because the name "MJIT" is an internal code name, it's inconsistent with
--jit while they are related to each other, and I want to discourage future
JIT implementation-specific (e.g. MJIT-specific) APIs by this rename.
[Feature #17490]
getaddrinfo_a() gets stuck after fork().
To avoid this, we need 1 second sleep to wait for internal
worker threads of getaddrinfo_a() to be finished, but that is unacceptable.
[Bug #17220] [Feature #17134] [Feature #17187]
We need stop worker threads in getaddrinfo_a() before fork().
This change adds a hook before fork() that cancel all outstanding requests
and wait for all ongoing requests. Then, it waits for all worker
threads to be finished.
Fixes [Bug #17220]
Receiving UnixSocket works fine if you don't provide a mode, and
I think it is reasonable to expect that you should not provide
a mode if klass.for_fd would not accept a mode.
Fixes [Bug #11778]
Before, Socket.getaddrinfo was using a blocking getaddrinfo(3) call.
That didn't allow to wrap it into Timeout.timeout or interrupt the thread in any way.
Combined with the default 10 sec resolv timeout on many Unix systems, this can
have a very noticeable effect on production Ruby apps being not
resilient to DNS outages and timing out name resolution, and being unable to fail fast even
with Timeout.timeout.
Since we already have support for getaddrinfo_a(3), the async version
of getaddrinfo, we should be able to make Socket.getaddrinfo leverage that
when getaddrinfo_a version is available in the system (hence #ifdef
HAVE_GETADDRINFO_A).
Related tickets:
https://bugs.ruby-lang.org/issues/16476https://bugs.ruby-lang.org/issues/16381https://bugs.ruby-lang.org/issues/14997
Addrinfo.getaddrinfo and .foreach now accepts :timeout in seconds as
a keyword argument. If getaddrinfo_a(3) is available, the timeout will be
applied for name resolution. Otherwise, it will be ignored.
Socket.tcp accepts :resolv_timeout to use this feature.
This commit is retry of 6382f5cc91.
Test was failed on Solaris machines which don't have "http" in
/etc/services. In this commit, use "ssh" instead.
Addrinfo.getaddrinfo and .foreach now accepts :timeout in seconds as
a keyword argument. If getaddrinfo_a(3) is available, the timeout will be
applied for name resolution. Otherwise, it will be ignored.
Socket.tcp accepts :resolv_timeout to use this feature.
There seems to be a compatibility problems with Rails +
Rack::Deflater; so we revert this incompatibility.
This effectively reverts r65922; but keeps the bugfixes to
better support non-blocking sockets and pipes for future use.
[Bug #15356] [Bug #14968]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@66093 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
All normal Ruby IO methods (IO#read, IO#gets, IO#write, ...) are
all capable of appearing to be "blocking" when presented with a
file description with the O_NONBLOCK flag set; so there is
little risk of incompatibility within Ruby-using programs.
The biggest compatibility risk is when spawning external
programs. As a result, stdin, stdout, and stderr are now always
made blocking before exec-family calls.
This change will make an event-oriented MJIT usable if it is
waiting on pipes on POSIX_like platforms.
It is ALSO necessary to take advantage of (proposed lightweight
concurrency (aka "auto-Fiber") or any similar proposal for
network concurrency: https://bugs.ruby-lang.org/issues/13618
Named-pipe (FIFO) are NOT yet non-blocking by default since
they are rarely-used and may introduce compatibility problems
and extra syscall overhead for a common path.
Please revert this commit if there are problems and if I am afk
since I am afk a lot, lately.
[ruby-core:89950] [Bug #14968]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@65922 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Detecting the presence of constants in C headers is insufficient,
as a Linux kernel can be built with CONFIG_IPV6=n
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@65055 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
CI failures are still happening from these tests, but try
to break out of it earlier instead of holding up the job.
[Bug #14898]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64484 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
It looks like we need to retry test_timestampns in addition
to test_timestamp; so share some code while we're at it.
cf. http://ci.rvm.jp/results/trunk-test@frontier/1153126
[ruby-core:88104] [Bug #14898]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64157 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
I theorize there can be UDP packet loss even over loopback if
the kernel is under memory pressure. Retry sending periodically
until recvmsg succeeds.
i[ruby-core:87842] [Bug #14898]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@63872 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* The warnings are shown by Thread.report_on_exception defaulting to
true. [Feature #14143] [ruby-core:83979]
* Improves tests by narrowing down the scope where an exception
is expected.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61188 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
IO#read_nonblock and IO#write_nonblock take into account
buffered data, so the Linux-only BasicSocket#read_nonblock
and BasicSocket#write_nonblock methods must, too.
This bug was only introduced in r58400
("socket: avoid fcntl for read/write_nonblock on Linux")
and does not affect any stable release.
* ext/socket/basicsocket.c (rsock_init_basicsocket):
* ext/socket/init.c (rsock_s_recvfrom_nonblock):
* ext/socket/init.c (rsock_init_socket_init):
* ext/socket/lib/socket.rb (def read_nonblock):
* ext/socket/lib/socket.rb (def write_nonblock):
* ext/socket/rubysocket.h (static inline void rsock_maybe_wait_fd):
* test/socket/test_basicsocket.rb (def test_read_write_nonblock):
[Feature #13362]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@60496 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Binding to a potentially public IP in a test can cause problems
if hit by a random port scanner or something...
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58725 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
On some systems with slower local sockets, :wait_readable may
happen and we should wait on it to drain the socket.
This is a possible fix for https://bugs.ruby-lang.org/issues/13491
* test/socket/test_basicsocket.rb (test_read_write_nonblock):
handle :wait_readable on read_nonblock
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58437 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
On platforms where MSG_DONTWAIT works reliably on all sockets
(so far, I know of Linux), we can avoid fcntl syscalls and
implement IO#write_nonblock and IO#read_nonblock in terms of the
socket-specific send and recv family of syscalls.
This avoids side effects on the socket, and also encourages
generic code to be written in cases where IO wrappers like
OpenSSL::SSL::SSLSocket are used.
Perhaps in the future, side-effect-free non-blocking I/O can
be standard on all files and OSes: https://cr.yp.to/unix/nonblock.html
* ext/socket/lib/socket.rb (read_nonblock, write_nonblock):
Linux-specific wrapper without side effects
[ruby-core:80780] [Feature #13362]
* test/socket/test_basicsocket.rb (test_read_write_nonblock):
new test
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@58400 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
