00:10:49 Inappropriate insult by Donald (https://crawl.develz.org/mantis/view.php?id=1702) by Niccus 00:51:37 -!- jarpiain_ is now known as jarpiain 01:30:34 -!- pineappl1 is now known as pineapple 01:51:44 -!- by has joined ##crawl-dev 01:52:31 !lm * type=crash 01:52:32 9. [2010-06-08] BirdoPrey the Sneak (L5 VpAs) ASSERT(item.base type == OBJ CORPSES && item.sub type == CORPSE BODY) in 'misc.cc' at line 155 failed on turn 3218. (D:4) 01:52:38 also, good morning 01:52:44 !lm * type=crash -tv 01:52:45 9. BirdoPrey, XL5 VpAs, T:3218 (milestone) requested for FooTV. 01:53:12 !lm * type=crash -tv:cancel 01:53:21 !lm * type=crash -tv:<0.2 01:53:35 9. BirdoPrey, XL5 VpAs, T:3218 (milestone) cancel requested for FooTV. 01:53:35 9. BirdoPrey, XL5 VpAs, T:3218 (milestone) requested for FooTV. 01:55:50 !lm * type=crash -tv:cancel 01:55:50 9. BirdoPrey, XL5 VpAs, T:3218 (milestone) cancel requested for FooTV. 01:56:13 !lm * type=crash -tv:<0 01:56:14 9. BirdoPrey, XL5 VpAs, T:3218 (milestone) requested for FooTV. 01:57:41 ??henzell 01:57:47 henzell[1/2]: Linley Henzell is the original creator of Crawl. He is also the ##crawl bot. If you want the source for *this bot*, go here: http://crawl.akrasiac.org/henzell/src/ or git clone http://crawl.akrasiac.org/henzell/src/henzell/.git 02:19:12 !tell greensnark For some reason, crash playback seems to start at ttyrec beginning. 02:19:12 by: OK, I'll let greensnark know. 02:36:43 Summon Ugly Thing summons a group in Pan (https://crawl.develz.org/mantis/view.php?id=1703) by evktalo 02:37:02 it always summons a group. 02:48:39 -!- ortoslon has joined ##crawl-dev 02:49:00 xale: No, it does not. 02:49:28 Hey ortoslon - played Hydorah yet? :) 02:49:49 yes, i'm on the last level now 02:51:32 i won't record a video of it though, TimW at http://www.youtube.com/indiegamescom#p/u is doing a good job 02:54:39 I managed to beat the first three levels! \o/ 02:56:16 Hi. 02:57:11 -!- ais523 has joined ##crawl-dev 03:04:27 -!- elly__ has joined ##crawl-dev 03:04:39 -!- sorear_ has joined ##crawl-dev 03:05:14 -!- elly has quit [Disconnected by services] 03:05:19 -!- elly__ is now known as elly 03:06:00 Hi due! 03:06:15 Hi eino! 03:09:15 -!- sorear has quit [*.net *.split] 03:09:16 -!- ixtli has quit [*.net *.split] 03:09:16 -!- Gretell has quit [*.net *.split] 03:09:19 -!- ixtli_ has joined ##crawl-dev 03:13:33 mmm, is -pedantic really necessary? 03:13:45 always 03:14:27 well, it's pretty buggy 03:15:02 elly: for catching errors, or is it actually needed for the compile? 03:15:11 no idea 03:15:13 I hope it's the first, but you never know with Crawl... 03:15:40 like, it complains about commas after the last enum, even though that's specifically allowed in C++98, C++03 and C++0x, not to mention C99. 03:16:16 --std=c99 allows that but you can't specify that for C++ 03:16:34 and that's a point even MSVC gets right 03:17:20 msvc allows a lot of useful non-standard things for what I've seen. 03:18:12 -!- ortoslon has quit [Read error: Connection reset by peer] 03:18:38 the issue with msvc is that it uses a different syntax for many of the nonstandard things than everyone else's 03:18:54 compare C99's %lld printf specifier to MSVC's 64, for instance 03:19:07 *compare C99's %lld printf specifier to MSVC's %I64, for instance 03:19:15 (%I is a control code in this client unless escaped) 03:19:51 If you've got to invent things, you might as well make them understandable. 03:20:28 -!- sorear_ is now known as sorear 03:20:28 I wonder if C ever will be a strict subset of C++ again? 03:20:46 IIRC the standards bodies want to move in that direction, just it's so hard to get anything standardised 03:21:12 not possible, C++ explicitely disallows some constructs 03:21:36 C99 has deprecated some of those constructs, hasn't it? 03:21:38 I wonder what kind of clang that lad yesterday used for Crawl. 03:21:47 maybe C will disallow them too in the future 03:21:47 and try this: "int class;" -- I can think of no way C++ could possibly allow that 03:22:02 The one from Fedora fails horribly at arcane things like and :D 03:22:02 hmm, perhaps 03:22:19 KiloByte: would you consider merging Luca__'s clang fixes? I'm not that comfortable with the makefiles 03:22:25 you might need just the one compatibility include, then 03:22:47 it wouldn't conflict with the C standards to say that #include or whatever would make the word "class" disallowed for use by the user 03:22:57 * Zao pokes Luca__ with a stick. 03:23:01 by: Debian's clang doesn't run C++ yet, lemme try to install it from source 03:23:17 much like #include means that you aren't allowed to mess with the definition of getc 03:23:58 -!- by has quit [Disconnected by services] 03:24:01 -!- by_ has joined ##crawl-dev 03:24:09 -!- by_ is now known as by 03:26:58 unify level id handling (https://crawl.develz.org/mantis/view.php?id=1704) by rob 03:30:35 morning rob 03:31:15 hi due 03:32:10 hmm, where's Gretell off to 03:33:25 @whereis Gretell 03:33:50 (is that an infinite loop?) 03:34:27 casmith789: I've seen botloops, they aren't pretty 03:34:29 but that wouldn't be one 03:37:50 Gretell's in a split 03:40:47 wow, that's a long split, more than half an hour 03:40:52 splits tend to be larger but shorter 03:42:05 Napkin is unsplit (even though he ragequit ##crawl-dev due to having a single bad mudation in a nice set) 03:42:53 in a DS? 03:43:20 btw, never think about compiling clang with -jX without having nearly a GB of memory per job 03:43:52 the system thrashes so bad even irssi takes >1s to respond to keypresses 03:44:05 -!- ogaz has quit [Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539]] 03:47:44 -!- Niccus has quit [Quit: out] 03:51:54 -!- syllogism has joined ##crawl-dev 03:51:57 -!- Textmode has quit [Quit: Over the ages, man has been certain of a great many things. some of these things were even true.] 03:54:22 that would seem to depend on a scheduler. :-) 03:56:18 the linux scheduler fails HARD in one-disk thrash death situations 03:56:22 is yours better? 03:56:54 (actually, any kernel mode I/O situation. 2GB core dump = 3 minute system unusability) 03:57:02 KiloByte: I'm confused as to why ragequitting a game would make you ragequit a channel 03:57:46 ais523: there's quitting, and then there's quitting 03:58:00 hmm, I suppose so 03:58:04 one is the 'Q' command, the other is a verb form most commonly applied to drugs 03:58:53 control-Q, surely? 04:00:00 * ais523 tries to remember what plain Q does in Crawl 04:00:30 ah, it's quiver, just like in NetHack 04:00:39 couldn't remember because I normally just use ( and ) 04:05:41 anyone know why DELAY_DESCENDING_STAIRS passes along you.absdepth0 as a delay parameter? 04:07:36 ... 04:07:53 that sounds outright wrong 04:08:28 unless it's just a case of epic mislabelling 04:08:34 I'd expect absdepth0 to not change between starting the delay and handling it 04:09:01 as far as I can tell, the old_level parameter to down_stairs is just needed for the portal spell 04:10:07 speaking of down_stairs()... don't even look at up_stairs() or you'll get a seizure due to witnessing the worst case of harmful code duplication ever 04:10:18 I am right now... 04:12:08 KiloByte: are the two functions actually different in important respects? that would be even worse 04:12:52 many things can happen only when going up or when going down 04:14:44 ugh 04:15:34 down_/up_stairs make me *cry*. 04:17:33 I've only read one of the Crawl sourcefiles, and it made me physically ill 04:17:36 I'm not sure I'll dare to read another 04:17:50 Some of them are good. 04:17:52 Some of them are... bad. 04:18:34 from what I've heard, the efforts of the current devteam are making things better on average 04:18:56 Yeah. 04:28:58 by: "delay parameter" means "parameter to the delay function" aka "userdata" 04:29:22 it has nothing to do with the number of turns. which is calculated in an unrelated place! 04:29:54 ais523: don't forget the thirteen years of previous devteams 04:30:28 sorear: yes, I understand that (I think) 04:31:20 the issue is that instead of calling down_stairs(you.absdepth0), _finish_delay calls down_stairs(delay.parm1), and main.cc starts DELAY_DESCENDING_STAIRS with parm1 == you.absdepth0 04:31:53 I can't think of any way that you.absdepth0 would change between starting and finishing the delay 04:32:20 and I'd like to get rid of the old_level argument to down_stairs, which seems to only be used for a portal hack 04:32:33 sorear: no, I know about them, that's the most likely explanation iMO 04:32:42 the current devteam are doing a good job, especially given what they have to work with 04:38:31 can a multi-turn action be interrupted by a user? 04:38:55 or by game events only. 04:39:33 at least some can be interrupted configurably 04:40:32 I'm sure you could write a Lua function that prints a message every turn and interrupt activities on that 04:40:43 not sure if there's a direct way to do it 05:02:15 our makefile is a monstrosity, using zoology instead of feature tests :( 05:02:41 looking at the compiler's version instead of checking if it accepts X is the worst 05:04:39 -!- Mu_ has joined ##crawl-dev 05:04:46 neunon takes it badly when I replace his tables with checks, but fixing clang support would be extra hairy otherwise 05:05:21 but gcc-opt.pl and gcc-opt-x86.pl are just beyond words 05:05:46 not to mention that they do the same thing as -march=native except for being badly incomplete 05:11:10 -!- ortoslon has joined ##crawl-dev 05:29:55 -!- ortoslon has quit [Quit: bye] 05:34:03 -!- eith has joined ##crawl-dev 05:45:09 03by * redf2169b753f 10/crawl-ref/source/stairs.cc: Also erase old level data in down_stairs if appropriate. 05:45:09 03by * re08777be5870 10/crawl-ref/source/files.cc: Don't reset you.prev_targ, you.prev_grd_targ in _save_level. 05:45:09 03by * r4b4c3a1ca466 10/crawl-ref/source/ (delay.cc main.cc): Don't use parm1 for DELAY_DESCENDING_STAIRS. 05:45:09 03by * rbfbd93d7ffc3 10/crawl-ref/source/ (9 files): Replace down_stairs old_level argument by forced destination. 05:45:09 03by * r7e4039d061d9 10/crawl-ref/source/stairs.cc: Don't consider Portal destination stairs known. 05:45:22 03by * r06a73d1d0edd 10/crawl-ref/source/files.cc: Make visitor's _load_level pass an invalid id as the old level. 05:45:33 03by * r3582d859caa6 10/crawl-ref/source/ (files.cc files.h stairs.cc startup.cc wiz-dgn.cc): Make load() take level_id old_level. 05:45:44 03by * r09062a67d643 10/crawl-ref/source/files.cc: Use shorter _make_filename some more. 05:46:01 woohooo 05:46:05 03by * rb60cdfa1fc43 10/crawl-ref/source/files.cc: Update travel cache before saving level. 05:46:10 looks *hoot* 06:11:21 03dolorous * redf888b2941d 10/crawl-ref/source/player.cc: Add spacing fixes. 06:14:53 by: Thanks for fixing tc save 06:14:54 greensnark: You have 1 message. Use !messages to read it. 06:15:01 I totally didn't think of that 06:15:03 !messages 06:15:03 (1/1) by said (3h 55m 51s ago): For some reason, crash playback seems to start at ttyrec beginning. 06:15:12 Hm 06:15:17 !lm * type=crash 06:15:17 11. [2010-06-08] MUMMY the Summoner (L15 VpTm) ASSERT(mimic->inv[MSLOT MISCELLANY] != NON ITEM) in 'mon-stuff.cc' at line 177 failed on turn 87084. (Snake:4) 06:15:35 ooh, more crashes 06:15:49 !lm * type=crash -tv:<0.3 06:15:49 11. MUMMY, XL15 VpTm, T:87084 (milestone) requested for FooTV. 06:16:10 Hm, what do you know 06:16:28 did it work now? 06:16:32 Can't say :) 06:16:43 !lm * type=crash -tv 06:16:43 11. MUMMY, XL15 VpTm, T:87084 (milestone) requested for FooTV. 06:17:06 Well, he tried to describe a mimic 06:17:09 crash crash crashy mccrash 06:17:21 But that's not a good test for whether seeking works correctly in FooTV :) 06:17:24 !lm * type=crash -tv -2 06:17:24 10. MUMMY, XL15 VpTm, T:87272 (milestone) requested for FooTV. 06:17:28 What 06:17:46 He must be very interested in mimics 06:18:27 obsessed? 06:18:58 !lm * type=crash 10 -tv:cancel 06:18:58 10. MUMMY, XL15 VpTm, T:87272 (milestone) cancel requested for FooTV. 06:19:02 !lm * type=crash 10 -tv:<0.3 06:19:02 10. MUMMY, XL15 VpTm, T:87272 (milestone) requested for FooTV. 06:19:44 * greensnark ponders 06:19:50 !lm * type=crash 10 x=time 06:19:50 10. [2010-06-08] [time=2010-06-08] MUMMY the Summoner (L15 VpTm) ASSERT(mimic->inv[MSLOT MISCELLANY] != NON ITEM) in 'mon-stuff.cc' at line 177 failed on turn 87272. (Snake:4) 06:19:54 Bleh 06:21:41 greensnark: I'd like to merge Luca__'s clientserver work (it seems very good), but am unsure about 0.7 -- what do you think? 06:22:17 Hm, it's probably best to wait until 0.7 is branched 06:24:32 -!- Henzell has quit [Remote host closed the connection] 06:25:08 Oh, Gretell is awol? 06:25:20 I should convert Gretell to basicbot so she reconnects 06:25:49 -!- Gretell has joined ##crawl-dev 06:26:05 -!- Henzell has joined ##crawl-dev 06:26:18 !lm * type=crash 10 x=time,rstart 06:26:20 10. [2010-06-08] [time=2010-06-08 [20100508111131S];rstart=2010-06-02 [20100502091145S]] MUMMY the Summoner (L15 VpTm) ASSERT(mimic->inv[MSLOT MISCELLANY] != NON ITEM) in 'mon-stuff.cc' at line 177 failed on turn 87272. (Snake:4) 06:33:39 ?/ mimics causes crash (https://crawl.develz.org/mantis/view.php?id=1705) by OG17 06:33:47 03kilobyte * rfc187f589343 10/crawl-ref/source/ (makefile util/gcc-opt-x86.pl util/gcc-opt.pl): Purge get-opt{,-x86}.pl, -march=native does the same without incomplete zoology. 06:33:48 03kilobyte * r30ac188b995f 10/crawl-ref/source/ (makefile util/gcc-gte.pl): Replace gcc version comparison by feature checks. 06:43:43 Allies detect mimics, attack without revealing them (https://crawl.develz.org/mantis/view.php?id=1706) by OG17 06:54:25 -!- ortoslon has joined ##crawl-dev 07:08:51 Summons' torment is broken (https://crawl.develz.org/mantis/view.php?id=1707) by OG17 07:13:30 by: Looks like FooTV seek was failing for crash milestones because the crash timestamp is after the timestamp of the last ttyrec frame 07:13:48 ah 07:13:54 I've fixed it to just use the last frame if the timestamp is beyond the end 07:14:13 thanks for looking into it! 07:14:29 No problem, it's a good thing to have working 07:14:38 is it ok to leave the crash milestone as is, then? 07:15:15 Yes 07:15:57 !lm * type=crash 10 -tv:<0.3 07:15:57 10. MUMMY, XL15 VpTm, T:87272 (milestone) requested for FooTV. 07:16:58 Huh :P 07:28:34 -!- Zaba has quit [Ping timeout: 264 seconds] 07:30:28 -!- ortoslon has quit [Ping timeout: 260 seconds] 07:32:09 -!- Zaba has joined ##crawl-dev 07:39:19 -!- ixtli_ has quit [Quit: ixtli_] 07:40:32 Ack, the Ugly Thing & Haunt level increases are behind a version bump.. I can't get them for my current game. :P 07:42:08 ^Qyes helps with that 07:42:35 Oh, I see you did :P 07:43:04 Master branch on CDO updated to: 0.7.0-a0-2275-g30ac188 (25.2) 07:43:09 ..needlessly :) 07:44:07 -!- bhaak is now known as demoqorqon 07:44:14 -!- demoqorqon is now known as bhaak 07:45:33 -!- ixtli has joined ##crawl-dev 07:52:59 -!- eith has quit [Ping timeout: 248 seconds] 07:56:03 short len = unmarshallShort(th); 07:56:07 can things go wrong there? 07:56:27 unmarshallShort returns int16_t 07:57:27 short could be not 16 bits 07:59:41 actually, the marshalled short is always 16 bit, and marshalled long is 32 (!), even though you'd expect 64 08:00:04 why would you expect that? 08:00:21 since sizeof(long) == 8 08:00:33 depending on the architecture, of course 08:00:51 elly: "may", I'd say, as it most can be 16-bit :D 08:01:07 +certainly 08:01:20 on all architectures currently supported by crawl, int is 32 bits, but long varies 08:01:57 expect the unexpected! 08:02:29 anyway, I'm wondering if (un)marshallString could have issues that explain https://crawl.develz.org/mantis/view.php?id=1466 08:02:33 at least it's not win64 where QWORD is not four, but one word long :p 08:02:54 the crashes seem to be in unmarshallString 08:03:10 on a lot of architectures it isn't. on most, I'd say. 08:03:35 I was wondering if unmarshallCString could return a negative value, say 08:05:54 Zao: I used clang svn head 08:06:08 released versions probably don't work 08:07:11 also, if using a clang release build, I don't think it's supposed to have huge memory consumption 08:07:26 I think it's supposed to use less memory than gcc 08:07:30 -!- Luca__ has quit [Remote host closed the connection] 08:07:42 I thought he was referring to a memory consumption when building clang itself. 08:07:45 -!- Luca__ has joined ##crawl-dev 08:08:37 unmarshallString looked horrible when I last read it 08:08:48 in particular, it imposes a 1000 char limit for no reason 08:09:02 instead of using ostringstream in the obvious way 08:10:00 btw, how about an overloaded operator << and operator >> instead of marshall* ? 08:10:22 then you can also overload operator & to both read and write like boost does 08:10:35 it would hide the type, which is bad 08:10:40 it's good 08:10:50 it just works then 08:10:57 it can be error-prone for integer types 08:11:13 we want to always specify the type explicitely, otherwise we'll get loads of unobvious errors 08:12:01 for integers you can use a variable-sized implementation 08:12:03 marshall is a mouthful though, it could be just a mem. fun with a type name 08:12:16 so they are serialized independent of the type length 08:12:22 like the one in my repository 08:12:43 only problem is signedness, but if you get that wrong it will immediately and obviously break 08:12:57 since the sign bit is put in bit 0... 08:13:05 where is it in your repo? 08:13:54 http://repo.or.cz/w/crawl/crawl-nettiles.git/commit/702392b41773381b20f3a5bb48dd01f10b2eec6e 08:14:14 oh I misunderstoof 08:14:37 look at http://www.boost.org/doc/libs/1_43_0/libs/serialization/doc/index.html 08:15:00 I thought you mean that you select a conversion function based on type size. 08:15:04 no 08:15:09 yeah I see now. 08:15:19 you send bytes with 7-bit of data and 1 end bit 08:15:44 clearly, the original type size doesn't matter because it works for arbitrarily long integers 08:16:00 signed/unsigned is necessarily different though unless you encode everything as signed 08:16:29 btw, perhaps the crawl approach could also be dropped in favor of Boost.Serialization 08:16:30 htonl and such are probably much more efficient though. 08:16:54 e.g. they don't have loops. 08:16:57 except, what would you need arbitrarily long integers if they get put into fixed-width variables right then? 08:16:59 in what sense? I think variable sized will result in a more compact output 08:17:17 KiloByte: integers are usually small, so the output is smaller 08:17:32 also, you can expand datatypes without breaking compatibility 08:21:01 03by * r0220ef6173ea 10/crawl-ref/source/ (tags.cc tags.h): Add some asserts to unmarshall(C)String. 08:28:13 heh, clang compiles slower than gcc, and produces minimally slower code. -O0: gcc 1m37s, clang 1m44s, -O2: gcc 2m07s, clang 2m09s. 08:28:36 evktalo (L8 HuSu) (D:8) 08:29:05 That was a crash. 08:29:24 http://crawl.develz.org/morgues/trunk/evktalo/crash-evktalo-20100608-132836.txt 08:29:30 KiloByte: with --enable-optimized right? 08:29:42 yes, it reports the assert message if any, and this wasn't an assert 08:29:43 compile times are stable, run times not really since I lack any decent reproducible tests (-seed works differently) 08:29:59 Luca__: of course, current SVN 08:31:24 I can't think of any long tests that would not involve randomness. And with randomness, we'd need a lot more iterations than I did. 08:32:38 actually, I'm not sure why -seed desyncs, we do everything strictly deterministically 08:32:56 by: I see. I took the downstairs and it crashed, probably generating the level. 08:33:09 no, saving the old level from the backtrace 08:33:43 maybe fsim? 08:33:52 Ah. 08:34:03 fsim is random too, right? 08:34:35 it may be far more stable than -arena 'cerebov v test spawner' that I'm using, though 08:34:36 it does a lot of iterations though, and maybe seed works 08:34:48 delay:0 of course 08:35:02 by: Should I try again? 08:35:24 time (for (( i=0; i<100; i++ ));do ./crawl -arena 'cerebov v test spawner delay:0' >/dev/null;done) 08:35:33 gcc 3m47s, clang 3m55s 08:36:02 Keskitalo: can you copy your save to be on the safe side? I fear it's a random crash, though 08:36:41 Okay. 08:37:05 It won't start. 08:38:20 By which I mean it says "starting (hashnumber)", then "hit enter to continue" and then goes back to CDO menu. 08:39:54 http://crawl.develz.org/saves/dumps/evktalo-30ac188-100608-1336.tar.bz2 08:40:11 clang produces heaps of valgrind errors 08:41:24 Luca__: Slightly tangential, but how's Boost with clang++? 08:41:32 Suicidal or kind of working? 08:41:40 Keskitalo: it's quite possible that the save is messed up because it crashed while saving 08:41:49 :/ 08:41:57 http://clang.llvm.org/cxx_status.html 08:42:06 they claim it passes all tests on darwin/x86-64 08:42:49 Not at losing the game, but not having useful save for the bug :) 08:44:39 the errors come from hundreds of places, but they all are either "Conditional jump or move depends on uninitialised value(s)" or "Invalid read of size 8" 08:45:32 the latter is typically when reading something four last bytes of a malloced area and four bytes past the end of it 08:46:20 it's possible clang throws away the rest of the data read, but it still would crash if it were mmapped not malloced 08:46:56 you could file a bug :-) 08:49:00 could std::vector::push_back failing to allocate cause a backtrace like Keskitalo's above? 08:52:43 could _pbuf be null? that's the first thing that comes to mind. 08:54:52 the game did apparently leave a huge core 08:55:52 actually.. it probably wouldn't get to the handler 08:55:52 if _pbuf were null, I wouldn't expect to see the std::vector functions in the trace 08:56:28 yep, it could very well be an allocation failure. 08:57:10 Could be the change to save maps in the level save is causing Crawl to hit ulimit 08:57:37 infinite loop calling writeByte? 08:58:47 greensnark: what change do you mean? 08:59:19 by: I changed level save to also save full map metadata 08:59:26 Which could be largeish 08:59:53 I don't recall what ulimit cdo uses 08:59:56 But it does use a ulimit 09:00:11 the core is among the maximal ones 09:00:19 You have the core? 09:00:24 How big is it? 09:00:48 77M 09:00:55 Hm, sounds pretty likely then 09:01:12 Could be map saving, or could be a leak 09:01:29 it doesn't seem to happen on every save, at least 09:01:44 Let me try generating a bunch of vaults and saving with a ulimit around 100M 09:01:59 -!- Cryp71c has joined ##crawl-dev 09:03:00 I can just list the cores, however 09:03:07 Morning 09:03:07 Cryp71c: You have 1 message. Use !messages to read it. 09:03:09 !messages 09:03:09 (1/1) casmith789 said (4h 43m 27s ago): I think my remembering of shadow imps was from 3s summoning other 5s; that's what's happening now to cause them to appear, anyway. 09:06:43 well well, looks like my lua marker read crashes are also likely to be out-of-memory caused 09:07:01 that would also explain why they aren't repeatable 09:08:12 could we detect the fact that we're failing to allocate in the crash handler? 09:09:12 -!- by has quit [Disconnected by services] 09:09:12 -!- by_ has joined ##crawl-dev 09:09:17 -!- by_ is now known as by 09:10:05 maybe put an exception-catching block aroud everything? 09:10:17 it should catch an std::bad_alloc 09:10:48 you could redefine new. 09:12:24 and whatever allocators are used by libraries, e.g. lua. 09:12:59 that probably won't be enough though if you have something like strdup() somewhere. 09:18:53 Oh, ^F shows the place where the item is now? 09:19:12 yes, there was some discussion on crd 09:19:26 It's neat, but I don't know if it needs to be the default action 09:19:28 on whether to hide that behind an option, or add extra modes, or whatever 09:19:50 it's an extra Enter (probably ok), though the extra redraw might be problematic 09:20:30 I get a reliable crash with ^F, choose item and hit enter to travel 09:20:35 Apparently in my code :) 09:21:33 I decided to just implement it as now because the stash tracker menu code scared me 09:22:05 Yeah, stash menu has a fairly tangled history 09:22:24 Looks like my changes to vault_placement don't survive level excursions very well 09:22:39 Still doesn't explain the OOM though :) 09:24:44 by, there are numerous uses of an in_array function, but its such a common implement and yet is not already present that I cannot help but think it was left out intentionally, do you know anything on those lines? 09:24:55 no 09:25:07 Would you be opposed to me implementing one? 09:25:55 no (no opinion on the matter) 09:26:15 in_array? 09:26:20 Any other developers want to weigh in? 09:26:37 What would in_array do? 09:26:59 greensnark, yeah, just a function which uses a for loop to check for the presence of a particular value in a particular array. 09:27:20 std::find? 09:28:36 I wasn't aware it could be used on arrays. 09:29:19 I thought std::find was only usable on the more object-oriented containers like lists and vectors. 09:30:05 You can use it on arrays 09:30:17 Don't know if it does the right thing on char * though, probably not 09:32:51 does anyone know about the "you can't press tab to select previously chosen character" bug on cdo atm? 09:33:49 don't thinks it's known 09:35:26 heh, the player dies of zero stats after a while in the arena 09:36:23 greensnark, how do you get pointers to .begin() and .end() of the array, then? Declare an iterator manually? 09:36:48 Cryp71c: int x[10]; start would be x and end would be x + 10 09:37:01 ah, that's what I'm doing wrong, thanks. 09:38:23 C++ knows how large an array is even after it's been passed to a function? 09:38:35 I thought it was the same as C, and forgot 09:38:44 ais523: It does not 09:38:45 no, unless it's \0 ended 09:38:50 That's why you have to specify the end 09:38:52 then you can count it 09:38:55 due to decaying the array into a pointer to its first element 09:38:57 ah, ok 09:39:39 Non-nagas can wear naga bardings. (https://crawl.develz.org/mantis/view.php?id=1709) by casmith789 09:39:39 You can't press tab to select previously chosen character (https://crawl.develz.org/mantis/view.php?id=1708) by casmith789 09:39:39 03by * r4109ace5a863 10/crawl-ref/source/ (tags.cc tags.h): Some file output asserts. 09:39:53 03by * r4faff9bab901 10/crawl-ref/source/arena.cc: Give player non-zero stats in arena. 09:40:11 casmith789: Nice bug with bardings <3 09:40:12 casmith789: thanks for filing 09:40:23 np 09:40:48 greensnark: heh, yeah I thought so 09:46:46 -!- violetj_ is now known as violetj 09:46:51 spriggans cant :( 09:47:05 trolls can 09:47:44 get your barding troll started, quick 09:47:59 can you acquire bardings? 09:48:07 not if you aren't a naga 09:48:09 or centaur 09:48:18 casmith789: how did you find that out? 09:48:26 as in, what made you think of trying? 09:48:30 it was white 09:53:39 03greensnark * rc8b90db1f2bc 10/crawl-ref/source/ (files.cc files.h travel.cc): Fix crash when using ^F to travel to an item. 09:54:45 03Cryp71c * re84710bde9e8 10/crawl-ref/source/mutation.cc: Restrict possible demonic guardian summons 09:54:51 Yay, deleting 47 lines to fix a crash :P 09:54:59 hurray 09:55:02 did they actually do anything useful? 09:55:08 Once upon a crime yeah 09:55:19 But later superseded by superior code 10:08:09 arena in valgrind claims that some memory is lost through dgn_register_place 10:09:54 What's the trace? 10:10:42 http://sprunge.us/MiJc 10:10:56 ...hrmm, is there a depth limit on when a Zig can appear? 10:10:59 -!- ortoslon has joined ##crawl-dev 10:10:59 it's very little, but I guess arena just generates one level 10:11:16 maybe it's just because crawl doesn't clean up on exit? 10:11:20 I have one on D:3...its quite surprising. 10:11:38 dat/des/portal/ziggurat.des or something 10:12:16 by: Right, Crawl doesn't toss level_vaults on exit, but the std::vector destructor will forget the pointers, ergo leak 10:16:31 Kyrris (L6 SEFE) (D:4) 10:17:55 what is Gretell reporting? 10:19:27 crashes 10:19:34 !lm kyrris 10:19:38 2668. Kyrris the Firebug (L6 SEFE), worshipper of Sif Muna, on D:4 on -01-, with 0 points after 4723 turns and 0:28:15. 10:19:47 oh 10:19:54 ah 10:20:24 Could well be the ^F crash :P 10:20:28