When running the tests if `NODE_TEST_DIR` is set to a device different
than the location of the test files (where this repo is checked out),
then the parallel/test-fs-link.js test will fail with
`EXDEV: cross-device link not permitted`. The code works fine (and is in
fact throwing an error as desired) but the test fails.
This commit first creates the "source" file in the same directory as the
"destination" (where the hardlink will be created).
PR-URL: https://github.com/nodejs/node/pull/4861
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Rich Trott <rtrott@gmail.com>
`test-assert.js` redeclares a variable with `var`. This change converts
it to a `const` declaration and wraps it in a standalone block to scope
it to just the test that uses it.
PR-URL: https://github.com/nodejs/node/pull/4854
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Roman Reiss <me@silverwind.io>
Reviewed-By: targos - Michaël Zasso <mic.besace@gmail.com>
Previously, port was assumed to be a number and would cause an abort in
cares_wrap. This change throws a TypeError if port is not a number
before we actually hit C++.
Fixes: https://github.com/nodejs/node/issues/4837
PR-URL: https://github.com/nodejs/node/pull/4839
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Roman Klauke <romaaan.git@gmail.com>
Reviewed-By: Brian White <mscdex@mscdex.net>
This commit applies new arrow function linting rules across the
codebase. As it turns out, the only offenders were in the test
directory.
PR-URL: https://github.com/nodejs/node/pull/4813
Reviewed-By: Rod Vagg <rod@vagg.org>
Reviewed-By: Stephen Belanger <admin@stephenbelanger.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
Reviewed-By: Saúl Ibarra Corretgé <saghul@gmail.com>
Reviewed-By: Michaël Zasso <mic.besace@gmail.com>
Reviewed-By: Matteo Collina <matteo.collina@gmail.com>
Reviewed-By: Evan Lucas <evanlucas@me.com>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
Reviewed-By: Roman Reiss <me@silverwind.io>
test-net-settimeout is unnecessarily complex. This change simplifies it.
PR-URL: https://github.com/nodejs/node/pull/4799
Reviewed-By: Brian White <mscdex@mscdex.net>
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Event emitters support symbols as event names. The process object
assumes that the event name is a string, and examines the first
three characters to check for signals. This causes an exception
if the event name is a symbol. This commit ensures that the
event name is a string before trying to slice() it.
PR-URL: https://github.com/nodejs/node/pull/4798
Reviewed-By: Sam Roberts <vieuxtech@gmail.com>
Reviewed-By: Michaël Zasso <mic.besace@gmail.com>
Reviewed-By: Wyatt Preul <wpreul@gmail.com>
Timer race results in some flakiness on slower devices in CI. Remove
unneeded setTimeout() and replace booleans with common.mustCall().
PR-URL: https://github.com/nodejs/node/pull/4793
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
test-http-exit-delay has a flaky history. Examination of the bug it was
written to find suggests that the test should be removed.
* The test is trying to find a delay of up to 1 second, but the delay
may also be much smaller than that. So the test will not catch the bug
all the time. It is therefore flaky when the bug exists.
* Experience has shown that the test is flaky as well when the bug is
absent in the code because it can sometimes take slower devices (such as
the Raspberry Pi devices that we have in continuous integration) more
than the allowed one second to run. Increasing the timeout for those
devices will make the test pass but will also mean that the test isn't
really testing anything because the delay it is trying to catch was a
delay of up to one second.
I don't think this is an appropriate test to run once on CI. If there is
to be a test for the issue in question, it should be a benchmark test
that is run a large number of times. We don't really have such tests in
CI yet
I would argue that this test is actively a problem. It does not reliably
catch the issue it is supposed to catch, nor can it likely be made to do
so. (To do so would likely require converting it to a benchmarking test
as previously described. We don't run those in CI, at least not at this
time.)
Because this test will have both false positives and false negatives,
especially on the slower devices, it contributes to a culture of
dismissing failed tests. It does not reliably identify an issue nor does
it reliably pass on a working code base. This test should be removed.
Ref: https://github.com/nodejs/node/pull/4277
PR-URL: https://github.com/nodejs/node/pull/4786
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Brian White <mscdex@mscdex.net>
* Exchange 20 millisecond timers for setImmediate().
* Do not attempt to unlink path that will have been guaranteed to be
removed by `common.refreshTmpDir()`
* Do not swallow errors thrown by failed creation of needed test
subdirectory. If that happens, we want to know about it.
* Use `common.isSunOS` in one place where it is applicable
PR-URL: https://github.com/nodejs/node/pull/4776
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Evan Lucas <evanlucas@me.com>
From time to time this test is failing in OS X because at least one of
the connections takes quite a long time (around 5 seconds) causing some
of the timers may fire before the test exited. To solve this, wait for
all the connections to be established before setting the timeouts and
unrefing the sockets.
PR-URL: https://github.com/nodejs/node/pull/4772
Reviewed-By: Rich Trott <rtrott@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Testing this wasn't really useful, besides Object.observe is going to be
deprecated.
Also this test fails with Chakra (#4765) for obvious reason.
PR-URL: https://github.com/nodejs/node/pull/4769
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Brian White <mscdex@mscdex.net>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Roman Reiss <me@silverwind.io>
This change adds a new event handler to the `error` event of the socket
after it has been used by the http_client.
The purpose of this change is to catch errors on *keep alived*
connections from idle sockets that otherwise will cause an uncaugh error
event on the application.
Fix: #3595
PR-URL: https://github.com/nodejs/node/pull/4482
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Helps in implementation of #6204, where some options passed to
`createSecurePair()` are ignored before this patch.
These options are very helpful if someone wants to pass
`options.servername` or `options.SNICallback` to securepair.
PR-URL: https://github.com/nodejs/node/pull/2441
Reviewed-By: Fedor Indutny <fedor@indutny.com>
In some conditions it can happen that the client-side socket is destroyed
before the server-side socket has gracefully closed, thus causing a
'ECONNRESET' error in this socket. To solve this, wait in the client-side
socket for the 'end' event before closing it.
PR-URL: https://github.com/nodejs/node/pull/4043
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Previously, test-cluster-disconnect-suicide-race had two issues:
* Magic numbers: How many times to spawn a worker was determined through
empirical experimentation. This means that as new platforms and new
CPU/RAM configurations are tested, the magic numbers require more
and more refinement. This brings us to...
* Non-determinism: The test seems to fail all the time when the bug
it tests for is present, but it's really a judgment based on sampling.
"Oh, with 8 workers per CPU, it fails about 80% of the time. Let's try
16..."
This revised version of the test takes a different approach. The fix
for the bug that the test was written for means that the disconnect
event will fire on a subsequent tick. So we check for that and the test
still fails when the fix is not in the code base and succeeds when it
is.
Advantages of this approach include:
* The test runs much faster.
* The test should be reliable on any new platform regardless of CPU and
RAM.
PR-URL: https://github.com/nodejs/node/pull/4739
Ref: https://github.com/nodejs/node/pull/4674
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: James M Snell <jasnell@gmail.com>
Make the byteLength work correctly when input is Buffer.
e.g:
```js
// The incomplete unicode string
Buffer.byteLength(new Buffer([0xe4, 0xb8, 0xad, 0xe6, 0x96]))
```
The old output: 9
The new output: 5
PR-URL: https://github.com/nodejs/node/pull/4738
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Brian White <mscdex@mscdex.net>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Previously, test-cluster-disconnect-leak had two issues:
* Magic numbers: How many times to spawn a worker was determined through
empirical experimentation. This means that as new platforms and new
CPU/RAM configurations are tested, the magic numbers require more
and more refinement. This brings us to...
* Non-determinism: The test *seems* to fail all the time when the bug
it tests for is present, but it's really a judgment based on sampling.
"Oh, with 8 workers per CPU, it fails about 80% of the time. Let's try
16..."
This revised version of the test takes a different approach. The fix
for the bug that the test was written for means that the `disconnect`
event will fire reliably for a single worker. So we check for that and
the test still fails when the fix is not in the code base and succeeds
when it is.
Advantages of this approach include:
* The test runs much faster.
* The test now works on Windows. The previous version skipped Windows.
* The test should be reliable on any new platform regardless of CPU and
RAM.
Ref: https://github.com/nodejs/node/pull/4674
PR-URL: https://github.com/nodejs/node/pull/4736
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Clean up OpenSSL error stack in `ECDH::Initialize`, some curves have
faulty implementations that are leaving dangling errors after
initializing the curve.
Fix: #4686
PR-URL: https://github.com/nodejs/node/pull/4689
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
Reviewed-By: Shigeki Ohtsu <ohtsu@iij.ad.jp>
Prior to this commit, the test was flaky because it was
executing the majority of its logic in a function called from
the client and multiple events on the server. This commit
simplifies the test by separating the server's connection and
listening events, and isolating the client logic.
Refs: https://github.com/nodejs/node/pull/4476
Refs: https://github.com/nodejs/node/pull/4644
PR-URL: https://github.com/nodejs/node/pull/4650
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
A 50ms timeout results in a race condition. Instead, enforce expected
order through callbacks. This has the side effect of speeding up the
test in most situations.
Ref: https://github.com/nodejs/node/pull/4476
PR-URL: https://github.com/nodejs/node/pull/4637
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Reviewed-By: James M Snell <jasnell@gmail.com>
In test-cluster-worker-wait-server-close, remove unneeded 1-second delay
and refactor to eliminate flakiness on FreeBSD.
PR-URL: https://github.com/nodejs/node/pull/4616
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Reviewed-By: James M Snell <jasnell@gmail.com>
Two tests were requiring the common module twice. This removes the
duplicate require statement in the tests.
PR-URL: https://github.com/nodejs/node/pull/4611
Reviewed-By: Brian White <mscdex@mscdex.net>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Reviewed-By: James M Snell <jasnell@gmail.com>
A few tests in test/gc include the http module twice. Remove duplicate
require().
PR-URL: https://github.com/nodejs/node/pull/4606
Reviewed-By: Brian White <mscdex@mscdex.net>
Reviewed-By: Michaël Zasso <mic.besace@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
It's not guaranteed that the socket data is received in the same chunk
as the upgrade response. Listen for the `data` event to make sure all
the data is received.
PR-URL: https://github.com/nodejs/node/pull/4602
Reviewed-By: Brian White <mscdex@mscdex.net>
Reviewed-By: James M Snell <jasnell@gmail.com>
Remove unnecessary `setImmediate()` that causes a minor race condition.
Stop the test after 3 occurrences rather than 5 to allow for slower
hosts running the test in parallel with other tests.
Fixes: https://github.com/nodejs/node/issues/4559
PR-URL: https://github.com/nodejs/node/pull/4599
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
If one were to set NODE_REPL_HISTORY to a string that contains only a
space (" "), then the history file would be created with that name
which can cause problems are certain systems.
PR-URL: https://github.com/nodejs/node/pull/4539
Reviewed-By: Sakthipriyan Vairamani <thechargingvolcano@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
It's not guaranteed that the socket data is received in the same chunk
as the upgrade response. Listen for the `data` event to make sure all
the data is received.
PR-URL: https://github.com/nodejs/node/pull/4520
Reviewed-By: Rich Trott <rtrott@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
It avoids the creation of unnecessary handles. This issue is causing
intermitent failures in `test-cluster-disconnect-race` on `FreeBSD`
and `OS X`.
The problem is that the `worker2.disconnect` is being called on the
master before the `queryServer` is handled, causing the worker to
be deleted, then the Server handle is created afterwards. Later on,
when `removeWorker` is called from the `exit` handler, there are no
workers left, but one handle, thus the `AssertionError`.
Add a new `test/sequential/test-cluster-disconnect-leak` based on
`test-cluster-disconnect-race` that creates lots of workers and fails
consistently without this patch.
PR-URL: https://github.com/nodejs/node/pull/4465
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Johan Bergström <bugs@bergstroem.nu>
Reviewed-By: Rich Trott <rtrott@gmail.com>
Since headers are stored in an empty literal object ({}) instead
of an object created with Object.create(null), care must be taken
with property names inherited from Object. Currently there are
only functions inherited, so we can safely check for existing
strings instead.
Fixes: https://github.com/nodejs/node/issues/4456
PR-URL: https://github.com/nodejs/node/pull/4460
Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Myles Borins <myles.borins@gmail.com>
Reviewed-By: Minwoo Jung <jmwsoft@gmail.com>
Before this commit, it was possible on Windows for the server's
'connection' handler to be called *after* the client socket's
'connect' handler. This caused the 'message' event to be missed
and the test would never end (timing out in CI). This problem
was more easily reproducible on a low resource (slow CPU)
Windows (2012r2) installation.
This commit waits until both handlers have been called before
sending the handle to the master process.
Fixes: https://github.com/nodejs/node/issues/3957
PR-URL: https://github.com/nodejs/node/pull/4444
Reviewed-By: Rich Trott <rtrott@gmail.com>
Windows would die with ECONNRESET most times when running
this particular test. This commit makes handling these errors
more tolerable.
PR-URL: https://github.com/nodejs/node/pull/4442
Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
There is no guarantee that the `suicide` property of a worker in the
master process is going to be set when the `disconnect` and `exit`
events are emitted.
To fix it, wait for the ACK of the suicide message from the master
before disconnecting the worker. Also, there's no need to send the
suicide message from the worker if the disconnection has been
initiated in the master.
Add `test-cluster-disconnect-suicide-race` that forks a lot of workers
to consistently reproduce the issue this patch tries to solve.
Modify `test-regress-GH-3238` so it checks both the `kill` and
`disconnect` cases. Also take into account that the `disconnect` event
may be received after the `exit` event.
PR-URL: https://github.com/nodejs/node/pull/4349
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Rich Trott <rtrott@gmail.com>
In some conditions it can happen that the client-side socket is
destroyed before the server-side socket has gracefully closed, thus
causing a 'ECONNRESET' error in this socket. To solve this, also close
gracefully in the client side.
PR-URL: https://github.com/nodejs/node/pull/3966
Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com>
Reviewed-By: James M Snell <jasnell@gmail.com>
Copy client CA certs and cert store when asynchronously selecting
`SecureContext` during `SNICallback`. We already copy private key,
certificate, and certificate chain, but the client CA certs were
missing.
Fix: #2772
PR-URL: https://github.com/nodejs/node/pull/3537
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Roman Reiss
Drew Folta
Rich Trott
Evan Lucas
cjihrig
Santiago Gimeno
Vladimir Kurchatkin
José F. Romaniello
Коренберг Марк
Michaël Zasso
Myles Borins
Jackson Tian
Fedor Indutny
Brian White
Santiago Gimeno
Ben Noordhuis
James M Snell