Perl 6 - the future is here, just unevenly distributed

IRC log for #perl11, 2014-09-25

| Channels | #perl11 index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:03 rurban joined #perl11
00:32 travis-ci joined #perl11
00:32 travis-ci [travis-ci] perl11/potion#775 (getops - 821f121 : Reini Urban): The build passed.
00:32 travis-ci [travis-ci] Change view : https://github.com/perl11/potion/compare/953f6560d937...821f1213a584
00:32 travis-ci [travis-ci] Build details : http://travis-ci.org/perl11/potion/builds/36211989
00:32 travis-ci left #perl11
01:50 bulk88 joined #perl11
05:43 travis-ci joined #perl11
05:43 travis-ci [travis-ci] perl11/potion#776 (getops - e1954f5 : Reini Urban): The build passed.
05:43 travis-ci [travis-ci] Change view : https://github.com/perl11/potion/compare/821f1213a584...e1954f520abf
05:43 travis-ci [travis-ci] Build details : http://travis-ci.org/perl11/potion/builds/36224579
05:43 travis-ci left #perl11
05:47 travis-ci joined #perl11
05:47 travis-ci [travis-ci] perl11/potion#777 (getops - 0394c35 : Reini Urban): The build passed.
05:47 travis-ci [travis-ci] Change view : https://github.com/perl11/potion/compare/e1954f520abf...0394c35b9f7e
05:47 travis-ci [travis-ci] Build details : http://travis-ci.org/perl11/potion/builds/36224828
05:47 travis-ci left #perl11
08:12 basiliscos joined #perl11
09:48 basiliscos joined #perl11
12:51 basiliscos1 joined #perl11
13:34 basiliscos joined #perl11
14:17 davido___ joined #perl11
15:21 basiliscos joined #perl11
16:05 rurban joined #perl11
16:44 basiliscos joined #perl11
18:13 travis-ci joined #perl11
18:13 travis-ci [travis-ci] perl11/potion#778 (getops - 3bc8c03 : Reini Urban): The build passed.
18:13 travis-ci [travis-ci] Change view : https://github.com/perl11/potion/compare/0394c35b9f7e...3bc8c03c69f1
18:13 travis-ci [travis-ci] Build details : http://travis-ci.org/perl11/potion/builds/36282253
18:13 travis-ci left #perl11
18:40 willthechill joined #perl11
18:45 willthechill joined #perl11
18:49 travis-ci joined #perl11
18:49 travis-ci [travis-ci] perl11/potion#779 (getops - 3ab2ddb : Reini Urban): The build passed.
18:49 travis-ci [travis-ci] Change view : https://github.com/perl11/potion/compare/3bc8c03c69f1...3ab2ddb5de3b
18:49 travis-ci [travis-ci] Build details : http://travis-ci.org/perl11/potion/builds/36285683
18:49 travis-ci left #perl11
18:49 willthechill rurban: what have you been up to with all the p2 and potion stuff?
18:53 rurban willthechill: having more time now. now I'm doing performance: faster array index access, to get to c-like speed, everything else is already as fast as c, just arrays not.
18:53 rurban did some benchmarks and examples, and arrays are horrible slow. now they are fast.
18:54 rurban for p2 I can only use unchecked fast array access only when properly typed. with potion I have 2 seperate syntax to represent fast but insecure or slow but secure.
18:54 rurban array[i] fast, array(i) slow
18:55 willthechill nice
18:55 rurban hash["key"] is always a method call to get the hash index, even if the key is a literal
18:55 rurban but it looks better than hash("key") or hash at("key")
18:56 rurban haven't used {} yet. this is still free to grab
18:57 rurban [] is enough to represent arrays and hashes and unevaluated lists (licks), like lisp lists.
18:57 rurban how was your meeting yesterday?
18:58 willthechill the Austin.pm meeting was on Tuesday night, it went very well, I presented both ShinyCMS and RapidApp running on Catalyst
18:58 willthechill finishing up my Shiny work now so I can get back to RPerl
19:03 rurban I think I'll soon need to add var-tracking to know compile-time types for my variables, to do better optimizations and checks.
19:04 rurban I only know explicitly declared types, now but I'm skipping implicit types derived by the syntax" arrays, hashes, calls, strings, int, numbers.
19:04 willthechill only strict typing allowed for rperl
19:05 willthechill I only use Perl to check that declared types are valid at Perl compile time, then C/C++ does the full type-checking at C/C++ compile time
19:05 rurban I'm more dynamic, but will handle some kind of simple type-inferencer. no proper subtypes and user-declarable types yet.
19:05 willthechill I want no inference, that is too much magic for RPerl
19:05 rurban C++ should be better than my type-checker
19:06 willthechill there is no need for me to re-invent full type checking when C/C++ does it for me
19:06 rurban no magic and no heuristics. only what's absolutely clear is stored.
19:06 willthechill I only have to partially implement it in Perl to make sure the types are valid in the first place, during variable declaration
19:07 willthechill okay well no heuristics and no magic is good then
19:07 rurban I need to seperate static types from dynamic ones at compile-time. currently I only have room for the static ones, and the dyanmic ones are stored in the value, but no  in the variable representing it.
19:08 rurban should be simple to implement
19:08 rurban but low prio
19:09 rurban and my compiler is getting bigger and bigger, need to simplify it a bit.
19:09 rurban too much special syntax
19:09 willthechill I really don't want dynamic types but I'm keeping a placeholder type called "scalar" which can be used for dynamic type in the future if medium-magic RPerl is needed
19:10 rurban for bool and undef also?
19:10 rurban I think the biggest problem is to check bool and undef with int, num, str
19:11 rurban or make bool just an int of 0 or 1
19:16 willthechill undef is tricky, it is different that NULL in C/C++
19:16 willthechill I have bool on the to-do list
19:17 willthechill bool will be a normal int in Perl (which is a scalar in Perl anyway), and bool will be a real bool in C/C++
19:17 willthechill one of the main points of RPerl is to match Perl types with C/C++ types
19:17 rurban currently I have a seperate bool type, as in perl6. for perl5 I'd need to special-case the checks
19:18 rurban but my undef/nil is thanksfully the same as NULL (0)
19:18 rurban I'm rather unsafe and fast, the libs and system checks NULL pointers already
19:19 willthechill I can let Perl use scalar int for bool in Perl data mode, and then have C/C++ do strict typechecking for bool as real 0/1 bool in C data mode
19:20 rurban my parser gives me still strong headaches, heredocs are a nightmare. maybe I'll switch to yacc. I'll have to check the perl6 and ruby syntax other people wrote with peg.
19:21 rurban a real static bool sounds cool
19:22 rurban maybe I'll also have to fork off perl5, as they went completely to war now. peal would be cool name, or just perl5i or perl5p
19:23 rurban but I rather improve p2 first, enough time until I get to the next compiler release, with modules and rperl support.
19:24 rurban with parrot this destruction period took 2 years. for perl5 it might be longer or shorter. but the damage is going on since years already.
19:32 willthechill well the Perl foundation owns the copyright to the name "Perl", and we must maintain compatibility with Perl 5 pumpkin core because that is the only true Perl
19:32 willthechill pumpkin is too ingrained in all operating systems and the whole internet to ever be removed
19:33 willthechill which is why RPerl is good because it works with the existing Perl and C infrastructure
19:33 willthechill if you break compatibility, you will not succeed
19:33 willthechill just look at every other failed project
19:34 rurban compat with perl5 is no problem, just project management and bugfixes/bad internals are the problem
19:34 rurban no compat because I don't want to have seperate compilers, and we need cpan. just do the fixes they refused to do.
19:35 rurban I wrote a big list of problems to be addressed but didn't published as it was too angry and personal
19:36 rurban But If I fork it can finally add type support, and improve methods and oo, and much more. all things they'll never do
19:37 rurban and proper readonly support
19:37 willthechill yes but if you fork, nobody will use your fork
19:37 rurban and faster hashes, realloc, proper cow, forbid binary symbols and such
19:37 willthechill then you will do all work for no reason
19:38 willthechill they will never accept your patches
19:38 rurban we will use it probably
19:38 rurban they will not do, they failed already
19:38 willthechill no then your fork will diverge from the main Perl and cause even more problems than we have now
19:38 willthechill we need to leave pumpkin alone
19:38 willthechill no forking, no fighting, just leave it alone
19:38 rurban no diverge. just the GV api probably
19:38 willthechill we need to create our own fixes from the outside
19:39 willthechill yes any fork will diverge, because everyone else will be working on the main branch and you will not ever be able to keep up with all their updates in your own branch
19:39 rurban pumpiun is big fail. he should control his devs and the roadmap, not the other way round.
19:39 rurban merges can be automated
19:39 willthechill hmmm
19:40 willthechill okay so maybe it would work if you could automate merges
19:40 rurban their track record is spectacularly bad, worse than the histories worst sw projects, such as isc bind or openssl.
19:40 willthechill STILL that would just be adding 1 more iron to the fire, when we have not even finished ONE iron to begin with!
19:40 willthechill we need to finish RPerl and P2 and Potion and B::CC first!!!
19:40 willthechill yes I am more than well aware of the problems with pumpkin Perl 5
19:40 willthechill I WROTE A WHOLE STORY ABOUT IT
19:40 rurban but they have a stupidified happy community. we are so happy (but still call the critics assholes)
19:41 willthechill http://rperl.org/the_book_of_rperl.html
19:41 rurban oh yes
19:41 rurban potion is almost finished. just threading is needed
19:41 willthechill let them be stupid and happy with pumpkin for now, we need to finish our own problems before attending to theirs
19:42 rurban p2 needs a library which is trivial, and the parser additions which are hard. esp. a gc safe parser
19:42 rurban well, I need to run/support a big company. bigger than booking.com
19:43 rurban we need a proper perl, not incapable devs gone rogue and completely insane
19:43 rurban we already had that with parrot before
19:43 rurban see where we are now. 5 years behind on the perl6 schedule
19:44 rurban and perl5 is unusable starting with 5.16
19:44 rurban I thought of getting a usable perl5 with 5.22, but not so. and 5.24 will be far away i fear.
19:45 rurban but on the other hand we ran with a stable and fast 5.6.2 for 15 years
19:45 rurban so staying with a maint-5.14 fork is fine (calling it 5.14.5)
19:46 rurban signatures are unusable anyway, cpan modules for signatures are much better
19:47 rurban and I'm waiting for them to get to sanity for several years already. I wrote detailled roadmaps and pathces already with no chance to be discussed and applied properly.
19:48 rurban typed scalars, arrays, readonly hashes+arrays for compile-time method optimizations, sized arrays.
19:48 rurban ...
19:48 basiliscos1 joined #perl11
19:50 rurban perfect hashes coming soon, and fast+small unicode tables
19:50 rurban nothing will break backcompat of course.
19:50 rurban just extensions to allow a faster perl and less memory
19:52 rurban they are only interested and capable of max 2% improvements, with some stupidies making it 10% slower (small strings, cow, hashes)
19:52 rurban I was talking about 50-200% better, not 2%
19:53 rurban similar to parrot vs moar
19:53 rurban or perl5 vs p2/rperl
19:54 rurban or gperl, which is in the rperl range
19:54 rurban tperl or cperl might be a good names also maybe
20:04 bulk88 " I wrote a big list of problems to be addressed but didn't published as it was too angry and personal" can I rewrite as my post? I've been planing a response to the ML but was away from a PC for last fri-sunday, mon/tue dealing with back log, today no time for code either.
20:07 willthechill rurban: why do you say 5.18 and 5.20 are unusable or unstable???
20:09 bulk88 also IMO jhi treats blead as his personal branch, and is the #1 breaker of blead for the last 2 months and that screws up bisecting
20:11 rurban 5.16 brought binary syms, almost fixed with 5.20 finally
20:12 rurban 5.18 broke hashes and cow (short strings)
20:12 bulk88 binary syms???? what r those?
20:13 rurban perl5 calls it unicode symbols
20:14 rurban but they support a embedded \0, and need a seperate len arg. so it's a binary blob
20:14 rurban and it needed several releases to support tthat in all the code places which treat symbols. most xs authors don't that, and dont; support it neither
20:14 rurban the whole GV api changed
20:15 rurban and it was never announced as such (just I cried foul)
20:15 bulk88 syber's method call patches?
20:15 rurban esp. when I found several use-after-free bugs with asan which easily exploit such binary payloads, DBI e.g.
20:16 rurban no, hugmeir's unicode symbols
20:16 rurban 5.15.6 or so
20:17 rurban syber's hash just has 32bit probnlems, and the hash function should be optimized to the limited pointer range they use
20:17 rurban intmaps are different than string maps, esp. with such a limited range
20:17 bulk88 if anyone is interested, here is a little war I won https://rt.perl.org/Public/Bug/Display.html?id=15667#txn-1309657 over 4 bytes in a struct
20:17 rurban but this is optimization, not security
20:18 willthechill well I use v5.18.2 because it ships in Ubuntu, no problems with RPerl so far
20:18 rurban yes, I fully support you there, a whole global flags in a struct for this tiny feature is nonsense.
20:18 bulk88 http://www.nntp.perl.org/group/perl.perl5.porters/2014/09/msg220432.html and the ML has no comment on SV struct's bad design and every other modern JIT/bytecode trick that everyone else uses
20:18 rurban it can be trivially solved without a global word in every GV
20:19 rurban you can only fix this with a better patch removing this nonsense. as I did with my bi-polar readonly removal patch
20:20 rurban i perlbugged it, but apparently the moderators killed it. which is a major scandal
20:20 rurban they did not only punished father to do such stupid things directly in blead, niot in a branch., they also punished teh critics, and refused to discuss and accept patches to fix that.
20:21 rurban well, only one critic here.
20:21 rurban but I appreciate your critic on the temp GV nonsense
20:21 rurban esp. since now also davem gone rogue
20:23 bulk88 http://perl5.git.perl.org/perl.git/commitdiff/522c63f2ecc9e36821528bb96cb98f39509068fb happened because I brought it up in #p5p after you were told to stop using RT
20:24 bulk88 since Linda Walsh used RT for months after she was banned from the ML
20:24 bulk88 but there was no policy at the time, so nobody said no RT
20:25 rurban interesting. so why didn't they ban rjbs for his actions then?
20:25 rurban he is the single worst nightmare for perl5 right now. letting his devs go rogue and silence code reviews and critics
20:26 rurban this is entirely unacceptable. big tunnel vision
20:26 bulk88 rjbs doesnt know C, and doesn't question internals choices
20:26 rurban they destroyed perl5 too much already, and everybody else is aware of that.
20:26 bulk88 also the p5p comitters seem to negotiate privatly over contraversy before comitting it, RFCs are rare
20:27 rurban but he is aware that perl5 is being actively destroyed, and he did his share in his decisions also
20:28 rurban smartmatch, given/when, signatures mop should be topics even he should understand
20:28 rurban I mean, he is one of the worst mid level developer I ever saw, but still.
20:28 bulk88 i dont use them, they aren't 5.8, and I have no comment
20:29 rurban dzil e.g.
20:29 rurban I posted statistical prove that his code is the worst, even worse than Adam Kennedies Module::Build mess
20:30 rurban or Moose, which actually does something useful
20:31 bulk88 rjbs is better than an iron fisted grey haired dictator that closes tickets and tells reporters to choke on it
20:33 bulk88 hands off management IMO is better than a "shared source" project that takes no public contributions of code at all
20:33 willthechill I stay out of p5p and pumpkin issues in general, too much of a clusterfunk and flamefest
20:33 rurban I don't care that much anymore, since there's now not a single capable developer for the internals left
20:33 rurban even the last one, davem gone rogue now, supporting all this nonsense.
20:34 bulk88 nonsense is specifically what issue?
20:34 willthechill our long-term goals can be to replace pumpkin 5 with a fully-backward-compatible suite such as P2 plus RPerl plus B::CC
20:34 rurban so perl5 is doomed. many cpan devs would be far better than the current committers.
20:34 bulk88 "since there's now not a single capable developer for the internals left" do I count ;)?
20:34 bulk88 jkjk
20:35 rurban demerqh having some psycholocal problems lately or such, nick clark gone away thanksfully
20:35 rurban bulk88, you are not committer
20:35 travis-ci joined #perl11
20:35 travis-ci [travis-ci] perl11/potion#780 (getops - cff1790 : Reini Urban): The build passed.
20:35 travis-ci [travis-ci] Change view : https://github.com/perl11/potion/compare/3ab2ddb5de3b...cff1790e9780
20:35 travis-ci [travis-ci] Build details : http://travis-ci.org/perl11/potion/builds/36295687
20:35 travis-ci left #perl11
20:35 rurban they make they rules, since they don;'t have to follow the rules.
20:36 rurban and rjbs just defending them, no matter what
20:36 rurban with technical insight he has no chance
20:36 rurban without
20:37 bulk88 you said capable developer, not capable committer
20:37 bulk88 ;)
20:37 rurban and shooting the messengers is rarely good, as they will have to go public then, which is 1000x worse than keeping it on the list
20:37 rurban oh, sorry.
20:37 rurban I meant committers, where only 3 of them are active developers.
20:38 rurban if dmq makes a nonsense change he at least pushes it some branch. if you publish a patch, it will be at least discussed somewhere
20:38 bulk88 ive never met demerphq IRL, he seems to be getting old and cranky, I dont think its pscological
20:38 rurban he said he is having personal problems
20:39 bulk88 continous integration would be a benefit for the p5p repo
20:39 rurban not reading and not understanding is a severe condition
20:39 rurban I called it "trolling", which is now accepted behavior on p5p
20:39 rurban even davem developed into a troll lately
20:41 rurban ok, have to leave now, offline for a few days
20:41 rurban potion work only
20:41 rurban one tiny jit problem remaining
20:44 rurban "Repeated unacceptable behavior" by people such as father or dmq do not get punished. not even warned. I'm trying to limit the damage a bit, but it's almost a find against windmills
20:45 rurban left #perl11
22:30 rurban joined #perl11
23:13 basiliscos joined #perl11

| Channels | #perl11 index | Today | | Search | Google Search | Plain-Text | summary