Before this change, _unrefActive would keep the unrefList sorted when
adding a new timer.
Because _unrefActive is called extremely frequently, this linear scan
(O(n) at worse) would make _unrefActive show high in the list of
contributors when profiling CPU usage.
This commit changes _unrefActive so that it doesn't try to keep the
unrefList sorted. The insertion thus happens in constant time.
However, when a timer expires, unrefTimeout has to go through the whole
unrefList because it's not ordered anymore.
It is usually not large enough to have a significant impact on
performance because:
- Most of the time, the timers will be removed before unrefTimeout is
called because their users (sockets mainly) cancel them when an I/O
operation takes place.
- If they're not, it means that some I/O took a long time to happen, and
the initiator of subsequents I/O operations that would add more timers
has to wait for them to complete.
With this change, _unrefActive does not show as a significant
contributor in CPU profiling reports anymore.
Fixes: https://github.com/joyent/node/issues/8160
Signed-off-by: Timothy J Fontaine <tjfontaine@gmail.com>
Conflicts:
lib/timers.js
Fixes: https://github.com/nodejs/node-convergence-archive/issues/23
Ref: https://github.com/nodejs/node/issues/268
PR-URL: https://github.com/nodejs/node/pull/2540
Reviewed-By: bnoordhuis - Ben Noordhuis <info@bnoordhuis.nl>
Both pointer arguments to memcmp are defined as non-null
and compiler optimizes upon that.
PR-URL: https://github.com/nodejs/node/pull/2544
Reviewed-By: trevnorris - Trevor Norris <trev.norris@gmail.com>
Reviewed-By: thefourtheye - Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
String concatenation in the assert messages has drastic impact on test
runtime. Removal of these messages is unlikely to affect debugging if
any breaking changes are made.
Previous time to run:
$ time ./iojs test/parallel/test-stringbytes-external.js
real 0m2.321s
user 0m2.256s
sys 0m0.092s
With fix:
$ time ./iojs test/parallel/test-stringbytes-external.js
real 0m0.518s
user 0m0.508s
sys 0m0.008s
PR-URL: https://github.com/nodejs/node/pull/2544
Reviewed-By: trevnorris - Trevor Norris <trev.norris@gmail.com>
Reviewed-By: thefourtheye - Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
The test was failing after adding 'use strict' because the windows CI
uses the autocrlf option of git which converts \r into \r\n on checkout.
Refactored the test to not read itself anymore and create a temp file on
the fly instead to avoid this line-ending issue.
PR-URL: https://github.com/nodejs/node/pull/2494
Reviewed-By: Joao Reis <reis@janeasystems.com>
This test was using fixturesDir to create temp files to test. This
patch replaces that with tmpDir and uses `assert` module to test.
Also, this test has been moved to `parallel`, from `sequential` mode.
PR-URL: https://github.com/nodejs/node/pull/2583
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Rich Trott <rtrott@gmail.com>
This reverts commit 6cd0e2664b.
This reverts commit 7a999a1376.
This reverts commit f337595441.
It turns out that on Windows, uv_pipe_getsockname() is a no-op for
client sockets. It slipped through testing because of a CI snafu.
PR-URL: https://github.com/nodejs/node/pull/2584
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
Complements the existing net.Socket#remoteFamily property.
PR-URL: https://github.com/nodejs/node/pull/956
Reviewed-By: Sam Roberts <vieuxtech@gmail.com>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
The implementation is a minor API change in that socket.address() now
returns a `{ address: '/path/to/socket' }` object, like it does for TCP
and UDP sockets. Before this commit, it returned `socket._pipeName`,
which is a string when present.
Change common.PIPE on Windows from '\\\\.\\pipe\\libuv-test' to
'\\\\?\\pipe\\libuv-test'. Windows converts the '.' to a '?' when
creating a named pipe, meaning that common.PIPE didn't match the
result from NtQueryInformationFile().
Fixes: https://github.com/nodejs/node/issues/954
PR-URL: https://github.com/nodejs/node/pull/956
Reviewed-By: Sam Roberts <vieuxtech@gmail.com>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Refer: https://github.com/nodejs/io.js/issues/1543
When this test fails, it leaves dead processes in the system. This
patch makes sure that the child processes exit first, in case of
errors.
PR-URL: https://github.com/nodejs/node/pull/2206
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
This poplulates the lists of flaky tests with tests that failed recently
in Jenkins.
PR-URL: https://github.com/nodejs/node/pull/2424
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: João Reis <reis@janeasystems.com>
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
Template configuration files for marking tests as flaky.
PR-URL: https://github.com/nodejs/node/pull/2424
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: João Reis <reis@janeasystems.com>
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
There's a bunch of stuff in test-child-process-spawnsync.js that seems
designed to test that it is in fact blocking/synchronous. However, that
code really just tests the OS sleep command. Change `sleep 1` to `sleep
0` and shave about one second off the test run.`
We check the return status to confirm the command is successful. The
tests in this file in general would not work if spawnSync() were
asynchronous. That includes this one, as a return status would not be
available if the command where asynchronous.
PR-URL: https://github.com/nodejs/node/pull/2542
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
The test had checked that a timer fired within a period after
spawnSync() returns. The result was a test that sometimes was
flaky.
Because there's no guarantee of how long a timer will take
before running, remove the check. There is a check that the
timer runs after spawnSync() so that is sufficient.
PR-URL: https://github.com/nodejs/node/pull/2535
Fixes: https://github.com/nodejs/node/issues/2470
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Make `make test-addons` part of the `make test-ci` target.
Use order-only prerequisites to make generating and building the add-ons
concurrency-safe when $JOBS > 1 and fudge the dependency on $(NODE_EXE)
so that add-ons are only rebuilt when needed instead of all the time.
PR-URL: https://github.com/nodejs/node/pull/2428
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Reviewed-By: Rod Vagg <rod@vagg.org>
This refactoring:
* eliminates the need for the external `curl` command
* speeds the test by running the two test requests simultaneously
* checks the type of error in the test that expects a failure
(previously, any error type would cause the test to pass)
PR-URL: https://github.com/nodejs/node/pull/2433
Reviewed-By: Brendan Ashworth <brendan.ashworth@me.com>
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
This commit replaces instances of io.js with Node.js, based on the
recent convergence. There are some remaining instances of io.js,
related to build and the installer.
Fixes: https://github.com/nodejs/node/issues/2361
PR-URL: https://github.com/nodejs/node/pull/2367
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: João Reis <reis@janeasystems.com>
This adds a new `--tls-cipher-list` command line switch
that can be used to override the built-in default cipher
list. The intent of this is to make it possible to enforce
an alternative default cipher list at the process level.
Overriding the default cipher list is still permitted at
the application level by changing the value of
`require('tls').DEFAULT_CIPHERS`.
As part of the change, the built in default list is moved
out of tls.js and into node_constants.h and node_constants.cc.
Two new constants are added to require('constants'):
* defaultCipherList (the active default cipher list)
* defaultCoreCipherList (the built-in default cipher list)
A test case and doc changes are included.
A new NODE_DEFINE_STRING_CONSTANT macro is also created in
node_internals.h
When node_constants is initialized, it will pick up either
the passed in command line switch or fallback to the default
built-in suite.
Within joyent/node, this change had originaly been wrapped
up with a number of other related commits involving the
removal of the RC4 cipher. This breaks out this isolated
change.
/cc @mhdawson, @misterdjules, @trevnorris, @indutny, @rvagg
Reviewed By: Ben Noordhuis <ben@strongloop.com>
PR-URL: https://github.com/nodejs/node/pull/2412
Refactored version of https://github.com/joyent/node/pull/25819
Removes integer keys (and keys starting with numbers) from
candidate list on repl tab complete. Refactored the originally
submitted change to simplify and ensure that the integer keys
do not show up on objects either.
Reviewed By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
PR-URL: https://github.com/nodejs/node/pull/2409
This resolvesjoyent/node#9272. `tlsSocket.getPeerCertificate` will
return an empty object when the peer does not provide a certificate,
but, prior to this, when the certificate is empty, `checkServerIdentity`
would throw because the `subject` wasn't present on the cert.
`checkServerIdentity` must return an error, not throw one, so this
returns an error when the cert is empty instead of throwing
a `TypeError`.
PR-URL: https://github.com/nodejs/node/pull/2343
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Reviewed-By: Shigeki Ohtsu <ohtsu@iij.ad.jp>
There is no way a line can be called after throwing an exception.
PR-URL: https://github.com/nodejs/node/pull/2289
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
Reviewed-By: Roman Reiss <me@silverwind.io>
Reviewed-By: Brendan Ashworth <brendan.ashworth@me.com>
If you have no history file written to disk, but the environment
variable set, `fs.readFileSync` will throw an ENOENT error,
but there's nothing to convert. The converter should ignore
ENOENT on that `fs.readFileSync` call.
Fixes: https://github.com/nodejs/node/issues/2449
PR-URL: https://github.com/nodejs/node/pull/2451
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
Previously 1000-1200ms, they're now (platform dependent) 50-100ms.
Improves test run time on my machine from 0m1.335s to 0m0.236s.
PR-URL: https://github.com/nodejs/node/pull/2429
Reviewed-By: Rich Trott <rtrott@gmail.com>
Previously the test had a massive timeout (3s!), reduce this to a
platform specific timeout of 50ms.
This test runs two servers at the same time in an attempt to compare
behaviour. I've added a check to make sure one event fires before the
other event, as is expected, but that is a possible race condition.
Improves test run time on my machine from 0m3.141s to 0m0.356s.
PR-URL: https://github.com/nodejs/node/pull/2429
Reviewed-By: Rich Trott <rtrott@gmail.com>
This includes the following changes:
- a more strict data check rather than a regex
- reduced number of annoying log calls
The most important of the changes is the annoying log calls, which
speeds up the test execution from about 0m1.130s to 0m0.481s on my
machine.
PR-URL: https://github.com/nodejs/node/pull/2429
Reviewed-By: Rich Trott <rtrott@gmail.com>
As per the discussion in #734, this patch deprecates the usage of
`EventEmitter.listenerCount` static function in the docs, and introduces
the `listenerCount` function in the prototype of `EventEmitter` itself.
PR-URL: https://github.com/nodejs/node/pull/2349
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Reviewed-By: Brian White <mscdex@mscdex.net>
v8::Isolate::GetCurrent() is not exactly deprecated at this point but
its use is strongly discouraged. Update the addon tests so they no
longer use it.
PR-URL: https://github.com/nodejs/node/pull/2427
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Add files in test/addon to the `make cpplint` rule and fix up existing
style issues. Tests scraped from doc/api/addon.md are filtered out
because those are predominantly for illustrative purposes.
PR-URL: https://github.com/nodejs/node/pull/2427
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
The test no longer waits about 5 seconds between callback invocations.
It now writes to the tmp directory rather than the fixtures directory.
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
PR-URL: https://github.com/nodejs/node/pull/2393
Errors might be injected into OpenSSL's error stack
without the return value of `PEM_read_bio_PrivateKey` being set to
`nullptr`. See the test of `test_bad_rsa_privkey.pem` for an
example.
PR-URL: https://github.com/nodejs/node/pull/2342
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Currently, v8 native deps must be built in order to run the log
processor on node profiling output. These scripts use node instead
of d8 to remove this dependency.
This change was originally proposed to the v8 team but since the
changes are not v8 specific, we have moved the proposal here. See:
https://codereview.chromium.org/1179173009/
PR-URL: https://github.com/nodejs/node/pull/2090
Reviewed-By: Ali Ijaz Sheikh <ofrobots@google.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
The exponent value was already in hex, but missing the 0x prefix which
could be confusing.
PR-URL: https://github.com/nodejs/io.js/pull/2320
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: James M Snell <jasnell@gmail.com>
v8 introduced the new flag `total_available_size` in version 4.4
and upwards. This flag is now available on `v8.getHeapStatistics`
with the name `total_available_size`. It contains the total
available heap size of v8.
Introduced with commit: v8/v8-git-mirror@0a1352a7
PR-URL: https://github.com/nodejs/io.js/pull/2348
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
According to docs, dropMembership() is automatically called
by the kernel when the socket is closed, and most apps will
never need to call it. It's called here as a sanity check
only so let's note that with a comment.
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
PR-URL: https://github.com/nodejs/io.js/pull/2062
Remove dead code paths that are created by assertions that will never
trigger. They may only trigger if either the `splitDeviceRe` or
`splitPathRe` regular expressions are modified. If at some point they
are modified, current unit tests will catch most of the resulting
errors and this commit adds extra tests to catch the remaining errors.
PR-URL: https://github.com/nodejs/io.js/pull/2282
Reviewed-By: Roman Reiss <me@silverwind.io>
This commit modifies util.inspect(obj) to additionally show the name of
the function that constructed the object. This often reveals useful
information about the object's prototype. In other words, instead of
> new Cls
{}
we have
> new Cls
Cls {}
This also works with exotic objects:
> class ArrayCls extends Array {}
> new ArrayCls(1, 2, 3)
ArrayCls [ 1, 2, 3 ]
The names of "trivial" constructors like Object and Array are not shown,
unless there is a mismatch between the object representation and the
prototype:
> Object.create([])
Array {}
This feature is inspired by browser devtools.
PR-URL: https://github.com/nodejs/io.js/pull/1935
Reviewed-By: Roman Reiss <me@silverwind.io>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
When TLS Session Ticket is renewed by server - no Certificate record is
to the client. We are prepared for empty certificate in this case, but
this relies on the session reuse check, which was implemented
incorrectly and was returning false when the TLS Session Ticket was
renewed.
Use session reuse check provided by OpenSSL instead.
Fix: https://github.com/nodejs/io.js/issues/2304
PR-URL: https://github.com/nodejs/io.js/pull/2312
Reviewed-By: Shigeki Ohtsu <ohtsu@iij.ad.jp>
- eliminate unnecessary intermediate process ("parent")
- children exit if runner dies unexpectedly (killed on a test timeout,
for example)
- use explicit messaging from children to parents to indicate when
worker is ready to accept http requests, rather than racing to see
whether runner will make request before worker is listening
PR-URL: https://github.com/nodejs/io.js/pull/1944
Reviewed-By: Johan Bergstrom <bugs@bergstroem.nu>
Reviewed-By: Ben Noordhuis <ben@strongloop.com>
Any time the connection state or the underlying handle itself changes,
the socket's name (aka, local address) can change.
To deal with this we need to reset the cached sockname any time we
set or unset the internal handle or an existing handle establishes a
connection.
PR-URL: https://github.com/nodejs/io.js/pull/2095
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
simple tests test-http-request-end.js, test-http-default-encoding.js
hangs in AIX. The root cause for both the failures is related to the
timing with which packets are sent between the client and server.
On the client side, one factor that affects the timing is Nagle's
algorithm. With Nagle enabled there may be a delay between two packets
as the stack may wait until either:
a. An acknowledgement for the first packet is received, or
b. 200 ms elapses.
before sending the second packet.
Similarly at the server side 2 sequential packages can be delivered to
the application either together or separatly.
On AIX we see that they are delivered separately to the server, while on
Linux delivered together. If we change the timing, for example disabling
Nagle on AIX we see the 2 packets delivered together and the tests pass.
In the test case simple/test-http-request-end.js, the client request
handler of the server receives and stores the data in a data callback,
closes the server in a request end callback, and writes to the client
and ends the response, in-line with the request receipt. An HTTP parser
module parses the incoming message, and invokes callback routines which
are registered for HTTP events (such as header, body, end etc.)
Because the termination sequence arrive in a separate packet, there is a
delay in parsing that message and identify that the client request ended
(and thereby invoke the request end call backhandler). Due to this delay,
the response close happens first, which in-turn destroys the server
socket leading to the fd and watcher removal from the uv loop abandoning
further events on this connection, and end call back never being called,
causing the reported hang. simple/test-http-default-encoding.js suffers
from the same problem.
Also, remove the timer logic from the test case. Test harness anyways
contain a timer which controls the individual tests so remove such
controls from the test case, as suggested by @tjfontaine
Reviewed-By: Michael Dawson <michael_dawson@ca.ibm.com>
PR-URL: https://github.com/joyent/node/pull/9432
PORT-FROM: joyent/node @ 13e1131406
PR-URL: https://github.com/nodejs/io.js/pull/2294
Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Julien Gilli
Karl Skomski
Trevor Norris
Rich Trott
Roman Reiss
Sakthipriyan Vairamani
Ben Noordhuis
Alexis Campailla
Fedor Indutny
Bradley Meck
João Reis
cjihrig
James M Snell
Brendan Ashworth
Mike Atkins
Michaël Zasso
Todd Kennedy
P.S.V.R
Matt Loring
muddletoes
Chad Johnston
Roman Klauke
Nathan Woltman
Christopher Monsanto
Sam Roberts
Ryan Graham
Gireesh Punathil
Jeremiah Senkpiel