DevelopersMeeting20161011Japan

Date: 2016/10/11 (Tue)

Time: 14:00- 18:00 (JST)

Place: SFDC Japan, Tokyo

Attendees/参加者: ko1, naruse, akr, amatsuda, sorah, nakada, duerst (Martin Dürst), shyouhei, hsbt, mrkn

Language: mostly Japanese (sorry for non native Japanese speakers)

Agenda

• next meeting?

• Nov. 3-4 RWC
• Nov. 5-6 (camp?)@ Shimane
• Nov. 10-12 RubyConf
• conclusion: next meeing will be some point at later half of November.

• dev-camp?

• hotel, transport, and arrangement are OK (by ruby-assn)
• seems enough people are willing to attend.

• st report (ko1)

• ko1 benchmarked 3 (Vladimir’s, FunnyFalcon’s, current one) implementations

• against kaminari-pagenated rails scaffold
• using ab

• he found most time are consumed in looking up of small tables (size <= 64).
• for such small tables, it seems everything are on cache, algorithm difference was negilible.
• lets merge (any one of them) at the camp.

About 2.4 timeframe

• https://bugs.ruby-lang.org/projects/ruby-trunk/wiki/ReleaseEngineering24
• no showstopper
• what about the Rational renovation?

• please nudge _tad_

(From RubyKaigi admin) for Keynote speakers

• Please send us keynote speaker's Shinkansen receipt (shyouhei)

• naruse: no. I don’t have one.

Carry-over from previous meeting(s)

• [Misc #12283] Obsolete ChangeLog and commit message in Git-style (shyouhei) situation and progress?

• naruse: no move.
• shyouhei: I write git-style commit log.
• mrkn: me too.
• Martin: the problem is not commit log, but to auto-generate ChanegLog.
• naruse: I’ll do this on camp.

• [Feature #12695] File.expand_path should resolve ~/ using /etc/passwd when HOME is not set (shyouhei) conclusion?

• waiting for matz.

• [Feature #12733] Bundle bundler to ruby core (shyouhei)

• hsbt: no progress.
• hsbt: rubygems master bundles bundler by default.  It sheds some light into its darkness.
• nobu: it will be merged when rubygems side is OK.
• hsbt: carefully watching rubygems’ progress.

• [Feature #6647] Exceptions raised in threads should be logged (shyouhei) Look at [ruby-core:77259]

• akr: the proposal can be okay, but we currently do not implemnt thread GC mode.

• [Feature #12744] Add str.reverse_each_char and str.reverse_chars (shyouhei) conclusion?

• naruse: I agree this makes things faster a bit, but will it really be needed?
• Martin: array allocation is avoided in the patch, but each_char generates lots of strings anyways.

• [Bug #11195] Add "no_proxy" parameter to Net::HTTP.new (shyouhei) Roman Bigler volunteers.

• akr: +1 to refacor find_proxy, then use it from Net::HTTP.
• let’s request a patch, then mentor by naruse.

• [Bug #12594] The class does not inherit from a module the modules that were included after the inclusion (shyouhei)

• nobu: I’ll fix this someday. hopefully before 2.4?

From attendees

• [Bug #12649] DateTime method calls hang (shyouhei) seems like a real bug?

• assign nobu.

• [Bug #12652] For Dir.home encode passed user (shyouhei) seems like a real bug?

• assign naruse

• [Bug #12509] Using qsort_s in mingw-w64 causes failures (shyouhei) who to handle this?

• assign nobu

• [Bug #8996] pthread_mutex_lock EINVAL (shyouhei) seems like a real bug?

• assign ko1

• [Bug #12776] Flaky test case: TestThread#test_thread_name (shyouhei) seems like a real bug, or....?

• closed, lets backport.

• [Feature #12656] Expand short paths with File.expand_path (shyouhei)

• assign cruby-windows

• [Bug #7877] E::Lazy#with_index should be lazy (shyouhei)

• assign nobu

• [Feature #12664] Multiline pretty-printing of multiline strings (shyouhei)

• nudge akr

• [Feature #4897] Define Math::TAU and BigMath.TAU. The "true" circle constant, Tau=2*Pi. See http://tauday.com/ (shyouhei)

• mrkn: I suggest Math::TWOPI
• akr: C has PI_2, but this means π/2
• wating for matz (or time?)

• [Bug #12674] io/wait: not handling the case when the socket is closed before doing wait_readable/writable with timeout (shyouhei) might be a design issue?

• assign nobu

• [Bug #12678] No way to set a timeout for TLS handshake when using Net::SMTP (shyouhei) should who handle this?

• assign shugo

• [Bug #6783] Infinite loop in inspect, not overriding inspect, to_s, and no known circular references. Stepping into inspect in debugger locks it up with 100% CPU. (shyouhei)

• akr: I have a conception of ritcher inspection framework that passes-around contexts, but have not implemented yet.

• [Bug #10222] require_relative and require should be compatible with each other (shyouhei)

• akr: require_relative expands real path, while require doesn’t.
• naruse: if we expand realpath on require, we need disk access every time we call require (even when the library is already loaded).
• naruse: what about adding both real path and expanded path to $LOADED_FEATURES at once for require?
• shyouhei: I think for this shown use case b.rb shall not be called twice.

• [Bug #12705] yielding args to a lambda uses block/proc rather than lambda/method semantics (shyouhei)

• akr: I think it’s a bug.  They should be consistent.
• akr: someone broke here at 2.2.
• ko1: maybe me.

• [Bug #4537] Incorrectly creating private method via attr_accessor (shyouhei) cf. [Bug #9005]

• shyouhei: is this fixed already?
• nobu: seems not.
• shyouhei: updated ruby -v field.

• [Feature #12721] public_module_function (shyouhei)

• waiting for matz.

• [Feature #12732] An option to pass to Integer, Float, to return nil instead of raise an exception (shyouhei)

• akr: it seems there are needs of other conversion methods than to_i or Integer
• nobu: they just want to detect validity, not exceptions.
• Martin: what about Integer?()
• Martin: but Upcase?() is a new concept.
• waiting for matz.

• [Feature #12745] String#(g)sub(!) should pass a MatchData to the block, not a String (shyouhei)

• akr: naming matters.
• shyouhei: is that we need a new method name?  What about changing gsub behaviour?
• akr: it breaks compatibility.
• akira: what about passing extra second argument to the block, which happen to be a copy of $~
• akr: that is possible.
• nobu: what about defineing MatchData#to_str ?
• shyouhei: it might be possible to extend the passed string to define method that returns $~.
• akr: adding a new method is the most clean way if we can find good method name.

• [Feature #12747] Add TracePoint#callee_id (shyouhei)        

• assign ko1.

• [Feature #12755] optimize instruction sequence (shyouhei)

• waiting matz.

• [Bug #12780] BigDecimal#round returns different types depending on argument (shyouhei)

• assign mrkn

• [Feature #12790] Better inspect for stdlib classes (shyouhei)

• akr: I can’t read the output.
• mrkn: yes.  I’ll fix this.  But it’s not that simple.

• [Feature #12005] define rb_cFixnum and rb_Bignum again? (akr)

• dependency hell

• situation: we “tried” to delete them.  Main difficulty is “gem ‘json’, ‘~> 1.8.0’”
• hsbt: I pushed this change (remove JSON’s version) to Rails 4

• C-level constants: rb_cFixnum and rb_cBignum

• akr: resurrection of these constants does _not_ solve the issue.

• warnigs for Ruby-level constants: Fixnum and Bignum

• nobu: lets see RC1’s response.

• [Bug #12828] Check whether macro RUBY is okay for protecting ruby-specific onigumo extensions (duerst)

• nobu: RUBY is already defined at defines.h
• naruse: doesn’t anyone want to use vanilla onigumo someday?
• akr: is that possible?  Not that easy than it sounds.

From non-attendees

Write your name and your interest (what do you want to ask and to whom?) please.

• [Feature #12760] Optional block argument for itself (Victor Shepelev)

• “itself” doesn’t sound right
• nobu: if itself takes block it should behave like tap.