13https://github.com/crawl/crawl/commit/366e295eb4c5 16:32:58 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-309-g366e295 16:34:08 "Man, your leg is broken! - No, it's okay. - Let me fix it! - What have you done?! Now I feel like it's broken!" 16:35:29 03advil02 07[devtest] * 0.23-a0-310-gc9f9e2d: Try several things at random 10(17 seconds ago, 1 file, 5+ 4-) 13https://github.com/crawl/crawl/commit/c9f9e2d7d701 16:35:53 classic crawl dev? 16:35:58 hah 16:36:13 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-310-gc9f9e2d 16:37:36 none of them worked 16:43:52 03advil02 07[devtest] * 0.23-a0-311-g034c931: Revert to master 10(4 minutes ago, 1 file, 20+ 22-) 13https://github.com/crawl/crawl/commit/034c931f8299 16:43:52 03advil02 07[devtest] * 0.23-a0-312-g4f1ac1f: Replace a bool with a vector 10(56 seconds ago, 1 file, 4+ 4-) 13https://github.com/crawl/crawl/commit/4f1ac1f3390e 16:44:16 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-312-g4f1ac1f 16:44:49 huh 16:44:56 that version also doesn't work 16:51:15 so it's some kind of interaction or something general about that function? 16:51:22 everything looks extremely normal to me 16:54:59 unless anyone has any ideas, I might just revert the fix on master and wait for johnstein to upgrade the chroot libraries 16:56:49 I tried doing a apt-get update and upgrade in the chroot and it didn’t do anything so I probably need to do a dist-upgrade or something. I really don’t understand the nuances of chroots beyond the super basics 16:57:13 this might end up being the catalyst I need to work on cbro 2.0 16:59:56 I sort of expect that this could happen again in an entirely unpredictable fashion 17:00:07 so reverting that commit is very much an emergency patch 17:03:44 03advil02 07[devtest] * 0.23-a0-313-ged3df24: Revert "Replace a bool with a vector" 10(62 seconds ago, 1 file, 4+ 4-) 13https://github.com/crawl/crawl/commit/ed3df24aff0e 17:04:06 03advil02 07[devtest] * 0.23-a0-314-g360067d: One more reimplementation try 10(16 seconds ago, 1 file, 4+ 4-) 13https://github.com/crawl/crawl/commit/360067d88d32 17:04:35 Experimental (devtest) branch on crawl.beRotato.org updated to: 0.23-a0-314-g360067d 17:04:35 wondering if this is something completely off the wall like pushing the data segment over some secret boundary 17:05:16 yeah it really must be 17:05:23 that reimplementation didn't work either 17:06:19 03advil02 07* 0.23-a0-292-g6cc01a3: Revert "Fix an uninitialized variable" 10(6 minutes ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/6cc01a3c85fb 17:06:31 boy that's tough 17:07:55 another guess I had was that maybe there's some libstdc++ bug in that version that later compilers know about, that is somehow instantiated here 17:08:34 but validate_basedirs just looks like really ordinary c++11 to me 17:08:48 I guess I wrote it so I'd think that 17:10:40 wow that's nuts 17:14:47 !source validate_basedirs 17:14:47 1/1. https://github.com/crawl/crawl/blob/master/crawl-ref/source/files.cc#L461 17:15:34 there is some change in how literals work at the next version, that's why I first tried moving the const vector out of the function 17:17:25 the only other semi-unusual thing I see is that there's a conditional call of a noreturn function? 17:17:46 also `for (auto subdir : data_subfolders)` is possibly wrong in a minor way 17:18:57 <|amethyst> what was the error when found was initialised there? 17:19:00 advil: which commit? 17:19:45 what is happening is that it causes the crawl binary on cbro to require a later version of libstdc++ than is present in the chroot 17:20:12 <|amethyst> that's bizarre 17:20:27 the reversion that seems to fix it (now trying on master, tested multiple times on the branch because I could believe it) is 6cc01a3 17:20:32 %git 17:20:32 07advil02 * 0.23-a0-292-g6cc01a3: Revert "Fix an uninitialized variable" 10(20 minutes ago, 1 file, 1+ 1-) 13https://github.com/crawl/crawl/commit/6cc01a3c85fb 17:20:39 *couldn't 17:21:08 <|amethyst> but otherwise, new builds on cbro are working? 17:21:09 Unstable branch on crawl.beRotato.org updated to: 0.23-a0-292-g6cc01a3 (34) 17:21:26 yeah, it's more of an interim solution until johnstein is able to work on the chroot 17:21:30 <|amethyst> my worry is that the revert only made it "work" because of ccache 17:22:21 hm 17:22:36 well, I did a lot of rebuilding, and devtest was built from scratch 17:22:55 6cc01a3 does seem to have worked 17:23:09 and not fast enough that I would have expected ccache to be being used? 17:23:14 <|amethyst> hm 17:23:55 I'm not sure what you have in mind for ccache hiding the effect, could you elaborate? 17:24:29 advil: what if you remove the `found` bool entirely and replace all the if-checks against it with if (true)? i'm wondering if, before the fix, the compiler analyses some codepath and says "well we could only get here via undefined behaviour (reading an undefined variable), so i'll just optimise by eliding this codepath entirely" 17:25:02 and one of those elided codepaths requires linking against some feature that doesn't exist in the relevant libstdc++ version 17:25:34 <|amethyst> hmm... if files.cc is now exactly the same as it used to be, including #includes, that particular file might be "compiled" by pulling the .o from the cache 17:25:37 <|amethyst> but 17:26:04 <|amethyst> if any headers it depended on, including system headers, changed, it wouldn't be pulled from the cache 17:26:08 <|amethyst> so I guess that's unlikely 17:26:12 ah 17:26:48 <|amethyst> amalloy's hypothesis seems reasonable, but I can't figure out what in those if (!found) branches would be any different from stuff we do all over the place 17:27:48 |amethyst: iirc ccache isn't set up on cbro 17:27:59 johnstein: can you confirm whether ccache is working on cbro? 17:28:07 I want to say that he never got to setting it up 17:28:10 <|amethyst> well, that definitely would rule out my hypothesis :) 17:28:19 I could be wrong, but that's my recollection 17:28:31 it rebuilds a *lot* slower than cao, so that's what I was assuming too 17:29:53 the end call (which is NORETURN) is in one of those branches 17:30:16 so we do have those around, but it's not super common 17:30:19 I think I have ccache *kinda* working 17:30:32 yeah, i wondered about end too, but it's used all over the place so i wouldn't expect this instance of it to be a problem 17:30:58 ie I know I installed and set it up 17:31:21 but I’ve always been underwhelmed by the performance gains almost like it wasn’t working 17:31:40 03advil02 07[devtest] * 0.23-a0-315-g61afaa9: Revert "One more reimplementation try" 10(4 minutes ago, 1 file, 4+ 4-) 13https://github.com/crawl/crawl/commit/61afaa96cbde 17:31:40 03advil02 07[devtest] * 0.23-a0-316-g1241a92: Try (a version of) amalloy's suggestion 10(54 seconds ago, 1 file, 2+ 2-) 13https://github.com/crawl/crawl/commit/1241a92dacf9 17:31:53 but in the past when looking into it t seemed like it actually was working. just slow 17:32:49 there must be a single build directory on cbro 17:32:53 advil: dir_exists, though, appears to be usde only in files.cc, and initfile.cc (plus being defined in syscalls.cc) 17:33:42 but i guess that's called outside of any if-check against found 17:33:58 yeah 17:36:22 ccache is installed 17:36:38 ah, ok 17:36:48 try `ccache -s`? 17:37:17 cache size should be maxed if it's consistently being used 17:37:25 or near to max 17:39:10 looks like 1.7 GB out of 2.0 GN 17:39:20 ok 17:39:24 well, something's using it 17:39:25 GIGANEWTONS OF COMPILED BYTES 17:40:10 yea. this is how I remember previous ccache conversations going. confirmation that it’s set up ok but still confused on why it’s so slow 17:40:30