Provide means to inspect information about the separate heap spaces
via a callable API. This is helpful to analyze memory issues.
Fixes: https://github.com/nodejs/node/issues/2079
PR-URL: https://github.com/nodejs/node/pull/4463
Reviewed-By: Colin Ihrig <cjihrig@gmail.com>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: James M Snell <jasnell@gmail.com>
The initial implementation of setPropByIndex() set the value of an Array
by index during development. Though the final form of the function
simply pushes passed values to an array as passed by arguments. Thus the
functions have been renamed to pushValueToArray() and
push_values_to_array_function() respectively.
Also add define for maximum number of arguments should be used before
hitting the limit of performance increase.
Fixes: 494227b "node: improve GetActiveRequests performance"
PR-URL: https://github.com/nodejs/node/pull/3780
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Fix node exiting due to an exception being thrown rather than emitting
an `'uncaughtException'` event on the process object when:
1. no error handler is set on the domain within which an error is thrown
2. an `'uncaughtException'` event listener is set on the process
Also fix an issue where the process would not abort in the proper
function call if an error is thrown within a domain with no error
handler and `--abort-on-uncaught-exception` is used.
Finally, change the behavior of --abort-on-uncaught-exception so that,
if the domain within which the error is thrown has no error handler, but
a domain further up the domains stack has one, the process will not
abort.
Fixes#3607 and #3653.
PR: #3654
PR-URL: https://github.com/nodejs/node/pull/3654
Reviewed-By: Chris Dickinson <chris@neversaw.us>
Call a user's callback to notify that the handle has been destroyed.
Only pass the id of the AsyncWrap instance since the object no longer
exists.
The object that's being destructed should never be inspected within the
callback or any time afterward.
This commit make a breaking change. The init callback will now be passed
arguments in the order of provider, id, parent.
PR-URL: https://github.com/nodejs/node/pull/3461
Reviewed-By: Fedor Indutny <fedor@indutny.com>
New instances of AsyncWrap are automatically assigned a unique id. The
value will be used in future commits to communicate additional
information via the async hooks.
While the largest value we can reliably communicate to JS is 2^53, even
if a new AsyncWrap is created every 100ns the uid won't reach its end
for 28.5 years.
PR-URL: https://github.com/nodejs/node/pull/3461
Reviewed-By: Fedor Indutny <fedor@indutny.com>
This fix is to be consistent implementation with ALPN. Tow NPN
protocol data in the persistent memebers move to hidden variables in
the wrap object.
PR-URL: https://github.com/nodejs/node/pull/2564
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
ALPN is added to tls according to RFC7301, which supersedes NPN.
When the server receives both NPN and ALPN extensions from the client,
ALPN takes precedence over NPN and the server does not send NPN
extension to the client. alpnProtocol in TLSSocket always returns
false when no selected protocol exists by ALPN.
In https server, http/1.1 token is always set when no
options.ALPNProtocols exists.
PR-URL: https://github.com/nodejs/node/pull/2564
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
v8 is faster at setting object properties in JS than C++. Even when it
requires calling into JS from native code. Make
process._getActiveRequests() faster by doing this when populating the
array containing request objects.
Simple benchmark:
for (let i = 0; i < 22; i++)
fs.open(__filename, 'r', function() { });
let t = process.hrtime();
for (let i = 0; i < 1e6; i++)
process._getActiveRequests();
t = process.hrtime(t);
console.log((t[0] * 1e9 + t[1]) / 1e6);
Results between the two:
Previous: 4406 ns/op
Patched: 690 ns/op 5.4x faster
PR-URL: https://github.com/nodejs/node/pull/3375
Reviewed-By: James Snell <jasnell@gmail.com>
Reviewed-By: Ben Noordhuis <ben@strongloop.com>
Previous logic didn't allow parent to propagate to the init callback
properly. The fix now allows the init callback to be called and receive
the parent if:
- async wrap callbacks are enabled and parent exists
- the init callback has been called on the parent and an init callback
exists then it will be called regardless of whether async wrap
callbacks are disabled.
Change the init/pre/post callback checks to see if it has been properly
set. This allows removal of the Environment "using_asyncwrap" variable.
Pass Isolate to a TryCatch instance.
Fixes: https://github.com/nodejs/node/issues/2986
PR-URL: https://github.com/nodejs/node/pull/3216
Reviewed-By: Rod Vagg <rod@vagg.org>
If the constructor can't assign a class id then the heap snapshot will
not be able to report the object. So ensure that all AsyncWrap instances
use a FunctionTemplate instance with an internal field count >= 1.
PR-URL: https://github.com/nodejs/node/pull/3139
Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-by: Stephen Belanger <admin@stephenbelanger.com>
Speeds up property lookups a little and it creates the string in the
old space straight away. It's a little easier on the garbage collector
because it doesn't have to track eternalized strings in the new space.
PR-URL: https://github.com/nodejs/node/pull/3060
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Overall construction time of Typed Arrays is faster in JS, but the
problem with using it normally is zero-fill of memory. Get around this
by using a flag in the ArrayBuffer::Allocator to trigger when memory
should or shouldn't be zero-filled.
Remove Buffer::Create() as it is no longer called.
The creation of the Uint8Array() was done at each callsite because at
the time of this patch there was a performance penalty for centralizing
the call in a single function.
PR-URL: https://github.com/nodejs/node/pull/2866
Reviewed-By: Fedor Indutny <fedor@indutny.com>
There might be a need to "kick off" the next tick queue and execute
events on it. Normally it is done through the `MakeCallback` interface,
but in case when it is not - we need a way to "kick them off" manually.
PR-URL: https://github.com/nodejs/node/pull/2355
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Expose and use in TLSWrap an `v8::External` wrap of the
`StreamBase*` pointer instead of guessing the ancestor C++ class in
`node_wrap.h`.
Make use of `StreamBase::Callback` structure for storing/passing both
callback and context in a single object.
Introduce `GetObject()` for future user-land usage, when a child class
is not going to be inherited from AsyncWrap.
PR-URL: https://github.com/nodejs/node/pull/2351
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
`enableTicketKeyCallback` and `onticketkeycallback` could be potentially
used to renew the TLS Session Tickets before they expire. However this
commit will introduce it only for private use yet, because we are not
sure about the API, and already need this feature for testing.
See: 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>
Upgrade the bundled V8 and update code in src/ and lib/ to the new API.
Notable backwards incompatible changes are the removal of the smalloc
module and dropped support for CESU-8 decoding. CESU-8 support can be
brought back if necessary by doing UTF-8 decoding ourselves.
This commit includes https://codereview.chromium.org/1192973004 to fix
a build error on python 2.6 systems. The original commit log follows:
Use optparse in js2c.py for python compatibility
Without this change, V8 won't build on RHEL/CentOS 6 because the
distro python is too old to know about the argparse module.
PR-URL: https://github.com/nodejs/io.js/pull/2022
Reviewed-By: Rod Vagg <rod@vagg.org>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
With V8 4.4 removing the external array data API currently used by
Buffer, the new implementation uses the Uint8Array to back Buffer.
Buffers now have a maximum size of Smi::kMaxLength, as defined by V8.
Which is ~2 GB on 64 bit and ~1 GB on 32 bit.
The flag --use-old-buffer allows using the old Buffer implementation.
This flag will be removed once V8 4.4 has landed.
The two JS Buffer implementations have been split into two files for
simplicity.
Use getter to return expected .parent/.offset values for backwards
compatibility.
PR-URL: https://github.com/nodejs/io.js/pull/1825
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Record the start time so we can make the return value of Timer.now()
relative to it, increasing the chances that it fits in a tagged integer
instead of a heap-allocated double, at least for the first one or two
billion milliseconds.
PR-URL: https://github.com/nodejs/io.js/pull/2256
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Use the --trace-sync-io flag to print a stack trace whenever a sync
method is used after the first tick, excluding during the process exit
event. (e.g. fs.readFileSync()) It does not track if the warning has
occurred at a specific location in the past and so will print the
warning every time.
Reason for not printing during the first tick of the appication is so
all necessary resources can be required. Also by excluding synchronous
calls during exit is necessary in case any data needs to be logged out
by the application before it shuts down.
Fixes: https://github.com/nodejs/io.js/issues/1674
PR-URL: https://github.com/nodejs/io.js/pull/1707
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
Reviewed-By: Petka Antonov <petka_antonov@hotmail.com>
Some calls to ReqWrap would get through the initial check and allow the
init callback to run, even though the callback had not been used on the
parent. Fix by explicitly checking if the parent has a queue.
Also change the name of the check, and internal field of AsyncHooks.
Other names were confusing.
PR-URL: https://github.com/iojs/io.js/pull/1614
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Do not enable ClientHello parser for async SNI/OCSP. Use new
OpenSSL-1.0.2's API `SSL_set_cert_cb` to pause the handshake process and
load the cert/OCSP response asynchronously. Hopefuly this will make
whole async SNI/OCSP process much faster and will eventually let us
remove the ClientHello parser itself (which is currently used only for
async session, see #1462 for the discussion of removing it).
NOTE: Ported our code to `SSL_CTX_add1_chain_cert` to use
`SSL_CTX_get0_chain_certs` in `CertCbDone`. Test provided for this
feature.
Fix: https://github.com/iojs/io.js/issues/1423
PR-URL: https://github.com/iojs/io.js/pull/1464
Reviewed-By: Shigeki Ohtsu <ohtsu@iij.ad.jp>
It's possible for an accessor or named interceptor to get called with
a different execution context than the one it lives in, see the test
case for an example using the debug API.
This commit fortifies against that by passing the environment as a
data property instead of looking it up through the current context.
Fixes: https://github.com/iojs/io.js/issues/1190 (again)
PR-URL: https://github.com/iojs/io.js/pull/1238
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Ensure that the debug context has an Environment assigned in case
a fatal error is raised.
The fatal exception handler in node.cc is not equipped to deal with
contexts that don't have one and can't easily be taught that due to
a deficiency in the V8 API: there is no way for the embedder to tell
if the data index is in use.
Fixes: https://github.com/iojs/io.js/issues/1190
PR-URL: https://github.com/iojs/io.js/pull/1229
Reviewed-By: Fedor Indutny <fedor@indutny.com>
If run with --abort-on-uncaught-exception, V8 will abort the process
whenever it does not see a JS-installed CatchClause in the stack. C++
TryCatch clauses are ignored. Domains work by setting a FatalException
handler which is ignored when running in abort mode.
This patch modifies MakeCallback to call its target function through a
JS function that installs a CatchClause and manually calls _fatalException
on error, if the process is both using domains and is in abort mode.
Semver: patch
PR-URL: https://github.com/iojs/io.js/pull/922
Fixes: https://github.com/iojs/io.js/issues/836
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Introduce a way to wrap plain-js `stream.Duplex` streams into C++
StreamBase's child class. With such method at hand it is now possible to
pass `stream.Duplex` instance as a `socket` parameter to
`tls.connect()`.
PR-URL: https://github.com/iojs/io.js/pull/926
Reviewed-By: Chris Dickinson <christopher.s.dickinson@gmail.com>
StreamBase is an improved way to write C++ streams. The class itself is
for separting `StreamWrap` (with the methods like `.writeAsciiString`,
`.writeBuffer`, `.writev`, etc) from the `HandleWrap` class, making
possible to write abstract C++ streams that are not bound to any uv
socket.
The following methods are important part of the abstraction (which
mimics libuv's stream API):
* Events:
* `OnAlloc(size_t size, uv_buf_t*)`
* `OnRead(ssize_t nread, const uv_buf_t*, uv_handle_type pending)`
* `OnAfterWrite(WriteWrap*)`
* Wrappers:
* `DoShutdown(ShutdownWrap*)`
* `DoTryWrite(uv_buf_t** bufs, size_t* count)`
* `DoWrite(WriteWrap*, uv_buf_t*, size_t count, uv_stream_t* handle)`
* `Error()`
* `ClearError()`
The implementation should provide all of these methods, thus providing
the access to the underlying resource (be it uv handle, TLS socket, or
anything else).
A C++ stream may consume the input of another stream by replacing the
event callbacks and proxying the writes. This kind of API is actually
used now for the TLSWrap implementation, making it possible to wrap TLS
stream into another TLS stream. Thus legacy API calls are no longer
required in `_tls_wrap.js`.
PR-URL: https://github.com/iojs/io.js/pull/840
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Reviewed-By: Chris Dickinson <christopher.s.dickinson@gmail.com>
This commit also breaks up req_wrap.h into req-wrap.h and req-wrap-inl.h
to work around a circular dependency issue in env.h.
PR-URL: https://github.com/iojs/io.js/pull/667
Reviewed-By: Fedor Indutny <fedor.indutny@gmail.com>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
AsyncWrap should always properly propagate asynchronous calls to any
child that is created. Regardless whether kCallInitHook is currently
active. The previous logic would always return early if kCallInitHook
wasn't set.
PR-URL: https://github.com/iojs/io.js/pull/732
Reviewed-by: Ben Noordhuis <info@bnoordhuis.nl>
* Include a description for the error message
* For rename, link, and symlink, include both the source and destination
path in the error message.
* Expose the destination path as the `dest` property on the error object.
* Fix a bug where `ThrowUVException()` would incorrectly delegate to
`Environment::TrowErrnoException()`.
API impact:
* Adds an extra overload for node::UVException() which takes 6
arguments.
PR: https://github.com/iojs/io.js/pull/675
Fixes: https://github.com/iojs/io.js/issues/207
Closes: https://github.com/iojs/io.js/pull/293
Reviewed-by: Ben Noordhuis <info@bnoordhuis.nl>
Since setting object properties in C++ can be slow, pass
data to JS using preallocated smalloc buffer and create
object in JS instead.
PR-URL: https://github.com/iojs/io.js/pull/469
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Use an array instead of an object to pass a parsed header chunk from c++
to javascript. This offers a 5-10% speedup on the http_simple benchmark,
as evidenced by running:
ab -k -t 100 -c 100 http://127.0.0.1:8000/bytes/100
PR: https://github.com/iojs/io.js/pull/292
Reviewed-by: Ben Noordhuis <info@bnoordhuis.nl>
The copyright and license notice is already in the LICENSE file. There
is no justifiable reason to also require that it be included in every
file, since the individual files are not individually distributed except
as part of the entire package.
Remove the 'gc' event from the v8 module and remove the supporting
infrastructure from src/. It gets the axe because:
1. There are currently no users. It was originally conceived as
an upstreamed subset of StrongLoop's strong-agent GC metrics,
but the strong-agent code base has evolved considerably since
that time and has no use anymore for what is in core.
2. The implementation is not quite sound. It calls into JS land
from inside the GC epilog and that is unsafe. We could fix
that by delaying the callback until a safe time but because
there are no users anyway, removing it is all around easier.
PR-URL: https://github.com/iojs/io.js/pull/174
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Call a user-defined callback at specific points in the lifetime of an
asynchronous event. Which are on instantiation, just before/after the
callback has been run.
**If any of these callbacks throws an exception, there is no forgiveness
or recovery. A message will be displayed and a core file dumped.**
Currently these only tie into AsyncWrap, meaning no call to a hook
callback will be made for timers or process.nextTick() events. Though
those will be added in a future commit.
Here are a few notes on how to make the hooks work:
- The "this" of all event hook callbacks is the request object.
- The zero field (kCallInitHook) of the flags object passed to
setupHooks() must be set != 0 before the init callback will be called.
- kCallInitHook only affects the calling of the init callback. If the
request object has been run through the create callback it will always
run the before/after callbacks. Regardless of kCallInitHook.
- In the init callback the property "_asyncQueue" must be attached to
the request object. e.g.
function initHook() {
this._asyncQueue = {};
}
- DO NOT inspect the properties of the object in the init callback.
Since the object is in the middle of being instantiated there are some
cases when a getter is not complete, and doing so will cause Node to
crash.
PR-URL: https://github.com/joyent/node/pull/8110
Signed-off-by: Trevor Norris <trev.norris@gmail.com>
Reviewed-by: Fedor Indutny <fedor@indutny.com>
Reviewed-by: Alexis Campailla <alexis@janeasystems.com>
Reviewed-by: Julien Gilli <julien.gilli@joyent.com>
Async Listener was the name of the user-facing JS API, and is being
completely removed. Instead low level hooks directly into the mechanism
that AL used will be introduced in a future commit.
PR-URL: https://github.com/joyent/node/pull/8110
Signed-off-by: Trevor Norris <trev.norris@gmail.com>
Reviewed-by: Fedor Indutny <fedor@indutny.com>
Reviewed-by: Alexis Campailla <alexis@janeasystems.com>
Reviewed-by: Julien Gilli <julien.gilli@joyent.com>
Call a user-defined callback at specific points in the lifetime of an
asynchronous event. Which are on instantiation, just before/after the
callback has been run.
**If any of these callbacks throws an exception, there is no forgiveness
or recovery. A message will be displayed and a core file dumped.**
Currently these only tie into AsyncWrap, meaning no call to a hook
callback will be made for timers or process.nextTick() events. Though
those will be added in a future commit.
Here are a few notes on how to make the hooks work:
- The "this" of all event hook callbacks is the request object.
- The zero field (kCallInitHook) of the flags object passed to
setupHooks() must be set != 0 before the init callback will be called.
- kCallInitHook only affects the calling of the init callback. If the
request object has been run through the create callback it will always
run the before/after callbacks. Regardless of kCallInitHook.
- In the init callback the property "_asyncQueue" must be attached to
the request object. e.g.
function initHook() {
this._asyncQueue = {};
}
- DO NOT inspect the properties of the object in the init callback.
Since the object is in the middle of being instantiated there are some
cases when a getter is not complete, and doing so will cause Node to
crash.
PR-URL: https://github.com/joyent/node/pull/8110
Signed-off-by: Trevor Norris <trev.norris@gmail.com>
Reviewed-by: Fedor Indutny <fedor@indutny.com>
Reviewed-by: Alexis Campailla <alexis@janeasystems.com>
Reviewed-by: Julien Gilli <julien.gilli@joyent.com>
Async Listener was the name of the user-facing JS API, and is being
completely removed. Instead low level hooks directly into the mechanism
that AL used will be introduced in a future commit.
PR-URL: https://github.com/joyent/node/pull/8110
Signed-off-by: Trevor Norris <trev.norris@gmail.com>
Reviewed-by: Fedor Indutny <fedor@indutny.com>
Reviewed-by: Alexis Campailla <alexis@janeasystems.com>
Reviewed-by: Julien Gilli <julien.gilli@joyent.com>
There is not much point in keeping it a separate project because it
doesn't build standalone, plus it makes applying changes to core more
difficult because of the implicit dependency on header files in src/.
Now that we are building with C++11 features enabled, replace use
of NULL with nullptr.
The benefit of using nullptr is that it can never be confused for
an integral type because it does not support implicit conversions
to integral types except boolean - unlike NULL, which is defined
as a literal `0`.
Attach the per-context execution environment directly to API functions.
Rationale:
* Gets node one step closer to multi-isolate readiness.
* Avoids multi-context confusion, e.g. when the caller and callee live
in different contexts.
* Avoids expensive calls to pthread_getspecific() on platforms where
V8 does not know how to use the thread-local storage directly.
(Linux, the BSDs.)
PR-URL: https://github.com/node-forward/node/pull/18
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Running fill() with an empty string would cause Node to hang
indefinitely. Now it will return without having operated on the buffer.
User facing function has been pulled into JS to perform all initial
value checks and coercions. The C++ method has been placed on the
"internal" object.
Coerced non-string values to numbers to match v0.10 support.
Simplified logic and changed a couple variable names.
Added tests for fill() and moved them all to the beginning of
buffer-test.js since many other tests depend on fill() working properly.
Fixes: https://github.com/joyent/node/issues/8469
Signed-off-by: Trevor Norris <trev.norris@gmail.com>
There is only one call site that uses it and that can do the checks
itself. Removes ~15 lines of code.
Reviewed-by: Trevor Norris <trev.norris@gmail.com>
Ben Ripkens
Brian White
Trevor Norris
Julien Gilli
Shigeki Ohtsu
Jeremy Whitlock
Ben Noordhuis
Fedor Indutny
Petka Antonov
Chris Dickinson
Bert Belder
Vladimir Kurchatkin
isaacs