[#40602] [ruby-trunk - Bug #5532][Open] Compile problem for bigdecimal on cygwin — Martin Dürst <duerst@...>
14 messages
2011/11/01
[#40646] [ruby-trunk - Bug #5541][Open] Better configure error message when llvm-gcc is the default compiler — Eric Hodel <[email protected]>
10 messages
2011/11/01
[#41137] [ruby-trunk - Bug #5541][Open] Better configure error message when llvm-gcc is the default compiler
— Motohiro KOSAKI <kosaki.motohiro@...>
2011/11/18
[#40647] [ruby-trunk - Bug #5542][Open] Ruby 1.9.3-p0 changed arity on default initialization method — Joshua Ballanco <jballanc@...>
7 messages
2011/11/01
[#40682] [ruby-trunk - Bug #5542] Ruby 1.9.3-p0 changed arity on default initialization method
— Joshua Ballanco <jballanc@...>
2011/11/02
[#40648] [ruby-trunk - Feature #5543][Open] rb_thread_blocking_region() API is poorly designed — Christopher Huff <cjameshuff@...>
14 messages
2011/11/01
[#40662] Relevant methods not appearing in RDoc — Andrew Grimm <andrew.j.grimm@...>
The method Module.private_constant isn't appearing in
6 messages
2011/11/01
[#40663] Re: Relevant methods not appearing in RDoc
— Eric Hodel <[email protected]>
2011/11/01
On Nov 1, 2011, at 3:45 PM, Andrew Grimm wrote:
[#40664] Re: Relevant methods not appearing in RDoc
— Andrew Grimm <andrew.j.grimm@...>
2011/11/01
On Wed, Nov 2, 2011 at 9:55 AM, Eric Hodel <[email protected]> wrote:
[#40684] [ruby-trunk - Feature #5555][Open] rename #include? to #includes? — Alexey Muranov <muranov@...>
20 messages
2011/11/02
[#40966] [ruby-trunk - Feature #5555] rename #include? to #includes?
— Andrew Grimm <andrew.j.grimm@...>
2011/11/12
[#43765] [ruby-trunk - Feature #5555][Rejected] rename #include? to #includes?
— "matz (Yukihiro Matsumoto)" <matz@...>
2012/03/27
[#44592] Re: [ruby-trunk - Feature #5555][Rejected] rename #include? to #includes?
— Roger Pack <rogerdpack2@...>
2012/04/24
> The basic naming for methods in standard class libraries are:
[#44602] Re: [ruby-trunk - Feature #5555][Rejected] rename #include? to #includes?
— Nikolai Weibull <now@...>
2012/04/24
On Tue, Apr 24, 2012 at 15:47, Roger Pack <[email protected]> wrote:
[#44603] Re: [ruby-trunk - Feature #5555][Rejected] rename #include? to #includes?
— Roger Pack <rogerdpack2@...>
2012/04/24
>> It always confuses me how that one defies the rule.
[#40688] [ruby-trunk - Bug #5556][Open] SIGHUP no longer ignored when sent to process group from a subprocess — Brian Ford <brixen@...>
12 messages
2011/11/02
[#40706] [ruby-trunk - Feature #5562][Open] Improvement of Windows IO performance — Hiroshi Shirosaki <h.shirosaki@...>
39 messages
2011/11/03
[#40737] [ruby-trunk - Bug #5570][Open] Encoding of environment variables on Windows — Nikolai Weibull <now@...>
11 messages
2011/11/04
[#43237] [ruby-trunk - Bug #5570][Feedback] Encoding of environment variables on Windows
— Nobuyoshi Nakada <nobu@...>
2012/03/11
[#43289] Re: [ruby-trunk - Bug #5570][Feedback] Encoding of environment variables on Windows
— "U.Nakamura" <usa@...>
2012/03/15
Hello,
[#40748] Proposal for sustainable branch maintenance — "Yuki Sonoda (Yugui)" <yugui@...>
-----BEGIN PGP SIGNED MESSAGE-----
14 messages
2011/11/05
[#40753] Re: Proposal for sustainable branch maintenance
— Trans <transfire@...>
2011/11/05
Why maintain 1.9.1 at all? I don't see the benefit. People need to
[#40813] Re: Proposal for sustainable branch maintenance
— "U.Nakamura" <usa@...>
2011/11/07
Hello,
[#40751] [ruby-trunk - Feature #5574][Open] Make arrays comparable — Arnau Sanchez <rbarnau@...>
8 messages
2011/11/05
[#40770] [ruby-trunk - Feature #5578][Open] Embedded YAML for Ruby 2.0 — Thomas Sawyer <transfire@...>
17 messages
2011/11/06
[#40816] [ruby-trunk - Feature #5578] Embedded YAML for Ruby 2.0
— Kurt Stephens <ks.ruby@...>
2011/11/07
[#40806] [ruby-trunk - Feature #5583][Open] Optionally typing — Yasushi ANDO <andyjpn@...>
21 messages
2011/11/07
[#40824] [ruby-trunk - Feature #5588][Open] add negation flag (v) to Regexp — Suraj Kurapati <sunaku@...>
38 messages
2011/11/08
[#40840] A quiz: Re: Second patch attempt for PATH_MAX fix in ruby1.9.1- 1.9.3~rc1-3 — Svante Signell <svante.signell@...>
Hello,
4 messages
2011/11/08
[#40845] [Ruby 1.8 - Bug #5599][Open] YAML.load_documents — Thomas Sawyer <transfire@...>
5 messages
2011/11/08
[#40865] IO.copy_stream creates files with restrictive permissions — Eric Wong <normalperson@...>
I'm not sure if this is a bug or intended as spec.
16 messages
2011/11/09
[#41151] Re: IO.copy_stream creates files with restrictive permissions
— Tanaka Akira <akr@...>
2011/11/19
2011/11/9 Eric Wong <[email protected]>:
[#41166] Re: IO.copy_stream creates files with restrictive permissions
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/20
>> I noticed when a file name argument is passed to the IO.copy_stream, the
[#41168] Re: IO.copy_stream creates files with restrictive permissions
— Clifford Heath <clifford.heath@...>
2011/11/20
On 20/11/2011, at 5:09 PM, KOSAKI Motohiro wrote:
[#41176] Re: IO.copy_stream creates files with restrictive permissions
— Tanaka Akira <akr@...>
2011/11/21
2011/11/20 Clifford Heath <[email protected]>:
[#41180] Re: IO.copy_stream creates files with restrictive permissions
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/21
>> I think documentation is the wrong answer. The security defects are not caused
[#41181] Re: IO.copy_stream creates files with restrictive permissions
— Tanaka Akira <akr@...>
2011/11/21
2011/11/22 KOSAKI Motohiro <[email protected]>:
[#41185] Re: IO.copy_stream creates files with restrictive permissions
— Clifford Heath <clifford.heath@...>
2011/11/22
On 22/11/2011, at 9:39 AM, Tanaka Akira wrote:
[#41196] Re: IO.copy_stream creates files with restrictive permissions
— Tanaka Akira <akr@...>
2011/11/22
2011/11/22 Clifford Heath <[email protected]>:
[#41219] Re: IO.copy_stream creates files with restrictive permissions
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/22
>> The umask that almost every Unix distribution has always had in /etc/profile - which is 022.
[#40867] Question regarding Ruby 2.0 backwards compatibility — Martin Bo煬et <martin.bosslet@...>
If I recall it correctly, the goal for 2.0 was to stay API-compatible
3 messages
2011/11/09
[#40898] [ruby-trunk - Feature #5605][Open] [PATCH] net/http: use IO.copy_stream for requests using body_stream — Eric Wong <normalperson@...>
7 messages
2011/11/10
[#40908] [ruby-trunk - Feature #5607][Open] Inconsistent reaction in Range of String — Yen-Nan Lin <redmine@...>
15 messages
2011/11/10
[#40941] [ruby-trunk - Feature #5617][Open] Allow install RubyGems into dediceted directory — Vit Ondruch <v.ondruch@...>
22 messages
2011/11/11
[#40943] [ruby-trunk - Bug #5619][Open] Segfault on OS X — Otto Hilska <otto@...>
9 messages
2011/11/11
[#40951] [Backport93 - Backport #5621][Open] Please backport thread-safe autoloading patch — Mike Perham <mperham@...>
25 messages
2011/11/12
[#40971] [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Mike Perham <mperham@...>
2011/11/12
[#40972] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Yehuda Katz <wycats@...>
2011/11/12
Unfortunately ruby-head has a deadlock in one of my go-to scenarios for
[#40976] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Hiroshi Nakamura <nahi@...>
2011/11/13
-----BEGIN PGP SIGNED MESSAGE-----
[#41128] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Charles Oliver Nutter <headius@...>
2011/11/18
On Sat, Nov 12, 2011 at 7:24 PM, Hiroshi Nakamura <[email protected]> wrote:
[#41129] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Hiroshi Nakamura <nahi@...>
2011/11/18
-----BEGIN PGP SIGNED MESSAGE-----
[#41142] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Charles Oliver Nutter <headius@...>
2011/11/18
On Fri, Nov 18, 2011 at 12:15 AM, Hiroshi Nakamura <[email protected]> wro=
[#41144] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Hiroshi Nakamura <nahi@...>
2011/11/18
(2011/11/19 6:31), Charles Oliver Nutter wrote:
[#41145] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Charles Oliver Nutter <headius@...>
2011/11/19
On Fri, Nov 18, 2011 at 5:24 PM, Hiroshi Nakamura <[email protected]> wrot=
[#41173] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Hiroshi Nakamura <nahi@...>
2011/11/21
(2011/11/19 9:08), Charles Oliver Nutter wrote:
[#40977] Re: [Backport93 - Backport #5621] Please backport thread-safe autoloading patch
— Hiroshi Nakamura <nahi@...>
2011/11/13
-----BEGIN PGP SIGNED MESSAGE-----
[#40982] [ruby-trunk - Bug #5625][Open] Remove profanity and pejoratives — Andrew Grimm <andrew.j.grimm@...>
30 messages
2011/11/13
[#40993] [ruby-trunk - Bug #5625][Third Party's Issue] Remove profanity and pejoratives
— Aaron Patterson <aaron@...>
2011/11/13
[#40996] Re: [ruby-trunk - Bug #5625][Third Party's Issue] Remove profanity and pejoratives
— Andrew Grimm <andrew.j.grimm@...>
2011/11/13
I was not aiming to protect children from the f-word. My intention was
[#41004] [ruby-trunk - Feature #5628][Open] Module#basename — Thomas Sawyer <transfire@...>
18 messages
2011/11/14
[#41024] [ruby-trunk - Feature #5632][Open] Attempt to open included class shades it instead. — Boris Stitnicky <boris@...>
12 messages
2011/11/14
[#41025] Proposal to add new methods: positive? negative? natural? — JosFrancisco Calvo Moreno <josefranciscocalvo@...>
Hi all!
11 messages
2011/11/14
[#41027] Re: Proposal to add new methods: positive? negative? natural?
— Jeremy Evans <code@...>
2011/11/14
On 11/15 12:58, Jos? Francisco Calvo Moreno wrote:
[#41031] Re: Proposal to add new methods: positive? negative? natural?
— JosFrancisco Calvo Moreno <josefranciscocalvo@...>
2011/11/14
Hi Jeremy,
[#41034] Re: Proposal to add new methods: positive? negative? natural?
— Chuck Remes <cremes.devlist@...>
2011/11/14
[#41036] Re: Proposal to add new methods: positive? negative? natural?
— JosFrancisco Calvo Moreno <josefranciscocalvo@...>
2011/11/14
Hi Chuck,
[#41037] Re: Proposal to add new methods: positive? negative? natural?
— Anurag Priyam <anurag08priyam@...>
2011/11/14
2011/11/14 Jos=E9 Francisco Calvo Moreno <[email protected]>:
[#41058] Re: Proposal to add new methods: positive? negative? natural?
— JosFrancisco Calvo Moreno <josefranciscocalvo@...>
2011/11/15
I don't have a deep understanding of ruby core base types but I want to
[#41064] Re: Proposal to add new methods: positive? negative? natural?
— "NARUSE, Yui" <naruse@...>
2011/11/15
On IEEE 754 Floating Point Numbers, there are positive 0.0 and negative 0.0=
[#41071] Re: Proposal to add new methods: positive? negative? natural?
— JosFrancisco Calvo Moreno <josefranciscocalvo@...>
2011/11/16
Do you suggest the following?
[#41038] [ruby-trunk - Bug #5634][Open] yield and binding — Thomas Sawyer <transfire@...>
17 messages
2011/11/14
[#50144] [ruby-trunk - Bug #5634] yield and binding
— "ko1 (Koichi Sasada)" <redmine@...>
2012/11/26
[#50210] [ruby-trunk - Bug #5634] yield and binding
— "ko1 (Koichi Sasada)" <redmine@...>
2012/11/27
[#41078] mkmf sets CFLAGS to contain flags that it shouldn't be setting — Nikolai Weibull <now@...>
Hi!
7 messages
2011/11/16
[#41091] Re: mkmf sets CFLAGS to contain flags that it shouldn't be setting
— Nobuyoshi Nakada <nobu@...>
2011/11/17
Hi,
[#41094] Re: mkmf sets CFLAGS to contain flags that it shouldn't be setting
— Nikolai Weibull <now@...>
2011/11/17
On Thu, Nov 17, 2011 at 07:24, Nobuyoshi Nakada <[email protected]> wrote:
[#41107] Re: mkmf sets CFLAGS to contain flags that it shouldn't be setting
— Nobuyoshi Nakada <nobu@...>
2011/11/17
Hi,
[#41108] Re: mkmf sets CFLAGS to contain flags that it shouldn't be setting
— Nikolai Weibull <now@...>
2011/11/17
On Thu, Nov 17, 2011 at 14:46, Nobuyoshi Nakada <[email protected]> wrote:
[#41079] Why doesn’t mkmf’s have_macro add a HAVE_X to $defs? — Nikolai Weibull <now@...>
Hi!
5 messages
2011/11/16
[#41092] Re: Why doesn’t mkmf’s have_macro add a HAVE_X to $defs?
— Nobuyoshi Nakada <nobu@...>
2011/11/17
Hi,
[#41086] [ruby-trunk - Feature #5644][Open] add Enumerable#exclude? antonym — Suraj Kurapati <sunaku@...>
14 messages
2011/11/17
[#41114] [Backport93 - Backport #5646][Open] Backport r33775 — Aaron Patterson <aaron@...>
9 messages
2011/11/17
[#41401] [Backport93 - Backport #5646][Closed] Backport r33775
— Aaron Patterson <aaron@...>
2011/11/30
[#41414] Re: [Backport93 - Backport #5646][Closed] Backport r33775
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/30
> Also, is there a script for doing backports? =A0It looks like the commit =
[#41420] Re: [Backport93 - Backport #5646][Closed] Backport r33775
— Aaron Patterson <tenderlove@...>
2011/12/01
On Thu, Dec 01, 2011 at 02:19:31AM +0900, KOSAKI Motohiro wrote:
[#41422] Re: [Backport93 - Backport #5646][Closed] Backport r33775
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/12/01
>> > Also, is there a script for doing backports? =A0It looks like the comm=
[#41116] Pathname['abc'] instead of Pathname('abc') global method as constructor alternative — "Alexander E. Fischer" <aef@...>
Hello,
3 messages
2011/11/17
[#41149] autoload will be dead — Yukihiro Matsumoto <matz@...>
Hi,
11 messages
2011/11/19
[#41150] Re: autoload will be dead
— Yehuda Katz <wycats@...>
2011/11/19
Does this also mean that we will not enhance const_missing to support
[#41160] [ruby-trunk - Bug #5650][Open] Add rb_enc_raise() to allow C extensions to raise errors with messages with correct encoding — Nikolai Weibull <now@...>
3 messages
2011/11/19
[#41161] conflict between 1.8.7 and 1.9.3 installations — Joel VanderWerf <joelvanderwerf@...>
6 messages
2011/11/19
[#41162] Re: conflict between 1.8.7 and 1.9.3 installations
— Nobuyoshi Nakada <nobu@...>
2011/11/20
Hi,
[#41164] Re: conflict between 1.8.7 and 1.9.3 installations
— Joel VanderWerf <joelvanderwerf@...>
2011/11/20
On 11/19/2011 04:55 PM, Nobuyoshi Nakada wrote:
[#41171] [ruby-trunk - Feature #5653] "I strongly discourage the use of autoload in any standard libraries" (Re: autoload will be dead) — Hiroshi Nakamura <nakahiro@...>
3 messages
2011/11/21
[#41175] [ruby-trunk - Feature #5654][Open] Introduce global lock to avoid concurrent require — Hiroshi Nakamura <nakahiro@...>
12 messages
2011/11/21
[#41186] [ruby-trunk - Bug #5657][Open] Constants in included modules aren't visible when defining classes with Class.new — Gary Bernhardt <gary.bernhardt@...>
5 messages
2011/11/22
[#41211] Availability of ruby/intern.h — Aaron Patterson <tenderlove@...>
Are functions accessible via ruby/intern.h considered to be "public"
4 messages
2011/11/22
[#41212] [ruby-trunk - Feature #5662][Open] inject-accumulate, or Haskell's mapAccum* — Edvard Majakari <edvard.majakari@...>
12 messages
2011/11/22
[#41213] [ruby-trunk - Bug #5663][Open] Combined map/select method — Yehuda Katz <wycats@...>
62 messages
2011/11/22
[#41217] [ruby-trunk - Bug #5663] Combined map/select method
— Rodrigo Rosenfeld Rosas <rr.rosas@...>
2011/11/22
[#41294] [ruby-trunk - Feature #5663] Combined map/select method
— Nobuyoshi Nakada <nobu@...>
2011/11/25
[#41391] [ruby-trunk - Feature #5663] Combined map/select method
— Rodrigo Rosenfeld Rosas <rr.rosas@...>
2011/11/29
[#45334] [ruby-trunk - Feature #5663] Combined map/select method
— "alexeymuranov (Alexey Muranov)" <redmine@...>
2012/05/31
[#41256] [ruby-trunk - Feature #2567] Net::HTTP does not handle encoding correctly — Yui NARUSE <naruse@...>
3 messages
2011/11/23
[#41262] [ruby-trunk - Feature #5666][Open] Make rb_path2class public — Aaron Patterson <aaron@...>
6 messages
2011/11/23
[#41302] [ruby-trunk - Feature #5673][Open] undef_method probably doesn't need to raise an error — Thomas Sawyer <transfire@...>
8 messages
2011/11/26
[#41314] [ruby-trunk - Bug #5675][Open] [mingw] static build fails due to socket extension build failure — Jon Forums <redmine@...>
7 messages
2011/11/26
[#41324] Re: [ruby-trunk - Bug #5675][Open] [mingw] static build fails due to socket extension build failure
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/27
> compiling ../../../ext/socket/socket.c
[#41330] Re: [ruby-trunk - Bug #5675][Open] [mingw] static build fails due to socket extension build failure
— Luis Lavena <luislavena@...>
2011/11/27
On Sun, Nov 27, 2011 at 4:11 AM, KOSAKI Motohiro
[#41332] Re: [ruby-trunk - Bug #5675][Open] [mingw] static build fails due to socket extension build failure
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/27
2011/11/27 Luis Lavena <[email protected]>:
[#41321] [Proposal] C API arrangement — Yugui <yugui@...>
Hi, mame-san
4 messages
2011/11/27
[#41336] Wrong encoding of Symbol — Andriy Tyurnikov <andriy.tyurnikov@...>
By default Symbols are encoded as US-ASCII,=20
10 messages
2011/11/27
[#41344] Re: Wrong encoding of Symbol
— Nobuyoshi Nakada <nobu@...>
2011/11/28
Hi,
[#41349] Re: Wrong encoding of Symbol
— Yukihiro Matsumoto <matz@...>
2011/11/28
Hi,
[#41353] Re: Wrong encoding of Symbol
— "Michel Demazure" <michel@...>
2011/11/28
Matz wrote :
[#41338] [ruby-trunk - Feature #5677][Open] IO C API — Martin Bosslet <Martin.Bosslet@...>
13 messages
2011/11/27
[#41340] [ruby-trunk - Feature #5678][Open] StringIO#to_str — Martin Bosslet <Martin.Bosslet@...>
8 messages
2011/11/28
[#41351] [ruby-trunk - Bug #5679][Open] Too many arguments for format warnings on mingw32 build — Heesob Park <phasis@...>
5 messages
2011/11/28
[#41368] Re: [ruby-trunk - Bug #5679][Open] Too many arguments for format warnings on mingw32 build
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/28
> I can see the following warnings during mingw32 build.
[#41370] [ruby-trunk - Bug #5681][Open] Conflcting types for rb_w32_inet_ntop caused by duplicate definition under MinGW — Luis Lavena <luislavena@...>
7 messages
2011/11/28
[#41373] Re: [ruby-trunk - Bug #5681][Open] Conflcting types for rb_w32_inet_ntop caused by duplicate definition under MinGW
— KOSAKI Motohiro <kosaki.motohiro@...>
2011/11/28
2011/11/28 Luis Lavena <[email protected]>:
[#41404] [ruby-trunk - Bug #5690][Open] Module#qualified_const_get — Yehuda Katz <wycats@...>
31 messages
2011/11/30
[#41406] Re: [ruby-trunk - Bug #5690][Open] Module#qualified_const_get
— Yukihiro Matsumoto <matz@...>
2011/11/30
Hi,
[#46001] [ruby-trunk - Bug #5690] Module#qualified_const_get
— "marcandre (Marc-Andre Lafortune)" <ruby-core@...>
2012/06/30
[#46645] [ruby-trunk - Feature #5690] Module#qualified_const_get
— "mame (Yusuke Endoh)" <mame@...>
2012/07/23
[ruby-core:40903] Re: [ruby-trunk - Feature #5605] [PATCH] net/http: use IO.copy_stream for requests using body_stream
From:
Eric Wong <normalperson@...>
Date:
2011-11-10 03:25:39 UTC
List:
ruby-core #40903
Eric Hodel <[email protected]> wrote: > For the chunked? half, would looping over > (({IO.copy_stream f, sock.io, 1024})) be faster? Yes, but its a rare case, needs to depend on a non-portable ioctl(), and isn't much faster. You can't blindly specify 1024 bytes because 'f' isn't guaranteed to have 1024 readable. But you need to write the "%x\r\n" size header before calling copy_stream, so I'm using IO#nread. Chunked uploads only make sense if: 1) f is a pipe/socket 2) you want to generate something like a Content-MD5 Trailer: on-the-fly and don't want to read f twice. In either case, sendfile() can't be used by IO.copy_stream. We can teach IO.copy_stream to splice() on Linux, maybe[1] I don't know if it's worth it to introduce the following patch, it only saves 200ms user CPU time on a 1G upload for an uncommon use case. I'm concerned about the portability of IO#nread. Even if IO#nread is available for the IO object, there's no guarantee the FIONREAD ioctl() is implemented (or implemented correctly) for the underlying file type given all the OSes Ruby supports. --- a/lib/net/http.rb +++ b/lib/net/http.rb @@ -21,6 +21,7 @@ require 'net/protocol' require 'uri' +require 'io/wait' autoload :OpenSSL, 'openssl' module Net #:nodoc: @@ -1965,9 +1966,25 @@ module Net #:nodoc: write_header sock, ver, path wait_for_continue sock, ver if sock.continue_timeout if chunked? - chunker = Chunker.new(sock) - IO.copy_stream(f, chunker) - chunker.finish + io = sock.io + if io.respond_to?(:nread) + seekable_p = f.stat.file? + f.wait + len = f.nread + while len > 0 + io.write("#{len.to_s(16)}\r\n") + IO.copy_stream(f, io, len) + f.pos += len if seekable_p + io.write("\r\n") + f.wait or break + len = f.nread + end + io.write("0\r\n\r\n") + else + chunker = Chunker.new(sock) + IO.copy_stream(f, chunker) + chunker.finish + end else # copy_stream can sendfile() to sock.io unless we use SSL. # If sock.io is an SSLSocket, copy_stream will hit SSL_write() --------------------------------------------------------------- Since chunked encoding is normally used for pipe/sockets of unknown length, I added the following to my test script to test: r, w = IO.pipe tmp.sysseek(0) pid = fork { IO.copy_stream(tmp.to_io, w) } w.close Net::HTTP.start(host, port) do |http| put = Net::HTTP::Put.new("/") put.body_stream = r put["Transfer-Encoding"] = "chunked" puts "with (pipe) Transfer-Encoding: chunked" bm = Benchmark.measure { res = http.request(put) } p [ res, res.body ] printf("utime=%0.3f stime=%0.3f\n", bm.utime, bm.stime) end Process.waitpid2(pid) > ------------------------ Results ------------------------------- > after > ----- > with Content-Length: 1073741825 > [#<Net::HTTPOK 200 OK readbody=true>, > "dcdd67a8f07b73796c0485890d48fa697966d09f\n"] > utime=0.010 stime=0.410 ^-- Oops, I realized stime=0.410 is high because I was on a cold cache, even with a sparse file. It should be 0.090 > with Transfer-Encoding: chunked > [#<Net::HTTPOK 200 OK readbody=true>, > "dcdd67a8f07b73796c0485890d48fa697966d09f\n"] > utime=0.320 stime=0.620 With a regular file, chunked becomes as fast as with Content-Length: --------------------------------------------------------------- with (file) Transfer-Encoding: chunked [#<Net::HTTPOK 200 OK readbody=true>, "dcdd67a8f07b73796c0485890d48fa697966d09f\n"] utime=0.010 stime=0.090 I think a better alternative would be to teach Net::HTTP to automatically set Content-Length if body_stream is a regular file. Chunked requests is are still an uncommon feature, I think. With a pipe, using IO.copy_stream calling Chunker --------------------------------------------------------------- with (pipe) Transfer-Encoding: chunked [#<Net::HTTPOK 200 OK readbody=true>, "dcdd67a8f07b73796c0485890d48fa697966d09f\n"] utime=0.350 stime=0.940 With a pipe, using IO#nread loop calling IO.copy_stream --------------------------------------------------------------- with (pipe) Transfer-Encoding: chunked [#<Net::HTTPOK 200 OK readbody=true>, "dcdd67a8f07b73796c0485890d48fa697966d09f\n"] utime=0.050 stime=0.940 [1] - We have to be careful since there are still old kernels with a buggy splice() implementation.