| [Coke] | (bah. the bisect pointed at that commit, but I'm having trouble getting a working version with the moarvm just before that, either.) | 00:33 | |
| TimToady | my cursory glance at perl6-all-modules seems to indicate most parser tests are success oriented, or use boolean context for failure, so that's close to okay | 00:49 | |
| Geth | rakudo/nom: cbfda90615 | (Aleks-Daniel Jakimenko-Aleksejev)++ | docs/ChangeLog Log all changes since 2017.07 Deliberately not logged: 0d7ea0a0 8cc64696 6723006f 4c8b1523 37b14836 d56573a7 de02bc3d 278e2918 d37be249 91be8bc1 8e960522 b8036aff d46f8371 875b0843 356dc404 e0d108bf 64064187 1d06770b 15debeec dc4ed746 50d38a1f 5a58498c 59e55646 99375e4a 48ce47e6 f083cfc6 a8016278 7fdbb49c f1464e22 c63c57e9 67455170 6e001867 59ba9c02 3c797b05 a256c26d 7599e0c3 c43ba0d0 d9f51da8 412c576b abf1cfe7 bf7063d1 c3da74d8 da4a0f50 29e41b45 c68762f4 6c76ed0a ce0b7b31 d151d8ad |
00:51 | |
| roast/6.c-errata: b0044b0751 | TimToady++ | 2 files fix tests that depend on failed parse --> Nil |
|||
| AlexDaniel | releasable6: status | 00:57 | |
| releasable6 | AlexDaniel, Next release in 3 days and ?18 hours. 2 blockers. 164 out of 165 commits logged | ||
| AlexDaniel, Details: gist.github.com/cd7f696d7be986c6f6...0adad0299b | |||
| AlexDaniel | releasable6: how can I log (or ignore) something if I don't know its sha yet? You stupid. | ||
| releasable6 | AlexDaniel, I cannot recognize this command. See wiki for some examples: github.com/perl6/whateverable/wiki/Releasable | ||
| timotimo | "added change of changelog to changelog" | 00:59 | |
| "added change of changing the changelog to changelog" | |||
| AlexDaniel | timotimo: but then you'll have to, yes | ||
| TimToady | .oO(What is the name of the changelog called?) |
01:00 | |
| AlexDaniel | This is how robot uprising is going to look like. They will torture us asking to include a sha in the commit message of the same damn commit | ||
| timotimo | we have to mine changelogcoin | 01:01 | |
| AlexDaniel | wait, but it's only 8 characters | ||
| which is… kinda a lot, but what's the estimate? | 01:02 | ||
| timotimo | should be doable with a little time investment | ||
| Zoffix | (+ Internal: + Introduced Mixy.total-positive [f49c49bb].... that's not a very internal-looking method name; wrong section?) | 01:03 | |
| Dunno, is this really worth putting in the changelog? ¯\_(?)_/¯ "+ Removed a 3-year old hack [e86bdfca]" | 01:04 | ||
| AlexDaniel | Zoffix: yea, I've been thinking about this too. It's not in roast yet, as far as I know, so decided to put it there | ||
| Zoffix | "Internal" section was originally meant as "large internal changes that may indirectly impact users". | ||
| AlexDaniel | oh | ||
| Zoffix | Just saying. I'm not here to tell you how to run the show :) | ||
| Zoffix re-disappears | |||
| AlexDaniel | feedback is welcome, thanks :) | 01:05 | |
| Geth | roast: b53616f8e6 | TimToady++ | 2 files fix tests that depend on failed parse --> Nil |
01:06 | |
| roast: 1fb68c4b7a | TimToady++ | S05-grammar/example.t generalize overly specific parsing test We shouldn't be testing to see that a parse specifically returns Nil, since in the future it will be returning Failure. In any case, that's not what this test is trying to test, so it shouldn't. |
|||
| rakudo/nom: 9501edae4f | TimToady++ | src/core/Grammar.pm return Failure from failed P6-level .parse The .parse and .parsefile methods will now return a Failure object that points to the high-water state of the parser, if available. (The .subparse method does not, since it is presumed to be for lower-level piecemeal matching that is likelier to be success-oriented and have its own failure recovery modes.) |
01:13 | ||
| TimToady | okay, we'll see if that flies in the ecosystem | 01:14 | |
| if not, it's an easy retraction | |||
| my inspection of perl6-all-modules did not turn up any danger signs, but we could yet be blindsided | |||
| otoh, we're getting an amount of bad press about not showing a default error message on grammar failure, so it might be worth a bit of heartburn | 01:15 | ||
| AlexDaniel | this looks very promising | 01:35 | |
| but oh Larry, our brave man. Three days before the release :) | 01:36 | ||
| TimToady | sometimes it's difficult to tell the brave from the foolhardy, except in retrospect :) | 01:37 | |
| TimToady just hopes perl6-all-modules is representative of most usage | |||
| otoh, it can be argued that people who don't test the return value of .parse deserve what they get :) | 01:38 | ||
| and in any case, the fixes should be relative trivial, if any are needed | 01:39 | ||
| *tively | |||
| samcv | TimToady, why would you show an error on false parses though? | 01:40 | |
| i mean we don't do it for regex | |||
| i mean i guess it could be optional? | 01:41 | ||
| not sure why they would expect it though. there's LOADS of times a failed parse is expected | |||
| AlexDaniel | it gives you a Failure, you don't have to sink it ;) | ||
| TimToady | and you can always use .subparse if you want finer control | 01:42 | |
| AlexDaniel | m: grammar G { rule TOP { <a> <a> }; rule a { <b> }; rule b { <c> }; rule c { $<d>=x } }; my $x = G.parse('x z'); say WHAT $x | ||
| camelia | (Failure)? | 01:50 | |
| AlexDaniel | m: grammar G { rule TOP { <a> <a> }; rule a { <b> }; rule b { <c> }; rule c { $<d>=x } }; G.parse('x z') | ||
| camelia | 5===SORRY!5=== Error while compiling <anon>?Confused?at <anon>:1?------> 3<BOL>7?5x z?Actually thrown at:? in block <unit> at <tmp> line 1?? | ||
| awwaiid | TimToady: I just checked a random script of mine which does .parse, and it indeed blindly tries to use the result -- I believe in a way that is no worse off with your change, and will be significantly proved if it takes your change into account (with only a single line addition). As another datapoint for you :) | ||
| AlexDaniel | c: HEAD~10 grammar G { rule TOP { <a> <a> }; rule a { <b> }; rule b { <c> }; rule c { $<d>=x } }; my $x = G.parse('x z'); say WHAT $x | ||
| committable6 | AlexDaniel, ¦HEAD~10: «(Any)» | ||
| awwaiid | significantly improved | ||
| TimToady | almost all the uses of .parse I saw were success oriented in that way :) | ||
| I guess ilbot missed that bit of conversation :) | 01:53 | ||
| travis-ci | Rakudo build failed. TimToady 'return Failure from failed P6-level .parse | 02:13 | |
| travis-ci.org/rakudo/rakudo/builds/264970177 github.com/rakudo/rakudo/compare/c...01edae4f73 | |||
| buggable | [travis build above] ? All failures are due to timeout (0), missing build log (0), GitHub connectivity (1), or failed make test (0). | ||
| samcv | yay more tests | 02:21 | |
| Geth | roast: 649d9dc503 | (Samantha McVey)++ | S03-operators/comparison.t Add tests to make sure synthetics compare properly These tests are to handle the fixes done here: "Fix MVM_string_compare to support deterministic comparing of synthetics" github.com/MoarVM/MoarVM/commit/ab...c0dcb693cd |
||
| roast: 8695092352 | (Samantha McVey)++ | S03-operators/comparison.t Change all tests in comparison.t to use is-deeply Also remove some trailing whitespace. |
|||
| BenGoldberg | m: sub foo { COMPILING::<bar> = my sub { 42 } }; package Baz { BEGIN foo(); bar() }; # What am I doing wrong? | 02:25 | |
| camelia | 5===SORRY!5=== Error while compiling <tmp>?Undeclared routine:? bar used at line 1. Did you mean 'VAR', 'bag'??? | ||
| samcv | all the 3 i added fail before i added that change. declared the strings in the order that would cause them to fail | 02:27 | |
| Geth | roast: 51c4ac08bd | (Samantha McVey)++ | S03-operators/comparison.t Add a large number of string comparison tests * Test comparing empty strings with cmp & leg = Add a function to test various cases: * Test same string but different in length ** Test reversed term returns opposite result * Same string of various lengths but last grapheme is different * Same length string but different character is in various locations ** Same as previous test, but the different character is a lower codepoint than previous test |
03:42 | |
| AlexDaniel | samcv++ | 03:51 | |
| samcv | i wonder how much work adding a string search algorithm would be | 03:56 | |
| since now ignoremark/ignorecase+ignoremark/plain index operations all go through the same function | 03:57 | ||
| well aside from the index when both are flat strings of the same type that use memmem | |||
| adding it before would have required either doing it for just plain search or doing what has now been done AND adding the algorithm. but now that it's all in one function it makes adding the algorithm work for all the functions which is pretty nice | 03:58 | ||
| lizmat | Files=1217, Tests=67435, 267 wallclock secs (13.61 usr 5.08 sys + 1795.67 cusr 176.89 csys = 1991.25 CPU) | 08:43 | |
| [Coke] | has someone written a script to do a bisect across rakudo/nqp/moarvm? | 12:46 | |
| brrt | don't think we have. also, arguably not easily bisectable, since the space is threedimensional | 12:52 | |
| yoleaux | 11:49Z <markmont> brrt: irclog.perlgeek.de/moarvm/2017-08-16#i_15024158 | ||
| nine | Zoffix: would you be ok with IRC::Client not just catching all exceptions and turning them into warnings? That's the reason why Geth keeps disappearing. Connection gets reset for some reason and the warning logged, but since Geth doesn't exit, systemd will not restart it. | 13:36 | |
| AlexDaniel | brrt: doesn't necessarily have to be. If I got it right, you can swap moar in some cases without any issues | ||
| brrt: and even if not, you can compile on demand. There are 12 steps IIRC, so that's 12×3 minutes approximately | 13:38 | ||
| brrt: but you can speed it up if you compile two rakudos ahead of time for the next step | |||
| that should halve the time, if I got it right | 13:39 | ||
| there's github.com/perl6/whateverable/issues/40 and maybe github.com/perl6/whateverable/issues/195 | 13:40 | ||
| whateverable right now has moar and rakudo built on every commit since 2014.01 | 13:41 | ||
| and it takes just a little bit of space actually | |||
| so even if it is threedimensional as you say, you can still squish it into a reasonable amount of space | |||
| oh also, 12 steps a wrong number. Should be less because you only have a couple of commits between bumps | 13:46 | ||
| Zoffix | nine: what's the warning? | ||
| AlexDaniel | that's considering the fact that rakudo bisect is a solved problem | ||
| Zoffix | I rather IRC::Client reconnect instead of crashing and relying on some support script to restart it | 13:47 | |
| nine | Zoffix: that would be absolutely fine, too, of course :) | 13:54 | |
| Zoffix: warning could be "Jul 29 23:15:52 camelia perl6[32400]: ??? _ Connection closed" | |||
| Zoffix: also "Aug 13 10:35:56 camelia perl6[27970]: [ERROR] [27970] [11] connection reset by peer in method send-response at /home/geth/.perl6/sources/41D120CC8C5AC1ADBAA2CC7E606E4F8710D84044 (HTTP::Server::Tiny) line 350" | |||
| Or this one: Aug 11 13:55:29 beta perl6[25007]: connection reset by peer | 13:55 | ||
| Zoffix | .in 3d fix IRC::Client/Geth: irclog.perlgeek.de/perl6-dev/2017-...i_15024568 | 14:00 | |
| yoleaux | Zoffix: I'll remind you on 19 Aug 2017 14:00Z | ||
| Zoffix | K, I'll try to fix this on the weekend | ||
| nine | ++Zoffix | ||
| [Coke] | AlexDaniel: can't use whatever for this because I'm specifically chasing something on OS X. | 14:12 | |
| (though maybe I can if I install it locally, I dunno) | 14:13 | ||
| AlexDaniel | [Coke]: I know, right. I'm planning to make bots easily runnable locally, right now they are not. | ||
| Geth | rakudo/nom: 63cdc426ad | (Zoffix Znet)++ (committed using GitHub Web editor) | src/core/Rakudo/Iterator.pm Fix two typos to unbreak JVM build Per - github.com/rakudo/rakudo/commit/3a...t-23677553 - github.com/rakudo/rakudo/commit/50...t-23677589 |
15:22 | |
| lizmat | Zoffix++ # catching JVM build errors /me let through | 15:27 | |
| Skarsnik | Hello, Dunno if it's interesting but some tests for NC fail on my chromebook (armv71) | 15:29 | |
| lizmat | hmmm I had given up on building JVM, but it appears it is now busted for me completely: | 15:32 | |
| gist.github.com/lizmat/ca7311a55f0...013a53e715 | |||
| lizmat tries a fresh clone | 15:34 | ||
| timotimo | lizmat: the method it refers to, String.codePoints, seems to be new in Java8? | ||
| lizmat | same result :-( | ||
| $ javac -version | 15:35 | ||
| javac 1.7.0_45 | |||
| timotimo | is java8 just java 1.8? or is it more silly than that? | 15:36 | |
| Skarsnik | gist.github.com/Skarsnik/8df31328a...93664d644f | ||
| timotimo | Skarsnik: little endian vs big endian? | 15:37 | |
| nine | timotimo: it's just that | ||
| timotimo | ok, so if we use String.codePoints we definitely should somehow explicitly require Java8 or newer | ||
| lizmat | timotimo: I have *no* idea | ||
| timotimo | so it bails out with a much more helpful error message | ||
| Skarsnik | hm | 15:39 | |
| can someone fix the test message in github.com/rakudo/rakudo/blob/nom/...args.t#L62 ? (replace uint8 by uint16 and uint32). I could do it and do a PR but I have to create a branch on my fork and sync it with upstream probably | 15:41 | ||
| timotimo | you can make changes to files directly on github, it'll create a correct branch for you if i'm not mistaken | 15:42 | |
| lizmat | hmmm... so I upgraded to 8.0.144 or so, now it still claims to use 1.7.045 :-( | ||
| Skarsnik | directly on rakudo repository or my fork? | 15:43 | |
| lizmat | Skarsnik: will do | ||
| Geth | rakudo/nom: 591f69af1e | (Elizabeth Mattijsen)++ | t/04-nativecall/02-simple-args.t Fix copy-pasto, Skarsnik++ |
15:45 | |
| lizmat gives up on JVM | 16:02 | ||
| Skarsnik | hm | 16:06 | |
| I added a printf in github.com/MoarVM/MoarVM/blob/mast...all.c#L239 to have the value of the ushort before the cast. I end with -15xxxxx instead of just 65518 | 16:08 | ||
| jnthn | TimToady: So I'm adding various atomic ops to enable doing various lock-free algos (as well as atomic increment/decrement). For now cas is a function, but while putting them in I just did github.com/rakudo/rakudo/blob/atom.../Atomic.pm | 16:10 | |
| TimToady: That is, threw them in a class so you write Atomic.assign($head, $node) or so | |||
| But my feeling is they want to be functions really | 16:11 | ||
| There's nothing method-y about them | |||
| I'm having a bit of a hard job naming them though, in that postfix:<++> is now we spell post-increment, and = is how we spell assignment, so atomic-assign and atomic-int feel a tad odd. | 16:12 | ||
| TimToady | well, that's why the setting is an outer scope, so it won't override a user's sub atomic ($particle) { "or so" } | ||
| jnthn | Any thoughts? :) | ||
| Sure, I'm not worried about users with existing functions | |||
| I'm worried about what to even call them in the first place :) | 16:13 | ||
| TimToady | but yeah, if you have an additional name there, I gotta think a little | ||
| Skarsnik | *65534 | ||
| jnthn | Yeah, my names so far are in the file I linked. I mean, we could just call them atomic-fetch, atomic-inc, atomic-add, etc. | ||
| TimToady | well, that's pretty clear, if a bit clunky | 16:14 | |
| jnthn | Yeah, feels it to me too | ||
| TimToady | I guess what bothers me is that you'd like to name them as variants of built-in operators where possible | ||
| jnthn | Right | ||
| TimToady | but maybe that's not doable | 16:15 | |
| jnthn | The inc/dec have the semantics of the postfix, not prefix, fwiw. | ||
| TimToady | atomic:<++> leaves out the category in a C++y way... | ||
| will we have both prefix and infix ones? | 16:16 | ||
| jnthn | Yeah | ||
| TimToady | atomic-infix:<+> is even longer, but how often are you going to call them anyway? | ||
| infix:<+>:atomic | 16:17 | ||
| jnthn | The operations (these are tied to what hardware typically provides) are ++, --, +=, =, .<> (that is, decont) | ||
| Note that it's += not just + | |||
| That is, the operation (at a low level) takes the address of an integer and another integer to add | 16:18 | ||
| TimToady | well, atomic-infix:<+=> would presumably just be a name, nothing special on the category | ||
| jnthn | Yeah | ||
| TimToady is carefully not suggesting capitalizing them :) | 16:19 | ||
| Skarsnik | :nuclearemojiunicode:+= | ||
| perlpilot | atomic::infix:<+> :-) | ||
| TimToady | INFIX:<+> | ||
| jnthn | Skarsnik: Haha, I already thought about looking for atomic-y Unicode chars and doing a meta-op :P | ||
| But...well no :P | |||
| Huffman coding wise, these will be rarely used | 16:20 | ||
| TimToady | $a A+= $b | ||
| but that don't work so well for prefix/postfix | |||
| jnthn | Yeah, but then you'd have to write $foo\A++ | ||
| lizmat | +++ ? | ||
| Skarsnik | hm, should I run the spec test on my chromebook? | 16:21 | |
| jnthn | Skarsnik: What's the worst that could happen? :) | ||
| Skarsnik | it explode ! | ||
| perlpilot | make sure you get video if that happens. | ||
| Perl 6 -- so advanced, it makes chromebooks explode. | 16:22 | ||
| Skarsnik | having NC test that fails bother me x) | ||
| TimToady | $foo? ++ | 16:23 | |
| jnthn | *groan* :) | ||
| Skarsnik | huhu | ||
| Zoffix | ? looks really nice :) | 16:24 | |
| nuka-infix:<+> | |||
| TimToady | $foo++? might be perceived more as a modifier | 16:26 | |
| Geth | rakudo/nom: 435f7e36c0 | (Elizabeth Mattijsen)++ | t/04-nativecall/02-simple-args.t Oops, Skarsnik++ |
||
| lizmat | .u ? | 16:27 | |
| yoleaux | U+2622 RADIOACTIVE SIGN [So] (?) | ||
| jnthn | True | ||
| How'd we Texas it? :) | |||
| lizmat | hmmm.. isn't this all about scalars ? | ||
| TimToady | why would we want to nuke Texas? | ||
| lizmat | perhaps use ? as a sigil ? | ||
| b2gills | postfix:<?++> | ||
| Zoffix | u: ? | 16:28 | |
| unicodable6 | Zoffix, U+269B ATOM SYMBOL [So] (?) | ||
| TimToady | if it's rare enough, I'd say the obscurity of unicode is a feature :) | ||
| Zoffix | lol :D | ||
| TimToady | I like ? | ||
| Skarsnik | I don't see it x) | ||
| TimToady | atomic orbitals and a nucleus | 16:29 | |
| though I don't think any atom has 3 symmetrical orbitals like that :) | |||
| Skarsnik | are some spec tests supposed to fail? | ||
| TimToady | not generally | ||
| Skarsnik | I have S17-supply-syntax that is marked dubious | ||
| TimToady | we do still have occasional glitches, so try rerunning the test by hand | 16:30 | |
| b2gills | The orbits are actually more like electron clouds of possible position in real atoms | ||
| TimToady | anything in S17 tends to have occasional races, though thankfully they are getting rarer | ||
| I had that one fail in my last run with TEST_JOBS=6, but running it alone works okay | 16:31 | ||
| jnthn | Yeah, I see that one occasionally get unhappy also | ||
| Didn't manage to figure it out yet. | 16:32 | ||
| lizmat | FWIW, I think the race conditions are now mostly in the tests :-) | ||
| jnthn | In general I like the idea of something that looks like the operators. I guess we could have a prefix of it on its own as the atomic fetch also | 16:34 | |
| I figure full-memory-barrier() can just be a function called that :P | |||
| I'd really like an ASCII form though. | |||
| lizmat | jnthn: but if you do such an op on a variable with a REPR that doesn't support it, wouldn't that crash and burn ? | 16:35 | |
| jnthn | lizmat: The set of things you can do this one is VERY constrained | ||
| You can do it on a native int (which may be held in a lexical, attribute, or native array) for the integer operations | 16:36 | ||
| Or a Scalar container for the reference ones | |||
| These are really quite low level | |||
| lizmat is surprised a Scalar container works at all | |||
| or is there some magic involving a native int involved there ? | |||
| jnthn | No, it's because you can CAS a pointer | ||
| lizmat | ah, ok | 16:37 | |
| jnthn | That's how most lock free data structures are done | ||
| (Yes, I'll do docs once we get the naming straightened out) | |||
| There are places in CORE.setting we'll be able to use these. Like the increment/decrement of loads in ThreadPoolScheduler | 16:38 | ||
| Zoffix | What's CAS stand for anyway? | 16:39 | |
| jnthn | (atomic) compare and swap | ||
| Zoffix | Ah, thanks. | ||
| Skarsnik | hm rakudo github page show that travis an appvoyer build failed. it that normal? | ||
| lizmat | jnthn: and I know one place in Damians Quantum:: modules that could use it as well | 16:40 | |
| Zoffix | Skarsnik: not really. Though since it's not reported by a bot, sometimes the failures on it can go unnoticed for a bit. | ||
| TimToady | for Texas we could just fall back on an :atomic operator modifier, I suppose | ||
| $a + $b :atomic | |||
| jnthn | Well, except only += is available as an atomic op, not + | ||
| TimToady | $a +- $b :atomic | ||
| jnthn | The point of these are they do an atomic update of a storage location | 16:41 | |
| Skarsnik | probably need new op then? | ||
| TimToady | $a += $b :atomic | ||
| lizmat | (atom++) | ||
| (atom+=) | |||
| etc ? | |||
| jnthn | Or what lizmat suggested | ||
| Which is consistent with our other texas forms I guess | 16:42 | ||
| Thoudh I'd atomic rather than atom :) | |||
| lizmat | agree :-) | ||
| Skarsnik | does other language like c++ has special op or something for that? | 16:43 | |
| jnthn | Skarsnik: Probably they just do it as functions | ||
| Which our operators are anyway | |||
| I think the main attraction to me of doing it with operators is | 16:44 | ||
| atomic-inc($a) # OK, is it returning the original or the new? | |||
| Whereas | |||
| $a(atomic++) doesn't leave you guessing | |||
| I only worry if there'd be ambiguity there with call syntax... :S | 16:45 | ||
| lizmat | ah, but wouldn't that be interpreted as a call of a Callable in $a ? | ||
| Skarsnik | std::atomic<long> value(0); | ||
| value++; //This is an atomic op | |||
| value += 5; //And so is this | |||
| jnthn | Depends if LTM wins... | ||
| Skarsnik | this is so ugly >< | ||
| jnthn | But yeah, that's a downside of the parens approach | ||
| We're back to "not enough ASCII" again :P | 16:46 | ||
| TimToady | I know, we could use special Unicode brackets instead :P | ||
| TimToady again profers the fact that $a += $b :atomic is already available | 16:47 | ||
| lizmat | m: sub postfix:<(atomic++)> (\a) { a++ }; my $a = 42; dd $a(atomic++); dd $a # actually, it works :-) | ||
| camelia | 42?Int $a = 43? | ||
| TimToady | but it's misleading to the eye | ||
| druther just have to write $a\A++ if it comes to that | 16:49 | ||
| at least it looks different from anything we have | |||
| jnthn | What could we do with regards to a prefix? | 16:50 | |
| TimToady | LTM wins there too | ||
| jnthn | True | ||
| TimToady | and $a A+= $b already looks metaoppy | ||
| ugexe | too bad you cant do something like Atomic[+=] or some such, like with R[+=] but with a postfix | 16:51 | |
| TimToady | well, we can, it'd just be a special case | ||
| jnthn | I kinda thought ?$a would be atomic load, but A$a is probably gonna be rathr problematic ;) | ||
| *rather | |||
| TimToady | well, that might need a bit of special casing, but it could work as a syntax, at least until someone defines class A | 16:53 | |
| b2gills | Isn't there already a problem if they defined class X | 16:54 | |
| TimToady | well, there used to be more of a problem | ||
| but the internals look up things in SETTING or so these days | |||
| jnthn | That was a namespace clash rather than a syntax clash :) | 16:55 | |
| b2gills | We could make it so you can't create a class with a single letter without doing ::(A) | ||
| ugexe | is there special handling of X:: ? if so, where? (this could be related to a weird bug I was hunting down) | ||
| TimToady | obviously we need a load prefix that is a no-op ordinarily :) | ||
| jnthn | Well, actually .<> is our spelling of decont | ||
| So a postfix A<> would be an option too | 16:56 | ||
| TimToady | A<> $x | ||
| but I still think there's a case to be made for leaving out the Texas version entirely, and using foo() notation in that case instead | 16:57 | ||
| Geth | rakudo/atomics: 46b1a7f277 | (Jonathan Worthington)++ | src/core/Atomic.pm Make full-barrier() just a sub. |
||
| jnthn | TimToady: Yeah, I'd also be good with that | ||
| sub atomic:<++>(int $x is rw) { $x++ }; my int $x = 42; atomic:<++>($x); say $x | 16:59 | ||
| m: sub atomic:<++>(int $x is rw) { $x++ }; my int $x = 42; atomic:<++>($x); say $x | |||
| camelia | 5===SORRY!5=== Error while compiling <tmp>?Cannot add tokens of category 'atomic'?at <tmp>:1?------> 3sub atomic:<++>7?5(int $x is rw) { $x++ }; my int $x = 42;? | ||
| jnthn | If we want that we'll have to allow it I guess | 17:00 | |
| Geth | roast: bd54af79c2 | (Jonathan Worthington)++ | S17-lowlevel/atomic.t Follow function -> method change. |
17:01 | |
| jnthn | Well, guess I should head home and have dinner :) | 17:06 | |
| travis-ci | Rakudo build passed. Zoffix Znet 'Fix two typos to unbreak JVM build | ||
| travis-ci.org/rakudo/rakudo/builds/265201947 github.com/rakudo/rakudo/compare/9...cdc426adbc | |||
| Skarsnik | make spectest take forever x) | 17:21 | |
| TimToady | we mostly run it in parallel using TEST_JOBS=4 make spectest or whatever | 17:27 | |
| (I use 6) | |||
| ((cuz I have 4 cores)) | |||
| Skarsnik | gist.github.com/Skarsnik/7f9836d34...72e41fc50b (the grep is here because I can't scroll while selecting stuff in the console) | 17:45 | |
| nine | m: use NativeCall; sub foo() { my sub memcpy(Pointer $d, Pointer $s, size_t $size) returns Pointer is native(Str) { * }; memcpy(Pointer, Pointer, 0); }; foo; foo; | 17:46 | |
| camelia | Type check failed for return value; expected NativeCall::Types::Pointer but got Whatever (*)? in sub memcpy at <tmp> line 1? in sub foo at <tmp> line 1? in block <unit> at <tmp> line 1?? | ||
| nine | ^^^ golf of RT #131898 | 17:47 | |
| synopsebot6 | Link: rt.perl.org/rt3/Public/Bug/Display...?id=131898 | ||
| nine | It's....really odd. When one moves the memcpy sub outside the foo sub it works | ||
| timotimo | and this happens at run-time? :\ | 17:49 | |
| Skarsnik | how he event found memcpy after the sub declaration? | ||
| nine | m: use NativeCall; use nqp; sub foo() { my sub memcpy(Pointer $d, Pointer $s, size_t $size) returns Pointer is native(Str) { * }; say nqp::objectid(&memcpy); memcpy(Pointer, Pointer, 0); }; foo; foo; | ||
| camelia | 22278832?22279328?Type check failed for return value; expected NativeCall::Types::Pointer but got Whatever (*)? in sub memcpy at <tmp> line 1? in sub foo at <tmp> line 1? in block <unit> at <tmp> line 1?? | ||
| Skarsnik | *even | ||
| nine | It's a different &memcpy object on the second run. But why then doesn't the still existing &CALL-ME method get called? | ||
| The same sub that creates the new subroutine body changes the invokespec so the body gets called instead of &CALL-ME. So how can the latter part be in effect but the new subroutine body be missing? | 17:50 | ||
| TimToady | Skarsnik: that looks like you're out of sync with nqp/moarvm, so maybe try . config.status and then remake | 17:51 | |
| nine | Or asked the other way round: how can the new &memcpy object have a changed invokespec? It has never been called and thus neither it's !create-optimized-call. | 17:53 | |
| timotimo | closure clones? | 17:54 | |
| nine | That would of course explain the different objectids | 17:58 | |
| But somehow nqp::bindattr seems to affect the clone while nqp::setinvokespec affects the original which then propagates into the clones. | 17:59 | ||
| Because of course the invocation_spec is part of the STable which is shared by the clone. | 18:04 | ||
| So now I feel like I understand the problem. But a solution has yet to reveal itself. | 18:06 | ||
| Skarsnik | config.status? | 18:10 | |
| it's weird, it's like an install from 2 weeks | 18:12 | ||
| geekosaur finds "config.status --recheck" a handy way to resync | |||
| (no ".""souce" needed, in fact it's likely a bad idea as you don't want to pollute your interactive shell with all its temporary vars) | |||
| Skarsnik | it's a git command? | 18:13 | |
| geekosaur | no | ||
| it's generated by configure, and is a quick way to regenerate configure results | 18:14 | ||
| with --recheck it basically reruns configure with everything it had the first time, without your having to know/remember how it was run before, so it will pick up changes to dependencies | |||
| without --recheck it just regenerates everything configure generated the first time, in case you lost/trashed them somehow | 18:15 | ||
| Skarsnik | unkwon option, hm | 18:16 | |
| too bad I can't install synergy on this :( | |||
| timotimo | well, we don't have a traditional configure script like autotools would create | ||
| Skarsnik | I mean it display the help of the config.pl | ||
| and says there is no --recheck option | 18:17 | ||
| geekosaur | oh, then you might well have to just rerun it. I still suspect the "." is wrong unless someone practiced very careful script hygiene | ||
| travis-ci | Rakudo build passed. Elizabeth Mattijsen 'Fix copy-pasto, Skarsnik++' | ||
| travis-ci.org/rakudo/rakudo/builds/265211901 github.com/rakudo/rakudo/compare/6...1f69af1e05 | |||
| Skarsnik | I ran sh config.status --recheck | ||
| timotimo | geekosaur: what do you mean your terminals aren't throw-away? :) | 18:22 | |
| samcv | nine, you regressed my change I made here: github.com/rakudo/rakudo/commit/96e6b338c with your commit github.com/rakudo/rakudo/commit/ec7bc25c4 | 18:29 | |
| Geth | rakudo/nom: 9a0afcbcec | (Stefan Seifert)++ | 2 files RT 131898: Fix native closures failing on the second run On the first call of a native sub we create a new subroutine body and replace the original (stored in the routine's $!do attribute). We also set the invocation spec so the body will be called instead of the CALL-ME method. This invocation spec is shared between an original routine and it's clones. If the routine is a clousre, the body won't be shared as only a clone of the routine will be actually visible in the lexical scope. So until someone comes up with a better idea, we won't create an optimized version for a closure anymore. |
18:32 | |
| nine | On a scale from 1 to OMG, how ugly is this workaround? ^^^ | ||
| Skarsnik | I say blue | 18:34 | |
| nine | samcv: oh, sorry about that. I do need to communicate the nesting level to the precompilation processes and RAKUDO_MODULE_DEBUG seemed like a good vehicle for that. | ||
| Zoffix | ninety :) | 18:36 | |
| samcv | so how did you change the functionality of the variable with that commit? | ||
| Zoffix | Also, IIRC callsame is much slower than just doing the actual call; which one is it calling self.Mu::clone? Perhaps it's faster to rewrite it to that | ||
| samcv | i believe my fix made it so if it was 0 or empty it was false and if it was some other nonempty string then it was true | 18:37 | |
| nine | samcv: well I assume I start out with 1 and can increment it for every level of nested precompilation so I know how far to indent the debug output. | ||
| samcv | if you set it to a nonnumber then rakudo won't compile | ||
| i would think that should make it assume a number of '1' instead of breaking everything | 18:38 | ||
| nine | samcv: to be honest that feels very much like a DIHWIDT | ||
| samcv | DIHWIDT? | ||
| "Doctor, It Hurts When I Do This." | |||
| hmm uhm | |||
| true | 18:39 | ||
| i mean i'm not totally against changing it as long as we get something out of it | |||
| but you should change the comment at the top of the file if we're changing what we accept | |||
| i'm not sure why your change causes that. at least the CRON job started failing immediately after that commit so you could have made another commit that did it? but that's the only commit since mine to the file i did which fixed it back to how it used to be many months ago | 18:41 | ||
| hmm ok it seems maybe in src/Perl6/ModuleLoader.nqp things may have stayed the same but in src/core/Process.pm it got changed to only accept numbers? | 18:42 | ||
| not sure what the results of your change do, to be honest | 18:43 | ||
| nine | I guess I should use $DEBUG instead of $rakudo-module-debug? Will it retain positive numbers other than 1? | ||
| Haha, I'm not sure what the nqp::radix business does either :) | |||
| samcv | i'm not against using rakudo-modelu-debug | ||
| well i think i wrote this code: github.com/rakudo/rakudo/commit/ec...3bdf2f2aL3 | |||
| but it does what the comment shows above it | 18:44 | ||
| nine | Meh...we'd still have to duplicate the same here: github.com/rakudo/rakudo/commit/ec...21d722ccR4 | 18:45 | |
| And at that point it really feels like just not worth the hassle. | 18:46 | ||
| If it's fine by you, I'll just remove the now misleading comment | |||
| samcv | yeah it's fine if you change it | ||
| just change the comment | |||
| and then we should be good | 18:47 | ||
| Geth | rakudo/nom: 3e70d44d47 | (Stefan Seifert)++ | src/Perl6/ModuleLoader.nqp Remove now misleading comment The RAKUDO_MODULE_DEBUG environment variable is now used for communicating the nesting level to precompilation processes, so it really should always be numeric. Would have been more user friendly to accept any truthy value but the implementation hassle just seems not worth it any longer. |
||
| samcv | cool thanks :-) | 18:49 | |
| NeuralAnomaly, status | 18:50 | ||
| NeuralAnomaly | samcv, [?] Next release will be in 2 days and 9 hours. Since last release, there are 50 new still-open tickets (4 unreviewed and 1 blockers) and 170 unreviewed commits. See perl6.fail/release/stats for details | ||
| samcv | m: sub foo(:color(:$colour)) { $colour + 1 }; my $s; for ^1000000 { $s += foo(:color($_)) }; say $s; say now - INIT now | 18:51 | |
| camelia | Unexpected named argument 'color' passed? in sub foo at <tmp> line 1? in block <unit> at <tmp> line 1?? | ||
| Skarsnik | how bugs end in this list? x) | 18:58 | |
| Zoffix | Which list? perl6.fail ? It's all of rakudo's currently-open bugs on one page. | 19:05 | |
| Skarsnik | hm, bug with a rakudo tag on rt? | 19:07 | |
| like I don't see this one rt.perl.org/Ticket/Display.html?id=126688 | |||
| AlexDaniel | by the way, what time of the day people expect the release to happen? | 19:08 | |
| for example, releasable6 currently reports slightly different time | |||
| releasable6: status | |||
| releasable6 | AlexDaniel, Next release in 2 days and ?23 hours. 1 blocker. 164 out of 171 commits logged | ||
| AlexDaniel, Details: gist.github.com/0628c6ffa9c5c00537...37921a0c41 | |||
| geekosaur | Skarsnik, it was merged into another ticket | 19:09 | |
| AlexDaniel | so one hour ago on Saturday | ||
| lucasb | m: sub f(:a(:$b)) {}; f(:a) for ^1000 | 19:10 | |
| camelia | ( no output ) | ||
| Skarsnik | how do you see that? for me it's merged with this bug number | ||
| lucasb | m: sub f(:a(:$b)) {}; f(:a) for ^10000 | ||
| camelia | Unexpected named argument 'a' passed? in sub f at <tmp> line 1? in block <unit> at <tmp> line 1?? | ||
| lucasb | known issue, right? | ||
| ^^ samcv showed above | |||
| AlexDaniel | Also, can I have access for “Upload the tarball and the signature to rakudo.org/downloads/rakudo and rakudo.perl6.org/downloads/rakudo/” ? | 19:11 | |
| geekosaur | Skarsnik, I think you just said the same thing. the point is, your ticket is no longer an independent ticket; it will be considered part of the ticket it was merged into | 19:12 | |
| so only that ticket should show up in lists | |||
| AlexDaniel | or is somebody willing to upload these for me? | ||
| Zoffix | Skarsnik: no, all open bugs. It's there on the main page perl6.fail/ You're probably looking at perl6.fail/release/stats , which only shows tickets since last release. | 19:13 | |
| AlexDaniel: what's your public key? | |||
| Skarsnik | Oh yeah, that was it | ||
| AlexDaniel | Zoffix: github.com/AlexDaniel.keys | ||
| Zoffix | how come it's so short? Mine is like 4 lines long: github.com/zoffixznet.keys | 19:14 | |
| lucasb | AlexDaniel: congrats on the commit bit and good luck on shipping the release saturday :) | ||
| AlexDaniel | lucasb: thanks! | 19:15 | |
| Zoffix: I'm wondering that too… hold on… | |||
| geekosaur | Zoffix, yours describes an RSA key with a long bit length; AlexDaniel's is an elliptic curve key | ||
| AlexDaniel | right! | ||
| Skarsnik | AlexDaniel, did you finish your Benchable stuff to check for memory usage? | 19:18 | |
| Zoffix | AlexDaniel: ok, ssh [email@hidden.address] the dirs with tarballs are in /home/rakudo/public_html/downloads/ and ssh [email@hidden.address] the dirs with tarballs are in /var/www/rakudo.perl6.org/public_html | ||
| AlexDaniel | Skarsnik: no, no yet. But why are you interested? | ||
| Skarsnik: because maybe I can do something quickly for you | 19:19 | ||
| Zoffix | As for when people expect release: there's no set time. | ||
| Skarsnik | To show that run 'echo', 'foo'; take nearly 100Mb now. since 2 release | 19:20 | |
| AlexDaniel | Zoffix: ssh-ed to [email@hidden.address] just fine, but can't get into [email@hidden.address] | ||
| Skarsnik | gist.github.com/Skarsnik/6adfba55f...e923112393 | ||
| well not 100Mb, but too much | 19:21 | ||
| AlexDaniel | it asks me for password which probably means that my key was not installed right | ||
| Skarsnik | even when forcing a gc | ||
| AlexDaniel | does it mean that it simply does not support ed25519 somehow? | 19:23 | |
| Zoffix | AlexDaniel: can you try once more? Don't seem I did any changes, but I copy-pasted the line with the key off p6c.org directly to rakudo.org now... | 19:24 | |
| AlexDaniel | Zoffix: nope, doesn't work. Let's try with rsa key maybe, but I'll have to make one now… | 19:25 | |
| samcv | i don't see RAKUDO_MODULE_DEBUG documented anywhere | ||
| and while MoarVM shows a list of respected ENV vars when doing moar --help, perl6 doesn't | |||
| maybe we want a command line option that will show a list of all ENV vars that influence behavior of Perl 6? | |||
| would be a nice option, since it can be a possible security/implementation issue to know what env vars can affect it. and relates directly to the binary itself, and not just the language | 19:26 | ||
| so could make sense to be able to be shown there | |||
| Zoffix | $ cat /etc/*release* | 19:27 | |
| CentOS release 5.8 (Final) | |||
| no idea if that supports ed25519 | |||
| AlexDaniel | Zoffix: ok try this one github.com/AlexDaniel.keys | 19:28 | |
| Zoffix | AlexDaniel: done. Did it work? | 19:29 | |
| ZofBot: the suspense is killing me! | 19:31 | ||
| ZofBot | Zoffix, x = ''; method y { nqp::iseq_s($!x, $!x) } } | ||
| Zoffix | Damn. I need to reset its brain. Nothing but code :( | ||
| AlexDaniel | Zoffix: I think it does, but I have to do something on my end… | 19:32 | |
| Zoffix | ZofBot: how are you feeling now? | 19:33 | |
| ZofBot | Zoffix, To open a ticket, email> "rakudobug@perl | ||
| AlexDaniel | it says “Authentication failed.”, why? | ||
| Zoffix | <3 | ||
| No idea. | 19:34 | ||
| moritz: are you around? AlexDaniel has trouble getting to [email@hidden.address] despite my trying to add a key to keys files :/ any ideas? | |||
| oh | 19:37 | ||
| AlexDaniel | on my end it feels like the key was not installed | ||
| Zoffix | AlexDaniel: can you try one more time? | ||
| AlexDaniel | doesn't work | 19:38 | |
| “Authentication failed.” | |||
| Zoffix | :| no idea | ||
| AlexDaniel | Zoffix: but what was the issue? | ||
| Zoffix | oh wait | ||
| AlexDaniel: can you try again? There are stray newlines in the "sha-rsa sig name" | 19:39 | ||
| were | |||
| AlexDaniel | works | ||
| Zoffix | \o/ | ||
| AlexDaniel | Zoffix: \o/ thanks! | ||
| Zoffix | moritz: nevermind :D | ||
| geekosaur | something "helpfully" line-wrapped it for you. possibly the browser | 19:40 | |
| Zoffix | :) | ||
| AlexDaniel | and ed25519 simply didn't work huh… I remember having this kind of issues with it two years ago or something | 19:42 | |
| geekosaur | if the sshd is old enough then it may not support it | ||
| Zoffix | Yeah, ed was on one line. uname -a tells me the install is from 2011... Still sporting perl 5.8.8 >_< | ||
| geekosaur | well, the kernel is. but it sounds like a downrev debian and its sshd probably does predate elliptic curve keys | 19:44 | |
| AlexDaniel | Zoffix: hm, maybe NeuralAnomaly should take a break, as it currently reports information which is a bit out of sync from what I do | ||
| NeuralAnomaly: status | |||
| NeuralAnomaly | AlexDaniel, [?] Next release will be in 2 days and 8 hours. Since last release, there are 50 new still-open tickets (4 unreviewed and 1 blockers) and 170 unreviewed commits. See perl6.fail/release/stats for details | ||
| AlexDaniel | releasable6: status | ||
| releasable6 | AlexDaniel, Next release in 2 days and ?23 hours. 1 blocker. 164 out of 171 commits logged | ||
| AlexDaniel, Details: gist.github.com/1d9f37c5e6bef048fa...f792161828 | |||
| AlexDaniel | or maybe not, just please don't wonder why there are 170 unreviewed commits :) | 19:45 | |
| Zoffix | NeuralAnomaly: status | 19:47 | |
| NeuralAnomaly | Zoffix, Use releasable6 bot instead. e.g. try: releasable6: status | ||
| Zoffix | NeuralAnomaly: botsnack | ||
| NeuralAnomaly | Zoffix, om nom nom nom | ||
| synopsebot6 | om nom nom | ||
| Zoffix | :) | ||
| <3 IRC::Client | |||
| travis-ci | Rakudo build passed. Elizabeth Mattijsen 'Oops, Skarsnik++' | 19:53 | |
| travis-ci.org/rakudo/rakudo/builds/265227064 github.com/rakudo/rakudo/compare/5...5f7e36c01c | |||
| Skarsnik | timotimo, after a rebuild I get the same fail for the spectest Moar is 2017.07.15-g07 ad rakudo 2017.07-69.g5a5 | 19:56 | |
| AlexDaniel | Zoffix: actually, I can make it respond to NeuralAnomaly | 19:59 | |
| I mean, when somebody says “NeuralAnomaly:” | |||
| but this works too | 20:00 | ||
| hm I can do it anyway | 20:02 | ||
| samcv | i'm excited for the release | 20:08 | |
| so i can get my knuth-morris-pratt string search algorithm merged in afterward :-)) | 20:09 | ||
| AlexDaniel | Skarsnik: oh, it is also using run… so we need unrestricted bot here, I see | 20:16 | |
| Skarsnik: thing is, you can already use Linux::Proc::Statm from within the bot | 20:17 | ||
| Skarsnik | yeah but you want to bisect based on this value and not a run time? | ||
| AlexDaniel | c: use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm<data> | 20:18 | |
| committable6 | AlexDaniel, ¦use: «Cannot find this revision (did you mean “all”?)» | ||
| AlexDaniel | c: HEAD use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm<data> | ||
| committable6 | AlexDaniel, ¦HEAD(3e70d44): «988260» | ||
| AlexDaniel | so in theory you can just “say get-statm<data> < 424242” and give it to bisectable | ||
| Skarsnik | hm that 900Mb what | 20:19 | |
| AlexDaniel squints | |||
| mch: use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm<data> | |||
| committable6 | AlexDaniel, ¦2015.12: «78820» ¦HEAD(3e70d44): «163332» | ||
| AlexDaniel | mch: use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm<data> | ||
| committable6 | AlexDaniel, ¦2015.12: «75144» ¦HEAD(3e70d44): «163336» | ||
| AlexDaniel | c: releases use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm<data> | 20:20 | |
| Skarsnik: I think it was the precompilation | |||
| Skarsnik | precomp was around 200Mb for me I think | ||
| or does it precomp all module? | |||
| committable6 | AlexDaniel, gist.github.com/f73cc92d8da4d09a9a...591e027e3d | ||
| AlexDaniel | just this one I think | ||
| am I… seeing this right? | 20:21 | ||
| c: releases use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm<data> | |||
| committable6 | AlexDaniel, gist.github.com/424dee4cd04b8f0dcb...1199157368 | ||
| Skarsnik | mch: use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm-human<data> | 20:22 | |
| committable6 | Skarsnik, ¦2015.12: «75.132 kB» ¦HEAD(3e70d44): «163.336 kB» | ||
| AlexDaniel | thing is, I don't know what it does when you point different versions to same precomp files | ||
| Skarsnik | I think it redoes it? | ||
| AlexDaniel | I don't know, but numbers seem to go down when you try it again | 20:23 | |
| Skarsnik: I guess you're right, I'll have to do it properly in benchable… | |||
| Skarsnik | what was the test for each release? | 20:24 | |
| c: releases use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm-human<data> | 20:25 | ||
| committable6 | Skarsnik, gist.github.com/4ffbb86b8e7802cf80...b7195ef6c6 | ||
| Skarsnik | still head is crazy | ||
| AlexDaniel | I thing I know how to bisect this XD | 20:26 | |
| Skarsnik | inline the code of l:p:statm | ||
| AlexDaniel | c: 9658dd98c9dd8ec use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm-human<data> | 20:27 | |
| committable6 | AlexDaniel, ¦9658dd9: «163.392 kB» | ||
| AlexDaniel | c: 9658dd98c9dd8ec^^ use lib ‘data/all-modules/Skarsnik/p6-linux-proc-statm’; use Linux::Proc::Statm; say get-statm-human<data> | ||
| committable6 | AlexDaniel, ¦9658dd98c9dd8ec^^: «85.620 kB» | ||
| AlexDaniel | there | ||
| jnthn: do you have any thoughts on this? ? | 20:28 | ||
| Skarsnik: I bisected this by running bisectable as many times as it was needed | |||
| Skarsnik: it feels like in the end the right stuff was precomped and I got to this stable result | |||
| Skarsnik | m: say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] | 20:29 | |
| camelia | 34066? | ||
| AlexDaniel | oh | 20:30 | |
| nice | |||
| Skarsnik | m: sub getpagesize is native(Str) returns int32 { * }; say getpagesize(); | ||
| camelia | 5===SORRY!5=== Error while compiling <tmp>?Can't use unknown trait 'is native' in a sub declaration.?at <tmp>:1? expecting any of:? rw raw hidden-from-backtrace hidden-from-USAGE? pure default DEPRECATED inlinable nodal? … | ||
| Skarsnik | m: use NativeCall; sub getpagesize is native(Str) returns int32 { * }; say getpagesize(); | ||
| camelia | 4096? | ||
| AlexDaniel | bisect: say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] < 100000 | ||
| bisectable6 | AlexDaniel, On both starting points (old=2015.12 new=3e70d44) the exit code is 0 and the output is identical as well | ||
| AlexDaniel, Output on both points: «True» | |||
| Skarsnik | m: say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] * 4096 | ||
| camelia | 140095488? | ||
| AlexDaniel | c: releases say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] | ||
| committable6 | AlexDaniel, gist.github.com/2fd4039561080668f7...93a3d282c8 | 20:31 | |
| Skarsnik | You need the pagesize | ||
| it depend on your system | |||
| AlexDaniel | bisect: say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] < 30000 | ||
| bisectable6 | AlexDaniel, Bisecting by output (old=2015.12 new=3e70d44) because on both starting points the exit code is 0 | ||
| AlexDaniel, bisect log: gist.github.com/5cf477671560bad212...7895708d94 | |||
| AlexDaniel, There are 3 candidates for the first “new” revision. See the log for more details | |||
| AlexDaniel | it's the same result | ||
| Skarsnik | m: say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] * 4096 / 1024; | 20:32 | |
| camelia | 136452? | ||
| Skarsnik | hm, that still a lot | ||
| 136mb | |||
| AlexDaniel | ok let's bisect your wget code. Changes are it's the same thing | 20:33 | |
| Skarsnik | c: release say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] * 4096 / 1024; | ||
| committable6 | Skarsnik, ¦release: «Cannot find this revision (did you mean “releases”?)» | ||
| Skarsnik | c: releases say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] * 4096 / 1024; | ||
| you can put echo "plop" | |||
| committable6 | Skarsnik, gist.github.com/19a3389f02aba56f13...fb297c61a2 | 20:34 | |
| Skarsnik | I think it's the same | ||
| yeah HEAD is big | |||
| in this | |||
| AlexDaniel | Skarsnik: is it identically bad with qqx, shell and run? | 20:35 | |
| Skarsnik | c: releases sub mem { say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] * 4096 / 1024 }; mem(); run 'yes'; mem(); | 20:36 | |
| Yeah, was the same | |||
| committable6 | Skarsnik, gist.github.com/e5a93da89dd32dcb11...e36ad4678f | ||
| Skarsnik | damn | ||
| AlexDaniel | it's ok I can do it here | 20:37 | |
| this is what I'm going to try: gist.github.com/AlexDaniel/4e97dcb...19b2b481a3 | |||
| Skarsnik | there is nqp::gc_force I think to force a gc pass | 20:39 | |
| AlexDaniel | Skarsnik: perhaps join #whateverable if you're interested to see it in action | 20:40 | |
| pmurias | if I find a serious security hole in rakudo what should I do besides fixing it? | 21:39 | |
| Geth | nqp: b15923522c | pmurias++ | src/vm/js/nqp-runtime/multicache.js [js] Stop caching captures with named arguments The partial way we did was causing problems |
21:40 | |
| nqp: dd25a92463 | pmurias++ | 3 files Fix serious security hole |
|||
| Skarsnik | repport it? | ||
| pmurias | when run in a directory perl6-m executes code from the blib subdirectory there | 21:42 | |
| [Coke] | pmurias: theoretically, there is a mailing list for that that ends up in a ticket queue. | ||
| which would give us a chance to remediate the issue before publishing it. | 21:43 | ||
| jnthn | AlexDaniel: The new spesh data collection initially (after that commit) logged closures, which could end up keeping a lot of stuff alive. That was later resolved, though. If HEAD is still costly, not sure. | 21:44 | |
| pmurias | [Coke]: you could easily find the issue from the commit | ||
| AlexDaniel | jnthn: RT #131914 and RT #131915 | ||
| synopsebot6 | Link: rt.perl.org/rt3/Public/Bug/Display...?id=131914 | ||
| Link: rt.perl.org/rt3/Public/Bug/Display...?id=131915 | |||
| AlexDaniel | jnthn: personally, I don't care about 131915 as much, but 131914 is a bit unfortunate :) | 21:45 | |
| pmurias | if you run perl6-m -e '123' in a say a random github repo clone it will execute the hostile bytecode from the repo | ||
| AlexDaniel | jnthn: I think our mistake is not including numbers for HEAD… yes. One second | ||
| jnthn | It's no surprise whatsoever that Proc costs as much as Proc::Async now :) | ||
| Or more | |||
| AlexDaniel | still LTA a bit | 21:46 | |
| pmurias | jnthn: I removed code from nqp that makes perl6-m execute code from $CWD | ||
| jnthn | pmurias: Alright, thanks. Well, it's disclosed now, so I guess all folks can do is upgrade... | ||
| AlexDaniel | jnthn: ¦9658dd98c9dd8ec^^: «58628» ¦9658dd9: «138220» ¦HEAD(3e70d44): «136436» | 21:47 | |
| jnthn | Should probably include a prominent notice about it in the release that's a couple of days away. | ||
| AlexDaniel | that's this code: say "/proc/$*PID/statm".IO.lines[0].split(/\s/)[5] * 4096 / 1024 | ||
| [Coke] | and make sure there's an RT for it if there isn't already. | ||
| pmurias++ | |||
| jnthn | Odd | 21:48 | |
| /usr/bin/time perl6-m -e 'sleep 2' | |||
| 0.11user 0.05system 0:02.22elapsed 7%CPU (0avgtext+0avgdata 70524maxresident)k | |||
| AlexDaniel | pmurias: can you create a ticket? | ||
| pmurias | AlexDaniel: no | ||
| jnthn | 70 is quite a lot less than 136 | ||
| pmurias | AlexDaniel: I mean I didn't create one | 21:49 | |
| AlexDaniel: what should a ticket mention | |||
| jnthn | Wonder what statm is reporting on | ||
| AlexDaniel | [Coke]: ? ? | ||
| Skarsnik | resident (2) resident set size | 21:50 | |
| (same as VmRSS in /proc/[pid]/status) | |||
| shared (3) number of resident shared pages (i.e., backed by a file) | |||
| (same as RssFile+RssShmem in /proc/[pid]/status) | |||
| text (4) text (code) | |||
| lib (5) library (unused since Linux 2.6; always 0) | |||
| data (6) data + stack | |||
| damn | |||
| 5 is data (the doc start at 1) | |||
| [Coke] | At some point, we should have the vulnerability described so that we can write a test for it. I don't know how much detail we want to start with in a public-facing ticket. some guidance from mst or p5 folk or people who have dealt with this in other OS projects would be nice. | 21:51 | |
| (if we knew what we were doing, we'd have already advertised the process) | |||
| (e.g. I just tried to duplicate this with a recentish perl6 and cannot) | |||
| ugexe | with PERL6LIB="" we didnt do anything special | 21:52 | |
| pmurias | [Coke]: you mean you can't duplicate the vulnerability I found | ||
| ? | |||
| [Coke] | pmurias: with the one line description you mentioned? right. | 21:54 | |
| if there's steps to reproduce, I didn't see those. | |||
| ugexe | its just dot @INC | ||
| same thing | |||
| wiki.gentoo.org/wiki/Project:Perl/...NC-Removal # same concept | 21:55 | ||
| [Coke] | so, why don't we try to follow what we'd want in a new ticket and create an item in the security queue. One sec, let me find that email addy. | ||
| travis-ci | NQP build passed. pmurias 'Fix serious security hole' | 21:56 | |
| travis-ci.org/perl6/nqp/builds/265341823 github.com/perl6/nqp/compare/99abf...25a9246389 | |||
| [Coke] | please open a security ticket by emailing to [email@hidden.address] | 21:57 | |
| ugexe | github.com/perl6/roast/commit/10cf...2519fcf3a9 # here is a test for a similar flaw at perl6 level | ||
| AlexDaniel | .tell stmuk you may be interested in this discussion: irclog.perlgeek.de/perl6-dev/2017-...i_15027543 | ||
| yoleaux | AlexDaniel: I'll pass your message to stmuk. | ||
| [Coke] | moritz & I have access to that today, so we can do a run through about how things would work from there. | 21:58 | |
| ugexe | from 'Avoid magic $cwd in module search path RT #130883 ' | 21:59 | |
| synopsebot6 | Link: rt.perl.org/rt3/Public/Bug/Display...?id=130883 | ||
| [Coke] | so, I'm confused I guess - this is a new vulnerability, or one that was fixed and we just want to advertise the fix? | 22:01 | |
| if it's pre-existing and reported, don't bother opening a new ticket. If it's new, then the exercise is probably still worth it. | |||
| [Coke] commutes. | |||
| ugexe | it gives prior art for how such a ticket might be handled, and a model to test it | ||
| pmurias | [Coke]: it's old pre-exisiting an fixed | ||
| * and | |||
| [Coke]: there are some details to reproduce it which I'll mention in the ticket I'm writing | 22:02 | ||
| [Coke] | ah, then ignore my new ticket crap. Use that for -new- security vulnerabilities. :) | ||
| pmurias | [Coke]: so should I explain the details on irc? | 22:04 | |
| [Coke]: I don't think it was reported | |||
| timotimo | [Coke]: the profiler will grab a template from a relative path if it can (it assumes it might be in the nqp repository) - is this security-relevant or just unfortunate? | 22:09 | |
| ugexe | its only a problem if you are running something with elevated privs that will load something from somewhere that someone with less privs can place code (like /tmp) | 22:11 | |
| pmurias | [Coke]: I sent the mail to perl6-security-report | ||
| ugexe | perl community seemed pretty split on if perl should fix dot @INC or if debian should just be more careful | 22:12 | |
| e.g. i dont think the security ramifications are huge here | 22:13 | ||
| pmurias | if you download a repo from github that you don't trust you can't run any Perl 6 programs in that directory | 22:14 | |
| ugexe | that is always the case | ||
| timotimo | ugexe: well, it does generate a html page that the user then probably opens right away, so could be a way to abuse a browser bug. if the profiler result is then hosted somewhere it could steal cookies and stuff | 22:15 | |
| ugexe | it has to have moarvm bytecode, which i assume also has to be generated specifically for your env/rakudo | ||
| pmurias | ugexe: do you expect cd ~/untrusted-directory;perl6-m -e 'say "hi"' to hijack your system? | 22:16 | |
| ugexe | m: use Test:from<NQP>; | ||
| camelia | ===SORRY!===?While looking for 'Test.moarvm': no such file or directory? | ||
| ugexe | i dont expect a Test.moarvm from some repo to run on my local system | ||
| pmurias | perl6-m would just load the blib/Perl6/Ops.moarvm from the $CWD | 22:18 | |
| ugexe | oh, it autoloads modules, not just sets up a lib that can be hijacked? | 22:19 | |
| pmurias | ugexe: yes | 22:20 | |
| ugexe | still, it would have to be precompiled already as well yes? | ||
| timotimo | ugexe: actually, you can build a .moarvm file to rely on no other files | ||
| or at least it should be possible | |||
| ugexe | doesn't its location on the file system matter? | ||
| or what built in? | 22:21 | ||
| built it^ | |||
| pmurias | you can avoid serialization dependencies in a custom built .moarvm | 22:22 | |
| I noticed the exploit because perl6-m was annoying me by loading bytecode from a different checkout | 22:24 | ||
| ugexe | ive seen it in gdb output but I never thought it could be exploited as such | 22:36 | |
| or maybe strace | |||
| stat("./NQPCORE.setting.moarvm", 0x7ffd00da2450) = -1 ENOENT (No such file or directory) | 22:40 | ||
| stat("blib/NQPCORE.setting.moarvm", 0x7ffd00da2450) = -1 ENOENT (No such file or directory) | |||
| stuff like that | |||
| travis-ci | Rakudo build failed. Stefan Seifert 'Remove now misleading comment | 22:52 | |
| travis-ci.org/rakudo/rakudo/builds/265279497 github.com/rakudo/rakudo/compare/9...70d44d47d1 | |||
| buggable | [travis build above] ? Did not recognize some failures. Check results manually. | ||
| ugexe | While looking for 'Perl6/ModuleLoader.moarvm': no such file or directory | 22:53 | |
| it must have pulled in that nqp commit | |||
| now i remember why i left this alone :) | 22:55 | ||
| pmurias: that breaks rakudo for some reason | 23:02 | ||
| pmurias | ugexe: I'll grab some sleep and if nobody fixes rakudo get round to it tomorrow morning | 23:17 | |
| ugexe | its in the Makefile. it needs the nqp equivilent of -I. or -Iblib/ | 23:18 | |
| ...but i dont know what that is | |||
| there we go. --module-lib | 23:20 | ||
| Zoffix | .tell [Coke] good news! I got paid :) | 23:26 | |
| yoleaux | Zoffix: I'll pass your message to [Coke]. | ||
| Zoffix | Spent it all in one place on a shiny new toy :D twitter.com/zoffix/status/897962614696882182 | ||
| [Coke] | Zoffix: (paid) glad to hear it, thanks for your work! | 23:46 | |
| yoleaux | 23:26Z <Zoffix> [Coke]: good news! I got paid :) | ||
| samcv | yay for getting paid | 23:59 |