| AlexDaniel | samcv: great! | 02:56 | |
| dogbert17: here? | 03:14 | ||
| hey what a cool fix I found to the annoying segv in whateverable | 05:30 | ||
| you just run it under valgrind and the segfault goes away! | |||
| awesome! :'( | |||
| yoleaux | AlexDaniel`: RT/GH weekly | 07:17 | |
| Geth | roast/run-with-prove: 175e3af168 | (Nick Logan)++ | 222 files Allow testing from rakudo or roast root While it makes sense to expect the roast to run from the rakudo directory, it is generally accepted that tests should be invoked from inside of the repo to be tested. |
07:36 | |
| roast/run-with-prove: 1b0a473aa9 | (Nick Logan)++ | 2 files Spawn procs in the same location Adds :%proc-args to Test::Util, which passes through its options to shell ala `shell( ..., |%proc-args )`. This allows commands that get spawned to know any relative paths used will point to the expected location (regardless if run from rakudo/ or roast/). |
|||
| roast/run-with-prove: 9bfb1201b3 | (Nick Logan)++ | 23 files Adjust relative lib / file paths to be more correct This removes more assumptions about the directory that the tests have been launched from, allowing them to pass when not cloned into rakudo/t/spec. |
|||
| ufobat_ | .tell Zoffix I've updated the PR github.com/rakudo/rakudo/pull/1249 with your suggestions | 09:21 | |
| yoleaux | ufobat_: I'll pass your message to Zoffix. | ||
| ufobat | what is ::CORE.WHO, or why does this actually not work github.com/rakudo/rakudo/blob/mast...n.pm#L2881 | 10:37 | |
| m: require IO::XXX | |||
| camelia | IO::XXX is a builtin type, not an external module in block <unit> at <tmp> line 1 |
||
| ufobat | in the 2nd iteration, for XXX, $ns{ 'XXX' }:exists seems to be True | 10:38 | |
| m: my $ns := ::CORE.WHO; say $ns<IO>.WHO<XXX>:exists' | 10:50 | ||
| camelia | 5===SORRY!5=== Error while compiling <tmp> Two terms in a row at <tmp>:1 ------> 3 ::CORE.WHO; say $ns<IO>.WHO<XXX>:exists7⏏5' expecting any of: infix infix stopper postfix statement end … |
||
| ufobat | m: my $ns := ::CORE.WHO; say $ns<IO>.WHO<XXX>:exists' | 10:51 | |
| camelia | 5===SORRY!5=== Error while compiling <tmp> Two terms in a row at <tmp>:1 ------> 3 ::CORE.WHO; say $ns<IO>.WHO<XXX>:exists7⏏5' expecting any of: infix infix stopper postfix statement end … |
||
| ufobat | m: my $ns := ::CORE.WHO; say $ns<IO>.WHO<XXX>:exists | ||
| camelia | False | ||
| [Tux] | Due to (very) high system load today, I don't think it is wise to do timing. System load is for the kids and is required | 11:58 | |
| tbrowder | after the release i hope someone can please evaluate my PR #1240 | 13:08 | |
| AlexDaniel | lizmat: alright, your answer makes sense after looking at this: github.com/torvalds/linux/blob/c9b...ys.c#L1731 | 14:17 | |
| although the comment there is still confusing | |||
| so everything is in KiB, alright | 14:18 | ||
| lizmat | aha! :-) | 14:23 | |
| dogbert17 | AlexDaniel: have you by any chance managed to golf down your whateverable SEGV ? | 14:28 | |
| AlexDaniel | dogbert17: it goes away once I reduce the amount of code | ||
| dogbert17: and if I run it under valgrind there's no SEGV | |||
| dogbert17: we had this issue for more than one release so that's not a blocker really as far as I'm concerned, but it would be great to fix it | 14:29 | ||
| because otherwise I can't really use whateverable as a canary | |||
| MasterDuke | AlexDaniel: what about running it under gdb? does it go away there? | ||
| AlexDaniel | oh, hmm | 14:30 | |
| I can try | |||
| dogbert17: it should be related to Proc::Async in some way, I think | |||
| more rambling here: github.com/perl6/whateverable/issu...-345493646 | 14:31 | ||
| dogbert17 | +1 on MasterDukes suggestion | ||
| AlexDaniel | dogbert17: not sure if my gdb-fu is strong enough, wanna try yourself? | 14:39 | |
| dogbert17 | what do I have to install? | 14:41 | |
| AlexDaniel | dogbert17: you'll actually have to do that on the server where the bots are running… give me your ssh pub key and we can move to #whateverable channel to discuss the rest | 14:42 | |
| dogbert17 | haha, here I come up as a NoOb, don't have an ssh pub key :) perhaps better to give the honors to MasterDuke | ||
| AlexDaniel | dogbert17: well, maybe it's a good idea to create one? :) | 14:44 | |
| it's not just for ssh-ing to servers, you can use it to commit to github too | 14:45 | ||
| to push* | |||
| dogbert17 | hmm, perhaps I should look into that | ||
| MasterDuke | i probably only have a couple minutes now, but will have more time this afternoon | ||
| AlexDaniel | dogbert17: I'd recommend ed25519 :) | 14:46 | |
| dogbert17: then, once you have the key generated, the easiest way is probably to add it to your github account, so that others can see it here: github.com/AlexDaniel.keys | 14:47 | ||
| well, here of course: gist.github.com/dogbert17.keys :) | |||
| I'm really hoping that one day we'll get stable enough so that I can declare any rakudo-related issue in whateverable a blocker… :) | 14:49 | ||
| dogbert17 | what about the other error, i.e. the Gumbo one, has that been solved? | 14:51 | |
| AlexDaniel | no | ||
| dogbert17 | I tried to reproduce it a couple of weeks ago but my 32 bit system seems immune to the problem | ||
| AlexDaniel | I'm currently working on a project that does some web scraping, so I simply built it in a way that expects MoarVM to panic every few runs… | 14:52 | |
| dogbert17 | :( | ||
| AlexDaniel | yes, 32-bit systems are immune to that issue | ||
| dogbert17 | and it's not jit related then ? | ||
| AlexDaniel | no, I don't think so | ||
| it is reproducible with MVM_JIT_DISABLE=1 for sure | 14:53 | ||
| dogbert17 | interesting | 14:55 | |
| AlexDaniel | (for anyone else wondering, the discussion is about RT#131003) | 14:56 | |
| synopsebot | RT#131003 [open]: rt.perl.org/Ticket/Display.html?id=131003 [SEVERE][SEGV] Heap corruption when using Gumbo | 14:57 | |
| dogbert17 | ok, I'll bite and setup a 64 bit vm temporarily | ||
| AlexDaniel | Geth: ver github.com/rakudo/rakudo/commit/40...2d348a2000 | 14:58 | |
| hey Geth ? | |||
| dogbert17 | then I can try out Linux Mint 18.3 (beta) at the same time | ||
| AlexDaniel | :) | 14:59 | |
| lizmat | T-Dose shutting down& | 15:49 | |
| dogbert17 | .oO(MoarVM panic: Heap corruption detected: pointer 0x7f8b7470c668 to past fromspace) |
15:51 | |
| AlexDaniel | .tell lizmat gist.github.com/AlexDaniel/bf7d916...604981b70e | 15:56 | |
| yoleaux | AlexDaniel: I'll pass your message to lizmat. | ||
| dogbert17 | AlexDaniel, timotimo, jnthn: gist.github.com/dogbert17/2a4be541...72c321dd65 | 16:13 | |
| AlexDaniel | who knows more about github.com/rakudo/rakudo/issues/1257 ? | 17:03 | |
| timotimo: ? | |||
| timotimo | do we forget to byteswap something somewhere? | 17:08 | |
| did previous versions build fine on the big endian systems? | 17:14 | ||
| hm, looks like | |||
| AlexDaniel | yeah, 2017.06 was ok at least | 17:16 | |
| timotimo | that's a really long span :\ | 18:10 | |
| AlexDaniel | timotimo: so what do you think we can do about it? | 18:34 | |
| if I had a system like that at hand I'd try bisecting, but I don't | |||
| timotimo | should be able to use qemu to run a different architecture on your machine | 18:35 | |
| it's probably dreadfully slow though | 18:36 | ||
| AlexDaniel | it is | 19:29 | |
| DrForr pushes a first cut at Format::Lisp to the ecosystem. | 19:30 | ||
| timotimo | DrForr: into the ecosystem rather than onto cpan? | 19:43 | |
| moritz | we should really update the docs to advise uploading to CPAN | 19:44 | |
| timotimo | we can also put it into the pr text template, right? | 19:48 | |
| moritz | but only after we update the docs, IMHO | ||
| moritz now very glad that he didn't put instructions on how upload modules into his book, but rather linked to the official docs | 19:49 | ||
| timotimo | of course | 19:51 | |
| DrForr | Yes, right now I'm on the ass end of the Intartubes and don't have the patience to both babysit the laptop and straighten out CPAN redentials. | ||
| Geth | roast/run-with-prove: 23846a42c6 | (Nick Logan)++ | 213 files Find local lib in a `FindBin`-esq manner * Use $?FILE to create absolute paths based on the layout of the spectest repo. * Some cases, such as when `run` is used and the cwd can be controlled, still use a relative path (to keep some testing on it) |
20:21 | |
| roast/run-with-prove: 1427f9c76e | (Nick Logan)++ | 24 files Absolutify references to 't/spec' The contents of 't/spec/foo'.IO.slurp depend on the users cwd but can be rewritten with $?FILE to work from any cwd. |
|||
| b2gills | m: my atomicint $a; my $s = supply { whenever Nil { $a⚛++; say 'Promise'; done } }; $s.act: *.say; $s.wait xx 5; say $a | 21:19 | |
| camelia | Promise Promise Promise Promise Promise Promise 6 |
||
| b2gills | It seems a little weird that `.wait` on a `supply` causes `whenever` to be called even after it is done | 21:21 | |
| AlexDaniel | oh gosh… this is probably too slow to be useful | 21:33 | |
| I have troubles starting the system, let alone trying to build rakudo… | 21:34 | ||
| oh… | 21:59 | ||
| jnthn | b2gills: You are tapping the supply twice: once in .act, once again in .wait | 22:01 | |
| Uh, actually 6 times :) | 22:02 | ||
| timotimo | b2gills: did you want to .share? | 22:10 | |
| MasterDuke | AlexDaniel: qemu is usually pretty fast for me, but you're emulating a big-endian architecture, right? | 22:14 | |
| timotimo | yeah, qemu on x86 for x86 is likely blazing fast | ||
| mips on x86, not so much | |||
| AlexDaniel | MasterDuke: actually I just noticed that there's some other problem… trying something right now | ||
| MasterDuke: are you up for some gdb-ing with whateverable? :) | 22:15 | ||
| MasterDuke | might have a couple minutes | 22:20 | |
| AlexDaniel: how do you have things set up? | 22:22 | ||
| AlexDaniel | → #whateverable | 22:26 |