| MasterDuke | timotimo: so it knows the method doesn't exist, but ^methods and ^private_method_table don't return anything. any other way to get the methods that are known? | 00:04 | |
| timotimo | i'd have to look through the roles that are mixed together for all the HOWs to figure that out | 00:07 | |
| but now i'll go to bed | |||
| MasterDuke | later... | 00:09 | |
| BenGoldberg | ===SORRY!===QAST::Block with cuid 1 has not appeared | 01:39 | |
| Oops. | |||
| m: gist.github.com/BenGoldberg1/d0764...549f4bc359 | |||
| camelia | ===SORRY!=== QAST::Block with cuid 9 has not appeared |
||
| Geth | rakudo/nom: aa3684218b | (Zoffix Znet)++ | src/core/Junction.pm Fix crash when Junctioning empty array after .elems'ing it `my @a; say +@a; say @a.all` crashes with `This type (Scalar) does not support elems` because the `.elems` call touches $!reified attr, making it nqp::attrinited. Then, when we go to make a shiny new Junction, we check if the $!reified is inited and use it if it is. However, it's inited because .elems touched it, not because it got stuff up in it. Fix by checking concreteness of $!reified instead of initness. |
02:01 | |
| roast: 932eb4c444 | (Zoffix Znet)++ | S03-junctions/misc.t Test Junctions on touched empty array don't crash Rakudo fix: github.com/rakudo/rakudo/commit/aa...218b1f668b |
02:02 | ||
| rakudo/nom: 61ecfd5117 | (Zoffix Znet)++ | src/core/Junction.pm Fix crashes/LTA errors in Junction.new w/wrong args Add a proto so we don't go into Mu.new's candidates. |
02:13 | ||
| roast: beea798132 | (Zoffix Znet)++ | S03-junctions/misc.t Test Junction.new with wrong args has sane errors Rakudo fix: github.com/rakudo/rakudo/commit/61...72b0e3cf20 |
02:14 | ||
| MasterDuke | interesting. i get different errors for missing methods depending on whether optimization is on or not | 02:29 | |
| `perl6 -e 'class C { method foo { self!bar }; method bar { say "hi" } }; C.new.foo'` gives `No such private method 'bar' for invocant of type 'C'` | |||
| `perl6 --optimize=0 -e 'class C { method foo { self!bar }; method bar { say "hi" } }; C.new.foo'` gives `No such private method '!bar' for invocant of type 'C'. Did you mean 'bar'? in method foo at -e line 1 in block <unit> at -e line 1` | 02:31 | ||
| nine | ugexe: the CORE dist is always force installed because back then, we knew as little about versioning the CORE dist as we do now :) | 06:22 | |
| [Tux] | This is Rakudo version 2017.05-347-g61ecfd511 built on MoarVM version 2017.05-25-g62bc54e9 | 08:40 | |
| csv-ip5xs 2.863 | |||
| test 13.064 | |||
| test-t 4.400 - 4.427 | |||
| csv-parser 13.238 | |||
| dogbert17 | I'm trying to update some docs and stumbled upon this text in the traps ducument | 11:20 | |
| "(Note that BUILDALL is a method, not a submethod. That's because by default, there is only one such method per class hierarchy, whereas BUILD is explicitly called per class. Which also explains why you need the nextsame inside BUILDALL, but not inside BUILD)." | |||
| isn't the mention on nextsame here incorrect, shouldn't it be callsame? (in the example above it actually says callsame): github.com/perl6/doc/blob/master/d....pod6#L129 | 11:21 | ||
| s/on/of/ | |||
| jnthn | Yes, it needs to be callsame there; were it nextsame, then the $!y = 18 would be dead code | 11:24 | |
| That really that should talk about TWEAK | 11:25 | ||
| dogbert17 | thx, I'll update it immediately | ||
| jnthn | Since TWEAK is a better solution to that problem than overriding BUILDALL | ||
| dogbert17 | yeah, I have noticed that some parts of the Lnaguage docs are a bit old in the tooth, probably written before TWEAK made an appearance I guess | 11:26 | |
| jnthn | Yeah, it surely was | ||
| dogbert17 | here's another callsame part that looks a bit old; github.com/perl6/doc/blob/master/d....pod6#L240 | 11:28 | |
| I'll see if I can improve that a tiny bit | 11:32 | ||
| MasterDuke | jnthn: did you notice my comments from last night about turning off optimizing and getting different errors for missing methods? | 11:33 | |
| jnthn | MasterDuke: Yes, but you missed the important part of the output. | 11:35 | |
| MasterDuke | yeah, one's compile time one isn't | ||
| jnthn | Right :) | ||
| That's 'cus the optimizer looks to try and produce better code for missing methods, and as a side-effect detects they are missing and can complain about it | 11:36 | ||
| MasterDuke | yep, i noticed that in Mu.pm's dispatch<!> | ||
| in there, if i iterate over `type`'s methods and private_method_table i see the user-defined methods (e.g., foo and bar) | 11:37 | ||
| so i tried sending `type` to the X::Method::NotFound.new instead of `SELF`, but that didn't seem to work | 11:38 | ||
| jnthn | Hm, odd. I dunno why; by the time we're in the optimizer the types are fully composed | 11:39 | |
| MasterDuke | wait, in the optimizing case dispatch<!> isn't called at all | ||
| oh, hm, need to re-compile to confirm that | 11:40 | ||
| jnthn | Yes, exactly | ||
| It looks at the private_method_table directly | 11:41 | ||
| lunch & | 11:42 | ||
| MasterDuke | ah, i see optimize_private_method_call() in Optimizer.pm | 11:58 | |
| hm, i still don't understand why what becomes the $.invocant in X::Method::NotFound doesn't have any of the local methods | 12:03 | ||
| Geth | rakudo/nom: 77d3c5465c | (Zoffix Znet)++ | src/core/operators.pm Fix infix:<orelse> calling .defined one too many times Check if we're at the last arg *before* we check definedness, otherwise we accidentally disarm Failures when we shouldn't |
12:27 | |
| roast: a215d2e021 | (Zoffix Znet)++ | S03-operators/orelse.t Test infix:<orelse> calls .defined exact number of times Rakudo fix: github.com/rakudo/rakudo/commit/77...5c771387bf |
|||
| rakudo/nom: 64e898f9ba | (Zoffix Znet)++ | src/core/Failure.pm Re-arm Failures when `fail`ing with them Maintain consistent interface that a call to &fail returns an explosive Failure, even if the arg was a handled Failure. |
12:29 | ||
| roast: 6c06e3c123 | (Zoffix Znet)++ | S04-exceptions/fail.t Test &fail re-arms handled Failures Rakudo fix: github.com/rakudo/rakudo/commit/64...baa159e201 |
|||
| roast: 53bb278ccf | (Zoffix Znet)++ | 3 files Spec andthen/orelse/notandthen must not call .defined on last arg |
13:45 | ||
| MasterDuke | Zoffix: is this expected? gist.github.com/Whateverable/760a5...bbe71b6445 | 14:04 | |
| jnthn: ^^^ the most recent change in behavior is because of github.com/rakudo/rakudo/commit/8e...01eccdc7d7 | 14:09 | ||
| lizmat: ^^^ and '-'.IO.slup actually breaking is because of github.com/rakudo/rakudo/commit/b4...7725fa53f3 | 14:14 | ||
| s/slup/slurp | |||
| dogbert17 | m: gist.github.com/dogbert17/63f81505...d718f9bcc8 | 14:44 | |
| camelia | False | ||
| dogbert17 | commit: 2016.07 gist.github.com/dogbert17/63f81505...d718f9bcc8 | ||
| committable6 | dogbert17, It looks like a URL, but mime type is ‘text/html; charset=utf-8’ while I was expecting something with ‘text/plain’ or ‘perl’ in it. I can only understand raw links, sorry. | ||
| dogbert17 | WAT | 14:45 | |
| mst | commit: 2016.07 gist.githubusercontent.com/dogbert...c/test.pl6 | 14:46 | |
| committable6 | mst, Successfully fetched the code from the provided URL. | ||
| mst, ¦2016.07: «False» | |||
| dogbert17 | mst, thx | 14:50 | |
| mst | *thanks | ||
| dogbert17 | commit: 2016.01 gist.githubusercontent.com/dogbert...c/test.pl6 | 14:51 | |
| committable6 | dogbert17, Successfully fetched the code from the provided URL. | ||
| dogbert17, ¦2016.01: «False» | |||
| dogbert17 | commit: 2015.06 gist.githubusercontent.com/dogbert...c/test.pl6 | ||
| committable6 | dogbert17, Successfully fetched the code from the provided URL. | 14:52 | |
| dogbert17, ¦2015.06: «False» | |||
| lizmat | . | ||
| dogbert17 | commit: 2015.01 gist.githubusercontent.com/dogbert...c/test.pl6 | 15:00 | |
| committable6 | dogbert17, Successfully fetched the code from the provided URL. | ||
| dogbert17, ¦2015.01: «False» | |||
| dogbert17 | anyone good at regexes here | 15:01 | |
| the example on line 514, should it return 'True' as the text claims or 'False'? github.com/perl6/doc/blob/master/d....pod6#L504 | 15:02 | ||
| timotimo | the example may want a .chomp | 15:12 | |
| dogbert17 | timotimo: will test | 15:17 | |
| timotimo: you mean like 'my $str = chomp q:to/EOS/; ...' | 15:18 | ||
| that seems to work :) timotimo++ | |||
| jnthn | MasterDuke: That looks like IO::Handle.open isn't repsecting :bin | 15:20 | |
| MasterDuke: In the case of '-' I mean | |||
| dogbert17 | m: my $str = "string with newline\n"; say so $str ~~ / line $$/; | 15:34 | |
| camelia | True | ||
| dogbert17 | commit: 2015.11 my $str = "string with newline\n"; say so $str ~~ / line $$/; | ||
| committable6 | dogbert17, ¦2015.11: «True» | ||
| dogbert17 | the last part of this doc piece must be incorrect: | 15:35 | |
| "$$ matches only at the end of a logical line, that is, before a newline character, or at the end of the string when the last character is not a newline character." | |||
| feels as if the last part 'when the last character is not a newline character' should be dropped | 15:36 | ||
| hmm, according to S05 the above code should return false | 15:43 | ||
| yoleaux | Zoffix: Future-Zoffix: github.com/perl6/doc/issues/401 | 15:45 | |
| TimToady | m: say "\n" ~~ m:g/$$/ | 16:03 | |
| camelia | (｢｣) | ||
| TimToady | m: say " " ~~ m:g/''/ | ||
| camelia | (｢｣ ｢｣) | ||
| TimToady | m: say " " ~~ m:g/$$/ | ||
| camelia | (｢｣) | ||
| TimToady | looks truish to me | ||
| we don't want the null string after a \n to match, since \n is generally construed as a terminator, not a separator | 16:05 | ||
| dogbert17 | TimToady: so does that mean that this no longer applies "^^ and $$ match line beginnings and endings. (The /m modifier is gone.) They are both zero-width assertions. $$ matches before any \n (logical newline), and also at the end of the string if the final character was not a \n." | 16:17 | |
| moritz | which part of this statement do you think is not true? | 16:23 | |
| sorry, missing a bit of context here | |||
| dogbert17 | moritz: in our docs, we have the following text "$$ matches only at the end of a logical line, that is, before a newline character, or at the end of the string when the last character is not a newline character." | 16:24 | |
| moritz | which seems to agree with S05 | 16:25 | |
| or am I missing something? | |||
| dogbert17 | well, my brain is probably mushy, but to me the texts implies that this code should return false | 16:26 | |
| m: my $str = "string with newline\n"; say so $str ~~ / line $$/; | |||
| camelia | True | ||
| moritz | it matches before the newline character | 16:27 | |
| which is exactly what both S05 and the docs say | |||
| m: say "string without newline" ~~ /line $$/ | |||
| camelia | ｢line｣ | ||
| moritz | and that's the second case: at the end of the string when there's no final newline character | 16:28 | |
| dogbert17 | gah, I need to wrap my brain around this | ||
| thanks for trying to help | 16:31 | ||
| MasterDuke | timotimo: does $foo.^methods get compiled to $foo.HOW.methods($foo)? is there any difference between those? | 19:04 | |
| jnthn | MasterDuke: Pretty sure that's exactly how it's compiled now; in the past there was a Mu.dispatch:<.^> helper but that caused issues aside from being a performance hit | 19:19 | |
| MasterDuke | m: use NativeCall; my \compar_t = Callable but role :: { method sub_signature { :(Pointer,Pointer --> int) } }; | ||
| camelia | X::Method::NotFound exception produced no message in block <unit> at <tmp> line 1 |
||
| MasterDuke | i'm trying to fix that ^^^ no message | ||
| it's dying here github.com/rakudo/rakudo/blob/nom/...on.pm#L180 | 19:20 | ||
| the nqp::can succeeds, but something about the ^methods call goes horribly wrong | 19:21 | ||
| the message before it gets there is `No such method 'mixin' for invocant of type 'Perl6::Metamodel::ParametricRoleGroupHOW'` | 19:22 | ||
| i stuck a note() in ParametricRoleGroupHOW's methods(), but it's not printing | 19:23 | ||
| timotimo | perhaps have to look into CurriedRoleHOW? | 19:24 | |
| MasterDuke | hm, it doesn't have a methods() | 19:25 | |
| might that be the problem? | |||
| timotimo | you looked through all the roles it assembles, too? | 19:26 | |
| MasterDuke | which it? | ||
| timotimo | CurriedRoleHOW | 19:27 | |
| all of the HOWs are made up of many roles usually | |||
| jnthn | I suspect nqp::can on a CurriedRoleHOW is about worthless in so far as .^find_method will always return a closure that will pun the role and then delegate the method to the pun | ||
| Well, on something whose .HOW is a CurriedRoleHOW, I mean | |||
| Hm, but it's doing it on the .HOW which should be OK | 19:28 | ||
| m: use nqp; say nqp::can(CurriedRoleHOW, 'methods') | |||
| camelia | 5===SORRY!5=== Error while compiling <tmp> Undeclared name: CurriedRoleHOW used at line 1 |
||
| jnthn | m: use nqp; say nqp::can(Metamodel::CurriedRoleHOW, 'methods') | ||
| camelia | 0 | ||
| MasterDuke | but ParametricRoleGroupHOW isn't/doesn't CurriedRoleHOW. why would i look there? | ||
| jnthn | m: use nqp; say nqp::can(Metamodel::ParametricRoleGroupHOW, 'methods') | 19:29 | |
| camelia | 1 | ||
| jnthn | m: role R { }; say R.HOW.^name | 19:30 | |
| camelia | Perl6::Metamodel::ParametricRoleGroupHOW | ||
| jnthn | m: role R { }; say R.^methods | ||
| camelia | () | ||
| jnthn | Hm, that looks OK | ||
| m: role R { }; say R.^methods(:all) | |||
| camelia | () | ||
| timotimo | i was guessing about the involvement of CurriedRoleHOW, because they are kinda related | 19:31 | |
| jnthn | Is methods actually the problem, or is the problem that you end up doing .^methods on an NQPClass and it gives back an NQP routine? | 19:32 | |
| MasterDuke | hm. let me see... | 19:33 | |
| i changed it to: for $!invocant.^methods(:all) -> $method { note("public method"); my $method_name = $method.name; | 19:37 | ||
| "public method" does not get printed | |||
| jnthn | Hm, ok | 19:40 | |
| Can you put a CATCH block in to see how it breaks? | |||
| MasterDuke | building that... | 19:42 | |
| `Unexpected named argument 'all' passed` | 19:44 | ||
| ParametricRoleGroupHOW's methods() just passes on a capture of the named parameters | 19:54 | ||
| jnthn | Hm, where do? | 19:56 | |
| *where to | |||
| I guess grep for the methods methods and throw a *%ignore on them | 19:57 | ||
| to slurp up any unknown ones | |||
| MasterDuke | huh. i have a note() as the first line in ParametricRoleGroupHOW's methods(), why doesn't that get printed? | 19:58 | |
| there are only two 'method methods' | 19:59 | ||
| src/Perl6/Metamodel/MethodContainer.nqp:45: method methods($obj, :$local, :$excl, :$all) { | 20:00 | ||
| src/Perl6/Metamodel/ParametricRoleGroupHOW.nqp:170: method methods($obj, *@pos, *%name) { | |||
| jnthn | MasterDuke: Huh, indeed... | 20:04 | |
| MasterDuke | jnthn: yeah. i guess i could always try() with :all and then do without if that throws, but that's not pretty | 20:14 | |
| any more debugging ideas? | 20:17 | ||
| timotimo | maybe it passes %name on to something else? | ||
| MasterDuke | `$c.HOW.methods($c, |@pos, |%name);` in ParametricRoleGroupHOW.nqp | 20:23 | |
| `for self.parents($obj, :all($all), :excl($excl)) {` in MethodContainer.nqp | |||
| only places they're used in those methods() | 20:24 | ||
| huh. that `Unexpected named argument 'all' passed` is from github.com/MoarVM/MoarVM/blob/mast...#L386-L388 not github.com/rakudo/rakudo/blob/nom/...P.nqp#L885 | 20:38 | ||
| any extra debug info i could get from an MVMArgProcContext? | 20:39 | ||
| here's a MVM_dump_backtrace when i broke at that line gist.github.com/MasterDuke17/2a92c...e7b8e2bbee | 20:42 | ||
| jnthn | Oh, wow, it's making it all the way to calling .methods on knowhow... | 20:43 | |
| oh wait, or is it | |||
| Oh, no | 20:44 | ||
| but | |||
| (/home/dan/Source/perl6/install/share/nqp/lib/nqpmo.moarvm:methods) | |||
| That's the giveaway | |||
| It's a method methods somewhere in src/how/ in NQP that likely wants a look | |||
| MasterDuke | ? | ||
| jnthn | The nqpmo being the cldue | ||
| *clue | |||
| MasterDuke | ah, there are three in NQP | ||
| and none have :$all | 20:45 | ||
| just :$local | |||
| jnthn | Well, problem found :) | ||
| So far as NQP's MOP is concerned, :$all being true or false would give the same results anyway | |||
| So can just add :$all in the signature for compat, and ignore its value | 20:46 | ||
| MasterDuke | ok, trying that now | 20:47 | |
| making progress, didn't die right away, now to take out some of my debugging prints... | 20:53 | ||
| ah ha! now `use NativeCall; my \compar_t = Callable but role :: { method sub_signature { :(Pointer,Pointer --> int) } };` just dies with `No such method 'mixin' for invocant of type 'Perl6::Metamodel::ParametricRoleGroupHOW'. Did you mean 'min'?` | 20:55 | ||
| jnthn++ | 20:56 | ||
| jnthn | MasterDuke++ # actually fixing it :) | 20:58 | |
| Geth | nqp: c4af44b05b | MasterDuke17++ | 3 files Add :$all to methods() for Perl 6 compatability For some types, calling .^methods(:all) in Rakudo would end up in NQP. None of the NQP methods() had a :$all parameter, so they would die with `Unexpected named argument 'all' passed`. Per irclog.perlgeek.de/perl6-dev/2017-...i_14681666 :$all doesn't have any effect, so this commit adds them only for compatability. BenGoldberg++ for the find, timotimo++ and jnthn++ for the debugging help. |
21:09 | |
| BenGoldberg | You're welcome :) | 21:11 | |
| timotimo | 4ice | 21:12 | |
| ... nice. | |||
| MasterDuke | m: class C { method foo { self!bar }; method bar { say "hi" } }; C.new.foo | ||
| camelia | 5===SORRY!5=== Error while compiling <tmp> No such private method 'bar' for invocant of type 'C' at <tmp>:1 ------> 3class C { method foo { self!7⏏5bar }; method bar { say "hi" } }; C.new. |
||
| MasterDuke | next thing on the list, how to get ^^^ to suggest the public 'bar'? | 21:13 | |
| timotimo | i'm not really sure it should, as private and public methods are so strongly different | ||
| you may need to also check what class the method belongs to because private methods are only valid in the same class | 21:14 | ||
| jnthn | So long as it indicates clearly that it's public, not private... | ||
| MasterDuke | heh. your past self disagrees. rt.perl.org/Ticket/Display.html?id=123078 | ||
| timotimo | if you only check if "self" is the right one, you may end up suggesting private methods that don't exist in the actual class | ||
| i'm smarter now :) | 21:15 | ||
| jnthn | Or more optimistic then :P | ||
| MasterDuke | yeah, i think what i just did (the way i added suggesting public/private methods, not the the NQP commit) is a little wrong, or less that optimal | 21:17 | |
| any thoughts on when/how to suggest public vs private? | 21:18 | ||
| BenGoldberg | If there's an exact name match. | 21:34 | |
| timotimo | (at least i hope i'm smarter) | 21:38 | |
| MasterDuke | for both cases? you tried to call a non-existing private method, only suggest a public one if it's exact? and if you tried to call a non-existing public method, only suggest a private one if it's exact? | ||
| BenGoldberg | No, I'm just saying that that's what I think *ought* to happen :) | 21:39 | |
| MasterDuke | i feel i should get some votes from @LARRY | 21:47 | |
| timotimo: i'm getting the private methods via ^private_method_table, so i believe they should only be for the same class | 21:55 | ||
| yoleaux | AlexDaniel: Future-AlexDaniel: Fix all 103 whateverable issues | 21:56 | |
| timotimo | that's right | 22:00 | |
| well, i hope that it is | |||
| you could look at dispatch:<!> | |||
| MasterDuke | that calls find_private_method(), which just looks in %!private_methods, which is what private_method_table() returns | 22:04 | |
| AlexDaniel | yoleaux fffuuuuuuu | 22:19 | |
| Geth | roast: 6a78d5ddf1 | (Zoffix Znet)++ (committed using GitHub Web editor) | packages/Test/Util.pm Remove duplicate .resolve The Win32 fix seems to have been untested[^1] as a duplicate .resolve was still in use. [1] github.com/perl6/roast/commit/b5b9...ada0f32ccd |
23:01 | |
| Zoffix | Did we come up with security bug policy at the Toolchain Summit? | 23:46 | |
| yoleaux | 01:40Z <BenGoldberg> Zoffix: There is no whatever-anything in the code: gist.github.com/BenGoldberg1/d0764...549f4bc359 | ||
| Zoffix | 'cause I found a security bug | ||
| And so like.. do I just fix it? Do we notify someone first? Do we notify users via some channels? | 23:48 | ||
| Zoffix don't yet got a fix, but is compiling a poor attempt | |||
| ugexe | policy is... policy | 23:50 | |
| in other words: any security policy I would imagine would be rakudo specific | 23:51 | ||
| Zoffix | poor attempt fails | ||
| Sure. But lizmat mentioned it would be a topic for PTS before PTS happened. | |||
| ugexe | heh, it was for perl5 at least | 23:53 | |
| Zoffix | heh | 23:54 |