| Date | Commit message (Collapse) |
|---|
|
Mainly this release reverts back to kgio 2.9.3 behavior by
supporting "autopush" functionality in the C extension.
Removing this feature in 2.10.0 was a monumental mistake(*) on
my part which caused a performance regression for ONE known user
of kgio on the deprecated (but supported) Rainbows! server.
This release does not affect known users outside of Rainbows!,
but we may have unknown users, too.
The attempt at a pure Ruby version of autopush had flaws and
regressions in real-world testing, so that was abandoned.
So yeah, kgio will be supported indefinitely even in the face
of internal changes to MRI, but I strongly discourage its use
for new projects.
Users who only support Ruby 2.3 (and later) are encouraged to
use `exception: false` with all *_nonblock methods.
4 changes since 2.10.0:
resurrect Kgio.autopush support in pure Ruby [REVERTED]
reinstate the original (and dangerous) autopush in C
TODO: update with Ruby 2.3 status
test: increase test data sizes to fill socket buffers
(*) - as was making this a separate project, at all
|
|
Some Debian build machines use insanely large socket
buffers, so increase our test buffers to large values to
force EAGAIN: https://bugs.debian.org/830353
In my experience, using setsockopt to clamp buffers to small sizes
can lead to unpredictable stalls, too, so it might not be good
to do in a test case, either.
|
|
Since we're on the cusp of Ruby 2.4, I guess 2.3 features
are stable.
|
|
The regression for existing users was unnacceptable and
completely poor judgement on my part. This change brings
us back to potentially not-future-compatible code which
will impose maintenance burdens on us in the face of
future Ruby changes.
But TODAY, it is the most performant option for folks who
need to use autopush.
Revert "resurrect Kgio.autopush support in pure Ruby"
and "remove autopush support and make it a no-op"
This reverts commits 64dc570f4b99f68b5ed792b36e7e8abc3df74927
and 4347980fa66115425fa8b765353c8b1bfe5dec24.
|
|
This avoids breaking compatibility for existing apps, but
is less performant (although safer and more resilient to future
changes in Ruby) than the previous C version.
|
|
The largely-unused autopush functionality is now a no-op.
Existing applications will probably not notice, and are
better off using MSG_MORE anyways if they're on Linux.
Our poll(2) wrapper now uses rb_hash_clear if available,
reducing global method cache dependence.
shortlog of changes since 2.9.3:
README: remove reference to Rainbows!
writev: update comment on preserving errno
poll: avoid rb_funcall for rb_hash_clear
minor doc updates
remove autopush support and make it a no-op
HACKING: update URL for Message-ID lookup
|
|
Prefer to use our own site since it is free of images+JavaScript
and may load faster if somebody is already reading bogomips.org
and has an open connection and hot DNS cache entry.
|
|
Maintaining global state in a library like kgio is ugly, and it
is not optimal from a performance standpoint compared to using
MSG_MORE.
TCP_CORK and TCP_NOPUSH require extra syscalls so it still offers
sub-optimal performance compared to MSG_MORE.
Instead, server developers should use MSG_MORE if their OS supports
it (and should add MSG_MORE to their OS if lacking it).
|
|
Ruby version changes, dropping Rubinius and MRI 1.8 support,
and clarifying a few other things.
"Linux kernel" coding style is probably a more accurate of what
we've always used as the style is probably better known than K&R
nowadays (and we do try to track Linux kernel changes (and
email-oriented development style) closely).
|
|
rb_funcall will always be slower, and currently hits the
global method cache in MRI.
|
|
Studying malloc implementations for a while, it's completely
reasonable for "free" to alter errno when making syscalls
to release/compact memory.
|
|
Rainbows! never took off and is mostly a stagnant project
at this point; so avoid confusing people with it.
|
|
Most notably there's a new RB_GC_GUARD to fix an occasional segfault at
load time for Ruby 1.8 users thanks to Petr Novodvorskiy for reporting
the issue.
There's been a new mailing list for a few months now at:
kgio-public@bogomips.org
No subscription is required, but if you wish to subscribe
(existing librelist subscribers must resubscribe)
kgio-public+subscribe@bogomips.org
There's also a bunch of code cleanups and packaging
cleanups. Our website is faster now thanks to olddoc:
http://bogomips.org/kgio/
shortlog:
test/lib_read_write: fix trywritev blocking test
new mailing list info: kgio-public@bogomips.org
relax license to allow LGPLv2.1 or later
gemspec: modernize for more recent rubygems
doc: switch to olddoc
remove references to rubyforge and ancient Rubies
TODO: updates for the future of kgio
tryopen: add RB_GC_GUARD for Ruby 1.8
LICENSE: prefer URL to FSF address
cleanup packaging harder
fixup -Wshorten-64-to-32 warnings
cleanup: avoid shadowing rb_str_subseq
cleanup: avoid shadowing rb_ary_subseq
pkg.mk: synchronize with my other projects
gemspec: remove invalid rdoc_options
gemspec: use SPDX license abbreviation
|
|
ref: http://spdx.org/licenses/
|
|
No longer needed for olddoc
|
|
We'll get EXTCONF_ARGS support as well as a muscle-memory-friendly
"make check" alias.
|
|
Define a MY_ARY_SUBSEQ macro which checks arg counts correctly and
avoids throwing off people who may potentially want to debug this
code.
|
|
Define a MY_STR_SUBSEQ macro which checks arg counts correctly and
avoids throwing off people who may potentially want to debug this
code.
|
|
These are emitted by clang on a Ubuntu Trusty amd64 system.
clang version 3.5-1ubuntu1 (trunk) (based on LLVM 3.5)
|
|
We don't need references to RubyForge or old images crap.
|
|
The FSF has and may move again physically, so the URL should be
more future-proof.
|
|
Calling rb_intern(RSTRING_PTR(err)) was unsafe as the `err'
value could be GC-ed inside rb_intern.
This only affected Ruby 1.8 as Errno.constants returns an
array of Symbols under 1.9+
Thanks to Petr Novodvorskiy for reporting the issue.
ref: <CAHGF7iJk9-wK5WbSiAgQaRye7y2RJttLBpykRBpnKkxFi=ASHg@mail.gmail.com>
Cc: Petr Novodvorskiy <nidd@skytap.com>
Cc: Lara Martin <lmartin@skytap.com>
Cc: Nick Astete <nastete@skytap.com>
|
|
The future of this project is to have no future.
|
|
|
|
wrongdoc was difficult to maintain because of the tidy-ffi
dependency and the HTML5 changes in Darkfish could not be
handled well by Tidy.
olddoc is superior as it generates leaner HTML which loads faster,
requires less scrolling and less processing power to render.
Aesthetic comparisons are subjective of course but completely
unimportant compared to speed and accessibility.
|
|
License field is commonly supported nowadays, and Rubyforge is dead.
Note: wrongdoc will be replaced with something more old-fashioned :)
|
|
In case I'm hit by a bus, the lesser evil is to allow the FSF
to update our license than to be stuck as LGPLv3-only in the
future.
|
|
You may (optionally) subscribe to the new mailing list at:
kgio-public+subscribe@bogomips.org
If your ISP prevents you from using port 25, port 587 (submission) is
open on bogomips.org. You may use Tor if you do not wish to expose your
IP when using port 587.
Any HTML mail will be flagged as spam, so please do not send it and
waste storage and bandwidth on it. Basic, old-fashioned mailing list
conventions apply: no top-posting, trim replies, attribute your quotes,
short signatures, etc.
This existing list will operate in public-inbox "hijack mode":
the new address is subscribed to the old librelist, so any messages
sent to the librelist are automatically archived to the new
public-inbox. Of course, not much happens here, anyways, so
nobody notices :)
Background:
Over the years, I've come to disagree with the subscription-required
posting policy of librelist. Combined with disabling of rsync archives,
one of my main reasons for choosing librelist back in the day are gone.
Lately, I've also been working on public-inbox, http://public-inbox.org/
an "archives-first" approach to mailing lists using git.
ssoma (git) archives are available at: git://bogomips.org/kgio-public/
The ssoma format is described at http://ssoma.public-inbox.org/
HTML archives are going to http://bogomips.org/kgio-public/
All archives imported from gmane, and posts to this librelist
also go there.
|
|
penultimate return value may be a subarray for writev-related
routines, not a string.
Caught by a pre-upload check by Christian Hofstaedtler
on the debian-ruby list for kgio 2.9.1
Tested-by: Christian Hofstaedtler <zeha@debian.org>
Cc: Hleb Valoshka <375gnu@gmail.com>
Cc: <debian-ruby@lists.debian.org>
Cc: Yura Sokolov <funny.falcon@gmail.com>
|
|
This release is for compatibility with future releases of mainline ruby,
as rb_thread_blocking_region is removed in r44955 of ruby trunk
This also avoids deprecation warnings fo rb_thread_blocking_region
2.0 and 2.1.
|
|
We've finally figured out how to avoid warnings when building on
C Ruby 2.x
|
|
It'll be OK to use rb_thread_call_without_gvl when
rb_thread_blocking_region is not detectable at all.
We still use rb_thread_blocking_region for Ruby 2.0-2.1 because
rb_thread_call_without_gvl was detectable in 1.9.3, but not
usable as an internal symbol.
ref: https://bugs.ruby-lang.org/issues/9502
|
|
Thanks to Christopher Rigor for this release.
Eric Wong (2):
only define and test kgio_syssend on 1.9+
various 1.8.7 fixes
|
|
Some errors in the code reorganization caused some compatibility
code to be dropped :x
|
|
Reported-by: Christopher Rigor <crigor@gmail.com>
|
|
This adds a new kgio_syssend method for Ruby 1.9+ which behaves
like BasicSocket#send, but will not raise exceptions on EAGAIN.
Eric Wong (12):
test_poll: remove race prone test_poll_EINTR_changed test
tryopen: remove RARRAY_PTR usage in init
read_write: remove the rest of RARRAY_PTR usage
my_writev: stylistic fixes
Rakefile: kill raa_update task
avoid downsizing casts
connect: constify RSTRING-related things
set RSTRING_MODIFIED where appropriate for Rubinius
split read_write.c into {read,write,writev}.c
add kgio_syssend method to wrap send(2)
write: correct check for various blocking regions
tryopen: additional debug information for bad Errno values
Hleb Valoshka (1):
Don't use deprecated api
|
|
In case users run into the same problem in the future.
ref: http://mid.gmane.org/20131227074129.GA6381@dcvr.yhbt.net
|
|
This will allow us to support past/future Rubies without
rb_thread_io_blocking_region but still have other ways
of releasing GVL.
|
|
This behaves like kgio_trywrite on GNU/Linux, but allows extra flags
to be specified. The main purpose of this is to support use of the
MSG_MORE flag on GNU/Linux.
|
|
This helps make our code more manageable, as well as isolating
RSTRING_MODIFIED usage to files which actually need it.
writev.c can eventually be fixed to avoid modifying RSTRING
pointers, but write.c can already benefit from const RSTRING_PTR
values right away.
|
|
Not yet tested, but it makes sense to do this.
|
|
This should enforce correctness with our non-use of RSTRING_MODIFIED
in this file, hopefully allowing Rubinius to optimize better.
|
|
It's always safe to cast a 32-bit value to a 64-bit value, but
compilers can warn going the other way (because VALUE is a pointer).
Tested with `ruby -v`
ruby 2.2.0dev (2014-01-19 trunk 44646) [x86_64-linux]
which enables -Wpointer-to-int-cast
|
|
Signed-off-by: Eric Wong <normalperson@yhbt.net>
|
|
RAA is dead
|
|
Rename 'str' to 'ary', as that's the whole point of supporting
writev... Also, long lines need to be wrapped at 80 columns
|
|
I have not benchmarked this, but this should not make a difference
as far as performance goes. This should also allow better
performance of better GCs in Ruby 2.1.0 and Rubinius.
|
|
This is trivially non-performance-critical, and can only
help with advanced GCs in Ruby 2.1.0 and Rubinius
|
|
We haven't figured out a way to reliably test this w/o races, so lets
just remove it for now and trust it works by reading the C code.
ref: <CAAB-KcnpvcG6=OZNsBmvv440OHfCWs6-eDD7L6oD=ziCRXPHLA@mail.gmail.com>
|
|
Improved error reporting for kgio_accept/kgio_tryaccept.
Minor size reduction throughout. There are also several
test case fixes for race conditions.
Thanks to Hleb Valoshka and the Debian project for all the
help with this release!
Eric Wong (7):
check syscall returns against < 0 instead of == -1
accept: more informative exception on unknown family
test_tryopen: skip EACCES test when euid == 0
test/lib_read_write: account for larger-than-normal pipes
test_poll: avoid potentially thread-unsafe test
test_poll: preserve original trap(:USR1) handler
test_poll: be less dependent on signal handler ordering
Hleb Valoshka (4):
Change prefix of temporary sockets to prevent races
Don't dump 20M in case of failure
Create own directory for every unix socket in unit tests
Close tempfile and unlink it immediately.
|