shyouhei: I think source_locations makes more sense than source_location.
akira: I started to confuse. Is it useful? We can already get Method#source_location.
[Feature #12063] KeyError#receiver and KeyError#name (shyouhei) status?
matz: I take KeyError#key.
[Bug #13397] #object_id should not be signed (shyouhei)
akr: isn’t it possible to somehow calculate so that we can avoid negative object_id? Because pointers are 8 byte aligned, while positive fixnums are of 262 space.
nobu: it’s true we can avoid negative numbers on some environment, but not always.
reject this specific issue and request for new issue that address the inspect helper.
[Feature #13396] Net::HTTP has no write timeout (shyouhei)
[Bug #13407] We have recv_nonblock but not send_nonblock... can we add it? (shyouhei)
[Feature #13395] Add a method to check for not nil (shyouhei)
[Feature #13434] better method definition in C API (shyouhei)
[Bug #13557] there's no way to pass backtrace locations as a massaged backtrace
[Bug #10290] segfault when calling a lambda recursively after rescuing SystemStackError
Fix confirmed?
nobu: the variant of setjmp seems to be the root cause. The proposed workaround works, but may impact performance. We have to measure before changing the default.
[Feature #10674] Net::HTTP retries idempotent requests once after a timeout, but its not configurable
[Bug #13542] MinGW trunk Builds - Summary of Issues
[Bug #13549] MinGW / Windows encoding - Two issues
[Bug #12159] Thread::Backtrace::Location#path returns absolute path for files loaded by require_relative (ko1)
ko1: I want to add Thread::Backtrace::Location#realpath to deprecate #absplute_path instead.
matz: no objection.
[Feature #13618] [PATCH] auto fiber schedule for rb_wait_for_single_fd and rb_waitpid (ko1)
[Bug #13576] File#to_path shall be deleted (shyouhei)
akr: did you try deleting this method?
shyouhei: no, but I should.
akr: we must start from adding warnings.
From non-attendees
Write your name and your interest (what do you want to ask and to whom?) please.
[Feature #13618] [PATCH] auto fiber schedule for rb_wait_for_single_fd and rb_waitpid (normalperson)
ko1: programmers transfer execution of fibers by hand. Eric proposes to introduce “auto”-Fiber which are OK to transfer each other on blocking situations. His implementation is great. rb_wait_for_single_fd is the ponit of switching. matz wanted this feature, too.
shyouhei: can auto Fibers be a subject to transfer or an object?
ko1: transfer happens only from an auto fiber to another and nothing else, AFAIK.
ko1: This is essentially a re-introduction of Ruby 1.8 threads (without time-tick). I’m concerning about thread difficulties like locking.
akr: is locking OK for auto Fibers?
ko1: Fibers has not needed locking because so far, Fiber.yield has been the only mechanism to transfer execution so it was very explicit. auto-Fiber changes that so there can be unintended transfers.
matz: do you want to reject this?
ko1: I don’t want to accept until we introduce a way to explicitly state where we can transfer, like async/await.
ko1: I think it’s NG to call routines that are not Fiber-aware from proposed auto-Fiber. I don’t think people can insert appropriate Thread.exclusive calls because if they could, we could never have experienced resource contentions.
matz: I think it’s hard to tag every and all of fiber-aware routines to be async. That sounds too big.
matz: show us your counter-proposal.
ko1: is it OK for you to accept Eric’s proposal as-is?
matz: I’m positive, but I know your concern so I won’t accept right now.
[Feature #12694] Want a String method to remove heading substr.
(sonots) I implemented with a name String#remove_prefix. If the name is ok, I want to merge it. Matz, could you decide? There were other candidates such as lchomp, trim_prefix. I dropped trim_prefix because trim is used to remove a list of characters rather than a substring in some languages, so I felt trim_prefix would raise confusion. I dropped lchomp because I just had never heard.
candidates discussed:
lchomp
chompl
guilotine
ltrim / rtrim
NG because PHP behaves differently
delete_prefix
amatsuda: remove_prefix sounds strange if it doesn’t change the receiver.
nobu: facets has lchomp.
akr: I don’t want chomp’s complex newline handling for this requested method.