It hits a compiler bug in gcc <= 4.4 similar to the issue that was
recently addressed in commit 157d2bc:
../deps/v8/include/v8.h: In function ‘char*
node::Buffer::Data(v8::Persistent&) [with TypeName = v8::Object]’:
../src/node_crypto.cc:1123: instantiated from here
../deps/v8/include/v8.h:876: error: ‘class v8::Data’ is not a
function,
../src/node_internals.h:356: error: conflict with ‘template char*
node::Buffer::Data(v8::Persistent&)’
../src/node_internals.h:357: error: in call to ‘Data’
Remove the helper function, it was only used in a couple of places.
Should fix the build on Ubuntu 10.04.
Fixes#5844.
Previously there was no way to pass a Function callback directly to
MakeCallback and support domains. The check has been added so that users
never have to worry about supporting domains while using MakeCallback.
X509_STORE_add_cert increment reference of passed `x509` cert,
`X509_free` must be called to avoid memory leak.
This is a back-port of commit c1db1ec from the master branch.
* uv: Upgrade to v0.10.12
* npm: Upgrade to 1.3.2
* windows: get proper errno (Ben Noordhuis)
* tls: only wait for finish if we haven't seen it (Timothy J Fontaine)
* http: Dump response when request is aborted (isaacs)
* http: use an unref'd timer to fix delay in exit (Peter Rust)
* zlib: level can be negative (Brian White)
* zlib: allow zero values for level and strategy (Brian White)
* buffer: add comment explaining buffer alignment (Ben Noordhuis)
* string_bytes: properly detect 64bit (Timothy J Fontaine)
* src: fix memory leak in UsingDomains() (Ben Noordhuis)
A pooled https agent may get a Connection: close, but never finish
destroying the socket as the prior request had already emitted finish
likely from a pipe.
Since the socket is not marked as destroyed it may get reused by the
agent pool and result in an ECONNRESET.
re: #5712#5739
Instead of destroying sockets when there are no pending requests, put
them in a freeSockets list, and unref() them so that they do not keep
the event loop open.
Also, set the default max sockets to Infinity, to prevent the awful
surprising deadlocks that happen when more connections are made.
When creating a slice, make sure to propagate the originating parent.
This is to prevent a buf.parent.parent.(etc) scenario.
Also speed up the constructor by preventing lookup of non-existant
properties by setting them beforehand in the prototype. (see
https://github.com/joyent/node/commit/7ce5a31#commitcomment-3332779)
Since the SlabAllocator was removed the buffer length/offset is no
longer sent to the onread callback. The benchmarks have been updated to
reflect that.
Fix bug where if dev passed a callback to Alloc then called AllocDispose
it wouldn't bother to pass the data to the callback and instead would
just free it.
There was previously up to a second exit delay when exiting node
right after an http request/response, due to the utcDate() function
doing a setTimeout to update the cached date/time.
Fixing this should increase the performance of our http tests.
This is a big commit that touches just about every file in the src/
directory. The V8 API has changed in significant ways. The most
important changes are:
* Binding functions take a const v8::FunctionCallbackInfo<T>& argument
rather than a const v8::Arguments& argument.
* Binding functions return void rather than v8::Handle<v8::Value>. The
return value is returned with the args.GetReturnValue().Set() family
of functions.
* v8::Persistent<T> no longer derives from v8::Handle<T> and no longer
allows you to directly dereference the object that the persistent
handle points to. This means that the common pattern of caching
oft-used JS values in a persistent handle no longer quite works,
you first need to reconstruct a v8::Local<T> from the persistent
handle with the Local<T>::New(isolate, persistent) factory method.
A handful of (internal) convenience classes and functions have been
added to make dealing with the new API a little easier.
The most visible one is node::Cached<T>, which wraps a v8::Persistent<T>
with some template sugar. It can hold arbitrary types but so far it's
exclusively used for v8::Strings (which was by far the most commonly
cached handle type.)
The previous commit removes our patch that builds V8 at -O2 rather
than -O3 so there is not much point in keeping the configure switch
around.
The reason it did so was to work around an assortment of compiler and
linker bugs. In particular, certain combinations of g++ and binutils
generate bad or no code when -ffunction-sections or -finline-functions
is enabled (which -O3 implicitly does.)
It was quite the problem back in the day because everyone and his dog
built from source. Now that we have prebuilt binaries and packages
available, there is no longer a pressing need to be so accommodating.
If you experience spurious (or possibly not so spurious) segmentation
faults after this commit, you need to upgrade your compiler/linker
toolchain.
UsingDomains() assigned process_tickCallback when it had already
been set by MakeCallback() a few frames down the call stack.
Dispose the handle first or we'll retain whatever is in the lexical
closure of the old process._tickCallback function.
Fixes#5795.
If a transform stream has objectMode = true, it should
allow falsey values other than (null) like 0, false, ''.
null is reserved to indicate stream eof but other falsey
values should flow through properly.
Now that Buffer instantiation has improved, the SlabAllocator is an
unnecessary layer of complexity preventing further performance
optimizations.
Currently there is a small performance loss with very small stream
requests, but this will soon be addressed.
Assert that when the client closes it has seen an error, this prevents
the test from timing out.
Also queue a second write in the case that we were able to send the
buffer before the other side closed the connection.
There was previously up to a second exit delay when exiting node
right after an http request/response, due to the utcDate() function
doing a setTimeout to update the cached date/time.
Fixing this should increase the performance of our http tests.
Assert that when the client closes it has seen an error, this prevents
the test from timing out.
Also queue a second write in the case that we were able to send the
buffer before the other side closed the connection.
Most TryCatch blocks have SetVerbose flag on, this tells V8 to report
uncaught exceptions to debugger.
FatalException handler is called from V8 Message listener instead from
the place where TryCatch was used. Otherwise uncaught exceptions are
logged twice.
See comment in `deps/v8/include/v8.h` for explanation of SetVerbose
flag:
> By default, exceptions that are caught by an external exception
> handler are not reported. Call SetVerbose with true on an
> external exception handler to have exceptions caught by the
> handler reported as if they were not caught.
The flag is used by `Isolate::ShouldReportException()`, which is called
by `Isolate::DoThrow()` to decide whether an exception is considered
uncaught.
Might cause write head running over read head, when there were no
allocation and `Commit()` was called. Source of at least one test
failure on windows (`simple/test-https-drain.js`).
gcc 4.2 on OS X gets confused about the call to node::Buffer::Data().
Fully qualify the function name to help it along.
Fixes the following build error:
../../deps/v8/include/v8.h: In function ‘char*
node::Buffer::Data(v8::Handle<v8::Value>)’:
../../deps/v8/include/v8.h:900: error: ‘class v8::Data’
is not a function,
../../src/node_buffer.h:38: error:
conflict with ‘char* node::Buffer::Data(v8::Handle<v8::Object>)’
../../src/node_buffer.cc:94: error:
in call to ‘Data’
Buffer(<String>) used to pass the string to js where it would then be
passed back to cpp for processing. Now only the buffer object
instantiation is done in js and the string is processed in cpp.
Also added a Buffer api that also accepts the encoding.
Old fill would take the char code of the first character and wrap around
the int to fit in the 127 range. Now fill will duplicate whatever string
is given through the entirety of the buffer.
Note: There is one bug around ending on a partial fill of any character
outside the ASCII range.
While the new Buffer implementation is much faster we still have the
necessity of using Buffer pools. This is undesirable because it may
still lead to unwanted memory retention, but for the time being this is
the best solution.
Because of this re-introduction, and since there is no more SlowBuffer
type, the SlowBuffer method has been re-purposed to return a non-pooled
Buffer instance. This will be helpful for developers to store data for
indeterminate lengths of time without introducing a memory leak.
Another change to Buffer pools was that they are only allocated if the
requested chunk is < poolSize / 2. This was done because allocations are
much quicker now, and it's a better use of the pool.
Memory allocations are now done through smalloc. The Buffer cc class has
been removed completely, but for backwards compatibility have left the
namespace as Buffer.
The .parent attribute is only set if the Buffer is a slice of an
allocation. Which is then set to the alloc object (not a Buffer).
The .offset attribute is now a ReadOnly set to 0, for backwards
compatibility. I'd like to remove it in the future (pre v1.0).
A few alterations have been made to how arguments are either coerced or
thrown. All primitives will now be coerced to their respective values,
and (most) all out of range index requests will throw.
The indexes that are coerced were left for backwards compatibility. For
example: Buffer slice operates more like Array slice, and coerces
instead of throwing out of range indexes. This may change in the future.
The reason for wanting to throw for out of range indexes is because
giving js access to raw memory has high potential risk. To mitigate that
it's easier to make sure the developer is always quickly alerted to the
fact that their code is attempting to access beyond memory bounds.
Because SlowBuffer will be deprecated, and simply returns a new Buffer
instance, all tests on SlowBuffer have been removed.
Heapdumps will now show usage under "smalloc" instead of "Buffer".
ParseArrayIndex was added to node_internals to support proper uint
argument checking/coercion for external array data indexes.
SlabAllocator had to be updated since handle_ no longer exists.
If the user knows the allocation is no longer needed then the memory can
be manually released.
Currently this will not ClearWeak the Persistent, so the callback will
still run.
If the user passed a ClearWeak callback, and then disposed the object,
the buffer callback argument will == NULL.
smalloc is a simple utility for quickly allocating external memory onto
js objects. This will be used to centralize how memory is managed in
node, and will become the backer for Buffers. So in the future crypto's
SlabBuffer, stream's SlabAllocator will be removed.
Note on the js API: because no arguments are optional the order of
arguments have been placed to match their cc counterparts as closely as
possible.
The console module has always been called 'stdio' in the
table-of-contents, but nowhere else, since its name is
'console'. This makes it difficult to find.
This is a back-port of commit 226a20d from the master branch.
Libuv may provide a NULL buffer to the uv_read callback in case of an
error, so with this assert we'd be using the api incorrectly. None of
the current DoRead implementations rely on this constraint, either.
The console module has always been called 'stdio' in the
table-of-contents, but nowhere else, since its name is
'console'. This makes it difficult to find.
Resolves minor discrepancies between android and standard POSIX systems.
In addition, some configure parameters were added, and a helper-script
for android configuration. Ideally, this script should be merged into
the standard configure script.
To build for android, source the android-configure script with an NDK
path:
source ./android-configure ~/android-ndk-r8d
This will create an android standalone toolchain and export the
necessary environment parameters.
After that, build as normal:
make -j8
After the build, you should now have android-compatible NodeJS binaries.
Suppress the following warning:
../../src/cares_wrap.cc: In function ‘v8::Handle<v8::Value>
node::cares_wrap::SetServers(const v8::Arguments&)’:
../../src/cares_wrap.cc:1017:5: warning: ‘uv_ret.uv_err_s::code’
may be used uninitialized in this function [-Wuninitialized]
Split `tls.js` into `_tls_legacy.js`, containing legacy
`createSecurePair` API, and `_tls_wrap.js` containing new code based on
`tls_wrap` binding.
Remove tests that are no longer useful/valid.
In case of cross-compilation host_arch_cc() function could return
target arch if CC was set to target arch compiler. Host arch
compiler should always be used in this case. This was broken
by commit 707863c.
This reverts commit a40133d10c.
Unfortunately, this breaks socket.io. Even though it's not strictly an
API change, it is too subtle and in too brittle an area of node, to be
done in a stable branch.
Conflicts:
doc/api/http.markdown
When large strings are used they cause v8's GC to spend a lot more time
cleaning up. In these cases it's much faster to use external string
resources.
UTF8 strings do not use external string resources because only one and
two byte external strings are supported.
EXTERN_APEX is the value at which v8's GC overtakes performance.
The following table has the type and buffer size that use to encode the
strings as rough estimates of the percentage of performance gain from
this patch (UTF8 is missing because they cannot be externalized).
encoding 128KB 1MB 5MB
-----------------------------
ASCII 58% 208% 250%
HEX 15% 74% 86%
BASE64 11% 74% 71%
UCS2 2% 225% 398%
BINARY 2234% 1728% 2305%
BINARY is so much faster across the board because of using the new v8
WriteOneByte API.
v8 has a new API to write out strings to memory. This has been
implemented.
One other change of note is BINARY encoded strings have a new
implementation. This has improved performance substantially.
Clang branch release_33 would optimize out a != NULL check because of
some undefined behavior. This is a floating patch as a backport of that
fix.
Committed: http://code.google.com/p/v8/source/detail?r=13570
Normalize the encoding in getEncoding() before using it. Fixes a
"AssertionError: Cannot change encoding" exception when the caller
mixes "utf8" and "utf-8".
Fixes#5655.
Before this commit NodeBIO never shrank, possibly consuming a lot of
memory (depending on reader's haste).
All buffers between write_head's child and read_head should be
deallocated on read, leaving only space left in write_head and in the
next buffer.
Commit 0bba5902 accidentally (or maybe erroneously) added node_isolate
to src/node.h and src/node_object_wrap.h.
Undo that, said variable is not for public consumption. Add-on authors
should use v8::Isolate::GetCurrent() instead.
I missed that while reviewing. Mea culpa.
Fixes#5639.
This is only relevant for CentOS 6.3 using kernel version 2.6.32.
On other linuxes and darwin, the `read` call gets an ECONNRESET in that
case. On sunos, the `write` call fails with EPIPE.
However, old CentOS will occasionally send an EOF instead of a
ECONNRESET or EPIPE when the client has been destroyed abruptly.
Make sure we don't keep trying to write or read more in that case.
Fixes#5504
However, there is still the question of what libuv should do when it
gets an EOF. Apparently in this case, it will continue trying to read,
which is almost certainly the wrong thing to do.
That should be fixed in libuv, even though this works around the issue.
In cases where there are multiple @-chars in a url, Node currently
parses the hostname and auth sections differently than web browsers.
This part of the bug is serious, and should be landed in v0.10, and
also ported to v0.8, and releases made as soon as possible.
The less serious issue is that there are many other sorts of malformed
urls which Node either accepts when it should reject, or interprets
differently than web browsers. For example, `http://a.com*foo` is
interpreted by Node like `http://a.com/*foo` when web browsers treat
this as `http://a.com%3Bfoo/`.
In general, *only* the `hostEndingChars` should be the characters that
delimit the host portion of the URL. Most of the current `nonHostChars`
that appear in the hostname should be escaped, but some of them (such as
`;` and `%` when it does not introduce a hex pair) should raise an
error.
We need to have a broader discussion about whether it's best to throw in
these cases, and potentially break extant programs, or return an object
that has every field set to `null` so that any attempt to read the
hostname/auth/etc. will appear to be empty.
In some cases, the http CONNECT/Upgrade API is unshifting an empty
bodyHead buffer onto the socket.
Normally, stream.unshift(chunk) does not set state.reading=false.
However, this check was not being done for the case when the chunk was
empty (either `''` or `Buffer(0)`), and as a result, it was causing the
socket to think that a read had completed, and to stop providing data.
This bug is not limited to http or web sockets, but rather would affect
any parser that unshifts data back onto the source stream without being
very careful to never unshift an empty chunk. Since the intent of
unshift is to *not* change the state.reading property, this is a bug.
Fixes#5557FixesLearnBoost/socket.io#1242
Remove the need to call start/stop the uv_idle spinner between
MakeCallbacks. The one place where the tick processor needs to be kicked
is where a user catches uncaughtException. For that we'll now use
setImmediate, which accomplishes the same task.
maxTickDepth checks have been removed for domains and replaced with a
flag that checks if the last callback threw. If it did then execution of
the remaining tickQueue is deferred to the spinner.
This is to prevent domains from entering a continuous loop when an error
callback also throws an error.
Removes the check for maxTickDepth for non-domain callbacks. So a user
can starve I/O by setting a recursive nextTick.
The domain case is more complex and will be addressed in another commit.
Previous code was calling uv_loop_delete() directly on a running loop,
which led to race condition aborts/segfaults within libuv. This change
changes the watchdog thread to call uv_run() with UV_RUN_ONCE so that
the call exits after either the timer times out or uv_async_send() is
called from the main thread in Watchdog::Destroy(). The timer/async
handles are then closed and uv_run() with UV_RUN_DEFAULT is called so
that libuv has a chance to cleanup before the thread exits. The main
thread meanwhile calls uv_thread_join() and then uv_loop_delete() to
complete the cleanup.
Before this, entering something like:
> JSON.parse('066');
resulted in the "..." prompt instead of displaying the expected
"SyntaxError: Unexpected number"
1. Emit `sslOutEnd` only when `_internallyPendingBytes() === 0`.
2. Read before checking `._halfRead`, otherwise we'll see only previous
value, and will invoke `._write` callback improperly.
3. Wait for both `end` and `finish` events in `.destroySoon`.
4. Unpipe encrypted stream from socket to prevent write after destroy.
Stream's `._write()` callback should be invoked only after it's opposite
stream has finished processing incoming data, otherwise `finish` event
fires too early and connection might be closed while there's some data
to send to the client.
see #5544
Quote from SSL_shutdown man page:
The output of SSL_get_error(3) may be misleading,
as an erroneous SSL_ERROR_SYSCALL may be flagged even though
no error occurred.
Also, handle all other errors to prevent assertion in `ClearError()`.
When writing bad data to EncryptedStream it'll first get to the
ClientHello parser, and, only after it will refuse it, to the OpenSSL.
But ClientHello parser has limited buffer and therefore write could
return `bytes_written` < `incoming_bytes`, which is not the case when
working with OpenSSL.
After such errors ClientHello parser disables itself and will
pass-through all data to the OpenSSL. So just trying to write data one
more time will throw the rest into OpenSSL and let it handle it.
Otherwise, writing an empty string causes the whole program to grind to
a halt when piping data into http messages.
This wasn't as much of a problem (though it WAS a bug) in 0.8 and
before, because our hyperactive 'drain' behavior would mean that some
*previous* write() would probably have a pending drain event, and cause
things to start moving again.
This saves a few calls to gettimeofday which can be expensive, and
potentially subject to clock drift. Instead use the loop time which
uses hrtime internally.
fixes#5497
This commit adds an optimization to the HTTP client that makes it
possible to:
* Pack the headers and the first chunk of the request body into a
single write().
* Pack the chunk header and the chunk itself into a single write().
Because only one write() system call is issued instead of several,
the chances of data ending up in a single TCP packet are phenomenally
higher: the benchmark with `type=buf size=32` jumps from 50 req/s to
7,500 req/s, a 150-fold increase.
This commit removes the check from e4b716ef that pushes binary encoded
strings into the slow path. The commit log mentions that:
We were assuming that any string can be concatenated safely to
CRLF. However, for hex, base64, or binary encoded writes, this
is not the case, and results in sending the incorrect response.
For hex and base64 strings that's certainly true but binary strings
are 'das Ding an sich': string.length is the same before and after
decoding.
Fixes#5528.
Consider a user on his Mac, who wants to cross-compile for his Linux ARM device:
./configure --dest-cpu=arm --dest-os=linux
Before this patch, for example, DTrace probes would incorrectly attempt to be
enabled because the configure script is running on a MacOS machine, even though
we're trying to compile a binary for `linux`.
With this patch, the `--dest-os` flag is respected throughout the configure
script, thus leaving DTrace probes disabled in this cross-compiling scenario.
When an internal api needs a timeout, they should use
timers._unrefActive since that won't hold the loop open. This solves
the problem where you might have unref'd the socket handle but the
timeout for the socket was still active.
Previously one could write anywhere in a buffer pool if they accidently
got their offset wrong. Mainly because the cc level checks only test
against the parent slow buffer and not against the js object properties.
So now we check to make sure values won't go beyond bounds without
letting the dev know.
Because of variations in different base64 implementation, it's been
decided to strip all padding from the end of a base64 string and
calculate its size from that.
Add localAddress and localPort properties to tls.CleartextStream.
Like remoteAddress and localPort, delegate to the backing net.Socket
object.
Refs #5502.
Instead of String::New every time, use a Persistent sym. This can be
accomplished in two ways:
1) Local<String> str = *persistent_str_sym;
2) Handle<String> str = persistent_str_sym;
I've chosen to use the latter method for simplicity's sake.
Other small changes include creating syms on Initialize and removing
unnecessary Local casting on return values.
The default encoding is 'buffer'. When the input is a string, treat it
as 'binary'. Fixes the following assertion:
node: ../src/string_bytes.cc:309: static size_t
node::StringBytes::StorageSize(v8::Handle<v8::Value>, node::encoding):
Assertion `0 && "buffer encoding specified but string provided"'
failed.
Introduced in 64fc34b2.
Fixes#5482.
Test case:
var t = setInterval(function() {}, 1);
process.nextTick(t.unref);
Output:
Assertion failed: (args.Holder()->InternalFieldCount() > 0),
function Unref, file ../src/handle_wrap.cc, line 78.
setInterval() returns a binding layer object. Make it stop doing that,
wrap the raw process.binding('timer_wrap').Timer object in a Timeout
object.
Fixes#4261.
Commit 38149bb changes http.get() and http.request() to escape unsafe
characters. However, that creates an incompatibility with v0.10 that
is difficult to work around: if you escape the path manually, then in
v0.11 it gets escaped twice. Change lib/http.js so it no longer tries
to fix up bad request paths, simply reject them with an exception.
The actual check is rather basic right now. The full check for illegal
characters is difficult to implement efficiently because it requires a
few characters of lookahead. That's why it currently only checks for
spaces because those are guaranteed to create an invalid request.
Fixes#5474.
getServers returns an array of ips that are currently being used for
resolution
setServers takes an array of ips that are to be used for resolution,
this will throw if there's invalid input but preserve the original
configuration
This also templatizes the Buffer::*Slice functions, and the template
function probably cannot be safely used outside of Node. However, it
also SHOULD not be used outside of Node, so this is arguably a feature
as well as a caveat.
Pretty much everything assumes strings to be utf-8, but crypto
traditionally used binary strings, so we need to keep the default
that way until most users get off of that pattern.
If there is an encoding, and we do 'stream.push(chunk, enc)', and the
encoding argument matches the stated encoding, then we're converting from
a string, to a buffer, and then back to a string. Of course, this is a
completely pointless bit of work, so it's best to avoid it when we know
that we can do so safely.
Empirical evidence suggests that OS-level load balancing (that is,
having multiple processes listen on a socket and have the operating
system wake up one when a connection comes in) produces skewed load
distributions on Linux, Solaris and possibly other operating systems.
The observed behavior is that a fraction of the listening processes
receive the majority of the connections. From the perspective of the
operating system, that somewhat makes sense: a task switch is expensive,
to be avoided whenever possible. That's why the operating system likes
to give preferential treatment to a few processes, because it reduces
the number of switches.
However, that rather subverts the purpose of the cluster module, which
is to distribute the load as evenly as possible. That's why this commit
adds (and defaults to) round-robin support, meaning that the master
process accepts connections and distributes them to the workers in a
round-robin fashion, effectively bypassing the operating system.
Round-robin is currently disabled on Windows due to how IOCP is wired
up. It works and you can select it manually but it probably results in
a heavy performance hit.
Fixes#4435.
Commit 9352c19 ("child_process: don't emit same handle twice") trades
one bug for another.
Before said commit, a handle was sometimes delivered with messages it
didn't belong to.
The bug fix introduced another bug that needs some explaining. On UNIX
systems, handles are basically file descriptors that are passed around
with the sendmsg() and recvmsg() system calls, using auxiliary data
(SCM_RIGHTS) as the transport.
node.js and libuv depend on the fact that none of the supported systems
ever emit more than one SCM_RIGHTS message from a recvmsg() syscall.
That assumption is something we should probably address someday for the
sake of portability but that's a separate discussion.
So, SCM_RIGHTS messages are never coalesced. SCM_RIGHTS and normal
messages however _are_ coalesced. That is, recvmsg() might return this:
recvmsg(); // { "message-with-fd", "message", "message" }
The operating system implicitly breaks pending messages along
SCM_RIGHTS boundaries. Most Unices break before such messages but Linux
also breaks _after_ them. When the sender looks like this:
sendmsg("message");
sendmsg("message-with-fd");
sendmsg("message");
Then on most Unices the receiver sees messages arriving like this:
recvmsg(); // { "message" }
recvmsg(); // { "message-with-fd", "message" }
The bug fix in commit 9352c19 assumes this behavior. On Linux however,
those messages can also come in like this:
recvmsg(); // { "message", "message-with-fd" }
recvmsg(); // { "message" }
In other words, it's incorrect to assume that the file descriptor is
always attached to the first message. This commit makes node wise up.
Fixes#5330.
This adds proper support for the following situation:
w.cork();
w.write(...);
w.cork();
w.write(...);
w.uncork();
w.write(...);
w.uncork();
This is relevant when you have a function (as we do in HTTP) that wants
to use cork, but in some cases, want to have a cork/uncork *around*
that function, without losing the benefits of writev.
In synchronous Writable streams (where the _write cb is called on the
current tick), the 'finish' event (and thus the end() callback) can in
some cases be called before all the write() callbacks are called.
Use a counter, and have stream.Transform rely on the 'prefinish' event
instead of the 'finish' event.
This has zero effect on most streams, but it corrects an edge case and
makes it perform more deterministically, which is a Good Thing.
uv_async_t handles for dispatching of debug messages and
emitting NODE_DEBUG_ENABLED used to be initialized every time
node::EnableDebug() was called, which happened every time
user sends a SIGUSR1.
Now they are initialized only once from node::Init() during
application start.
Implement support for debugging cluster workers. Each worker process
is assigned a new debug port in an increasing sequence.
I.e. when master process uses port 5858, then worker 1 uses port 5859,
worker 2 uses port 5860, and so on.
Introduce new command-line parameter '--debug-port=' which sets debug_port
but does not start debugger. This option works for all node processes, it
is not specific to cluster workers.
Fixesjoyent/node#5318.
Preserve default install prefix seen in process.config, but use DESTDIR
for installing to deliniate 32/64 versions, avoid conflicts with PREFIX
settings in config.mk
Preserve default install prefix seen in process.config, but use DESTDIR
for installing to deliniate 32/64 versions, avoid conflicts with PREFIX
settings in config.mk
Change vcbuild.bat to ignore VCINSTALLDIR environment variable,
always check for specific VS version and set GYP_MSVS_VERSION
accordingly. Otherwise GYP generates project files in format
that cannot be compiled by VS2012.
When developer calls setBreakpoint with an unknown script name,
we convert the script name into regular expression matching all
paths ending with given name (name can be a relative path too).
To create such breakpoint in V8, we use type `scriptRegEx`
instead of `scriptId` for `setbreakpoint` request.
To restore such breakpoint, we save the original script name
send by the user. We use this original name to set (restore)
breakpoint in the new child process.
This is a back-port of commit 5db936d from the master branch.
After much investigation it turns out that the affected servers are
buggy. user-service.condenastdigital.com:443 in particular seems to
reject large TLS handshake records. Cutting down the number of
advertised ciphers or disabling SNI fixes the issue.
Similarly, passing { secureOptions: constants.SSL_OP_NO_TLSv1_2 }
seems to fix most connection issues with IIS servers.
Having to work around buggy servers is annoying for our users but not
a reason to downgrade OpenSSL. Therefore, revert it.
This reverts commit 4fdb8acdae.
Add a watchdog class which executes a timer in a separate event loop in
a separate thread that will terminate v8 execution if it expires.
Add timeout argument to functions in vm module which use the watchdog
if a non-zero timeout is specified.
This commit undoes the downgrade from OpenSSL v1.0.1e to v1.0.0f,
effectively upgrading OpenSSL to v1.0.1e again. The reason for the
downgrade was to work around compatibility issues with certain TLS
servers in the stable branch. See the commit log of 4fdb8ac and the
linked issue for details. We're going to revisit that in the master
branch.
This reverts commit 4fdb8acdae.
Forward-port the comments from commit 01e2920 (v0.10) to the master
branch. Everything else from that patch already exists in master.
It didn't merge cleanly because lib/http.js has been split up in
several files.
Several people have reported issues with IIS and Resin servers (or maybe
SSL terminators sitting in front of those servers) that are fixed by
downgrading OpenSSL. The AESNI performance improvements were nice but
stability is more important. Downgrade OpenSSL from 1.0.1e to 1.0.0f.
Fixes#5360 (and others).
When developer calls setBreakpoint with an unknown script name,
we convert the script name into regular expression matching all
paths ending with given name (name can be a relative path too).
To create such breakpoint in V8, we use type `scriptRegEx`
instead of `scriptId` for `setbreakpoint` request.
To restore such breakpoint, we save the original script name
send by the user. We use this original name to set (restore)
breakpoint in the new child process.
Fixed a bug in debugger repl where `restart` command did not work
when a custom debug port was specified via command-line option
--port={number}.
File test/simple/helper-debugger-repl.js was extracted
from test/simple/test-debugger-repl.js
Fixed a bug in debugger repl where `restart` command did not work
when a custom debug port was specified via command-line option
--port={number}.
File test/simple/helper-debugger-repl.js was extracted
from test/simple/test-debugger-repl.js
Errors in leaf child processes weren't picked up by the test runner
because they didn't get bubbled up to the main process. Don't forcibly
kill the child processes; tell them to quit gracefully, then inspect
their exit codes.
This change introduces support for the common PREFIX variable in the
Makefile and install.py, instead of having /usr/local hardcoded. This
makes it much easier to install node to custom locations e.g. in a
user's home directory.
The PREFIX variable defaults to /usr/local.
The uname function can return any non-negative int to indicate success.
Strange, but that's how it is documented. This also fixes a similar
buffer overflow in the even more unlikely event that info.release is
> 255 characters, similar to how 78c5de5 did for info.sysname.
Fixes#3740
In the case of pipelined requests, you can have a situation where
the socket gets destroyed via one req/res object, but then trying
to destroy *another* req/res on the same socket will cause it to
call undefined.destroy(), since it was already removed from that
message.
Add a guard to OutgoingMessage.destroy and IncomingMessage.destroy
to prevent this error.
Fixes#3740
In the case of pipelined requests, you can have a situation where
the socket gets destroyed via one req/res object, but then trying
to destroy *another* req/res on the same socket will cause it to
call undefined.destroy(), since it was already removed from that
message.
Add a guard to OutgoingMessage.destroy and IncomingMessage.destroy
to prevent this error.
It needs to apply the Transform class when the _readableState,
_writableState, or _transformState properties are accessed,
otherwise things like setEncoding and on('data') don't work
properly.
Also, the methods wrappers are no longer needed, since they're only
problematic because they access the undefined properties.
Clean up and DRY the cluster source code. Fix a few bugs while we're
here:
* Short-lived handles in long-lived worker processes were never
reclaimed, resulting in resource leaks.
* Handles in the master process are now closed when the last worker
that holds a reference to them quits. Previously, they were only
closed at cluster shutdown.
* The cluster object no longer exposes functions/properties that are
only valid in the 'other' process, e.g. cluster.fork() is no longer
exported in worker processes.
So much goodness and still manages to reduce the line count from 590
to 320.
An absolute path will always open the same location regardless of your
current working directory. For posix, this just means path.charAt(0) ===
'/', but on Windows it's a little more complicated.
Fixesjoyent/node#5299.
V8 was upgraded to 3.18 in commit 9f68226. The knobs that control the
ARM build have changed in a number of ways. This commit patches the
configure script to reflect that. Should fix the Raspberry Pi build.
Fixes#5329.
4716dc6 made assert.equal() and related functions work better by
generating a better toString() from the expected, actual, and operator
values passed to fail(). Unfortunately, this was accomplished by putting
the generated message into the error's `name` property. When you passed
in a custom error message, the error would put the custom error into
`name` *and* `message`, resulting in helpful string representations like
"AssertionError: Oh no: Oh no".
This commit resolves that issue by storing the generated message in the
`message` property while leaving the error's name alone and adding
a regression test so that this doesn't pop back up later.
Closes#5292.
I broke dgram.Socket#bind(port, cb) almost a year ago in 332fea5a but
it wasn't until today that someone complained and none of the tests
caught it because they all either specify the address or omit the
callback.
Anyway, now it works again and does what you expect: it binds the
socket to the "any" address ("0.0.0.0" for IPv4 and "::" for IPv6.)
process.stdout isn't fully initialized yet by the time the test starts
when invoked with `python tools/test.py`. Use process.stdin instead and
force initialization with process.stdin.resume().
Fix a NULL pointer dereference in src/handle_wrap.cc which is really a
use-after-close bug.
The test checks that unref() after close() works on process.stdout but
this bug affects everything that derives from HandleWrap. I discovered
it because child processes would sometimes quit for no reason (that is,
no reason until I turned on core dumps.)
When LD_LIBRARY_PATH is overriden for custom builds we need to preserve
it for child processes. To be sure we preserve whole environment of
parent process and just add TEST_INIT variable to it.
When LD_LIBRARY_PATH is overriden for custom builds we need to preserve
it for forked process. There are possibly other environment variables
that could cause test failures so we preserve whole environment of
parent process.
Fix a (rather academic) buffer overflow. MAXHOSTNAMELEN is 256 on most
platforms, which means the buffer wasn't big enough to hold the
trailing nul byte on a system with a maximum length hostname.
Make http.request() and friends escape unsafe characters in the request
path. That is, a request for '/foo bar' is now escaped as '/foo%20bar'.
Before this commit, the path was used as-is in the request status line,
creating an invalid HTTP request ("GET /foo bar HTTP/1.1").
Fixes#4381.
Fix#5272
The consumption of a readable stream is a dance with 3 partners.
1. The specific stream Author (A)
2. The Stream Base class (B), and
3. The Consumer of the stream (C)
When B calls the _read() method that A implements, it sets a 'reading'
flag, so that parallel calls to _read() can be avoided. When A calls
stream.push(), B knows that it's safe to start calling _read() again.
If the consumer C is some kind of parser that wants in some cases to
pass the source stream off to some other party, but not before "putting
back" some bit of previously consumed data (as in the case of Node's
websocket http upgrade implementation). So, stream.unshift() will
generally *never* be called by A, but *only* called by C.
Prior to this patch, stream.unshift() *also* unset the state.reading
flag, meaning that C could indicate the end of a read, and B would
dutifully fire off another _read() call to A. This is inappropriate.
In the case of fs streams, and other variably-laggy streams that don't
tolerate overlapped _read() calls, this causes big problems.
Also, calling stream.shift() after the 'end' event did not raise any
kind of error, but would cause very strange behavior indeed. Calling it
after the EOF chunk was seen, but before the 'end' event was fired would
also cause weird behavior, and could lead to data being lost, since it
would not emit another 'readable' event.
This change makes it so that:
1. stream.unshift() does *not* set state.reading = false
2. stream.unshift() is allowed up until the 'end' event.
3. unshifting onto a EOF-encountered and zero-length (but not yet
end-emitted) stream will defer the 'end' event until the new data is
consumed.
4. pushing onto a EOF-encountered stream is now an error.
So, if you read(), you have that single tick to safely unshift() data
back into the stream, even if the null chunk was pushed, and the length
was 0.
Don't scan the whole string for a "NODE_" substring, just check that
the string starts with the expected prefix.
This is a reprise of dbbfbe7 but this time for the child_process
module.
Don't scan the whole string for a "NODE_CLUSTER_" substring, just check
that the string starts with the expected prefix. The linear scan was
causing a noticeable (but unsurprising) slowdown on messages with a
large .cmd string property.
Call SetPointerInInternalField(0, NULL) rather than
SetInternalField(0, Undefined()).
Fixes the following spurious NULL pointer dereference in debug builds:
#0 0x03ad2821 in v8::internal::FixedArrayBase::length ()
#1 0x03ad1dfc in v8::internal::FixedArray::get ()
#2 0x03ae05dd in v8::internal::Context::global_object ()
#3 0x03b6b87d in v8::internal::Context::builtins ()
#4 0x03ae1871 in v8::internal::Isolate::js_builtins_object ()
#5 0x03ab4fab in v8::CallV8HeapFunction ()
#6 0x03ab4d4a in v8::Value::Equals ()
#7 0x03b4f38b in CheckEqualsHelper ()
#8 0x03ac0f4b in v8::Object::SetInternalField ()
#9 0x06a99ddd in node::ObjectWrap::~ObjectWrap ()
#10 0x06a8b051 in node::Buffer::~Buffer ()
#11 0x06a8afbb in node::Buffer::~Buffer ()
#12 0x06a8af5e in node::Buffer::~Buffer ()
#13 0x06a9e569 in node::ObjectWrap::WeakCallback ()
We should go to next buffer if *current* one is full, not the next one.
Otherwise we may hop through buffers and written data will become
interleaved, which will lead to failure.
This change shouldn't have landed in the stable branch. It's a feature,
not a bug fix.
This reverts commit 58f93ffc4a.
This reverts commit 8c8ebe49b6.
This reverts commit ba0f7b8066.
This reverts commit 21f3c5c367.
Buffer.byteLength() works only for string inputs. Thus, when connection
has pending Buffer to write, it should just use it's length instead of
throwing exception.
Since 049903e, an end callback could be called before a write
callback if end() is called before the write is done. This patch
resolves the issue.
In collaboration with @gne
Fixesfelixge/node-formidable#209Fixes#5215
We were assuming that any string can be concatenated safely to
CRLF. However, for hex, base64, or binary encoded writes, this
is not the case, and results in sending the incorrect response.
An unusual edge case, but certainly a bug.
DH_compute_secret() may return key that is smaller than input buffer,
in such cases key should be left-padded because it is a BN (big number).
fix#5239
We should go to next buffer if *current* one is full, not the next one.
Otherwise we may hop through buffers and written data will become
interleaved, which will lead to failure.
On Linux, positional writes don't work when the file is opened in
append mode. The kernel ignores the position argument and always
appends the data to the end of the file.
To quote the man page:
POSIX requires that opening a file with the O_APPEND flag should have
no affect on the location at which pwrite() writes data. However, on
Linux, if a file is opened with O_APPEND, pwrite() appends data to the
end of the file, regardless of the value of offset.
RFC 6125 explicitly states that a client "MUST NOT seek a match
for a reference identifier of CN-ID if the presented identifiers
include a DNS-ID, SRV-ID, URI-ID, or any application-specific
identifier types supported by the client", but it MAY do so if
none of the mentioned identifier types (but others) are present.
Always define v8_postmortem_support, even if the platform does not
support it. Commit d8852aa adds a rule that references it in node.gyp.
Fixes the Windows build.
The DTrace probes were updated to accomodate platforms that can't
handle structs, update the prototypes for ETW but it's not necessary
to do anything with the new arguments as it's redundant information.
OSX and other DTrace implementations don't support dereferencing
structs in probes. To accomodate that pass members from the struct as
arguments so that DTrace is useful on those systems.
If an http response has an 'end' handler that throws, then the socket
will never be released back into the pool.
Granted, we do NOT guarantee that throwing will never have adverse
effects on Node internal state. Such a guarantee cannot be reasonably
made in a shared-global mutable-state side-effecty language like
JavaScript. However, in this case, it's a rather trivial patch to
increase our resilience a little bit, so it seems like a win.
There is no semantic change in this case, except that some event
listeners are removed, and the `'free'` event is emitted on nextTick, so
that you can schedule another request which will re-use the same socket.
From the user's point of view, there should be no detectable difference.
Closes#5107
The tests did not agree with the test comments. Tests first and second
were both testing the !state.reading case. Now second tests the
state.reading && state.length case.
Fixesjoyent/node#5183
The v0.8 Stream.pipe() method automatically destroyed the destination
stream whenever the src stream closed. However, this caused a lot of
problems, and was removed by popular demand. (Many userland modules
still have a no-op destroy() method just because of this.) It was also
very hazardous because this would be done even if { end: false } was
passed in the pipe options.
In v0.10, we decided that the 'close' event and destroy() method are
application-specific, and pipe() doesn't automatically call destroy().
However, TLS actually depended (silently) on this behavior. So, in this
case, we should just go ahead and destroy the thing when close happens.
Closes#5145
The DTrace probes were updated to accomodate platforms that can't
handle structs, update the prototypes for ETW but it's not necessary
to do anything with the new arguments as it's redundant information.
Expand the JSON representation of Buffer to include type information
so that it can be deserialized in JSON.parse() without context.
Fixes#5110.
Fixes#5143.
OSX and other DTrace implementations don't support dereferencing
structs in probes. To accomodate that pass members from the struct as
arguments so that DTrace is useful on those systems.
Pass the Isolate to Persistent<Function>::New(). Fixes the following
warning:
../../src/node.cc: In function ‘v8::Handle<v8::Value>
node::UsingDomains(const v8::Arguments&)’:
../../src/node.cc:921: warning: ‘New’ is deprecated
declared at ../../deps/v8/include/v8.h:4438)
In cases where a stream may have data added to the read queue before the
user adds a 'readable' event, there is never any indication that it's
time to start reading.
True, there's already data there, which the user would get if they
checked However, as we use 'readable' event listening as the signal to
start the flow of data with a read(0) call internally, we ought to
trigger the same effect (ie, emitting a 'readable' event) even if the
'readable' listener is added after the first emission.
To avoid confusing weirdness, only the *first* 'readable' event listener
is granted this privileged status. After we've started the flow (or,
alerted the consumer that the flow has started) we don't need to start
it again. At that point, it's the consumer's responsibility to consume
the stream.
Closes#5141
Have the formatter filter out vt100 color codes when calculating the
line width. Stops it from unnecessarily splitting strings over multiple
lines.
Fixes#5039.
A llvm/clang bug on Darwin ia32 makes these tests fail 100% of
the time. Since no one really seems to mind overly much, and we
can't reasonably fix this in node anyway, just accept both types
of NaN for now.
Calling `this.pair.encrypted._internallyPendingBytes()` before
handling/resetting error will result in assertion failure:
../src/node_crypto.cc:962: void node::crypto::Connection::ClearError():
Assertion `handle_->Get(String::New("error"))->BooleanValue() == false'
failed.
see #5058
Microsoft's IIS doesn't support it, and is not replying with ServerHello
after receiving ClientHello which contains it.
The good way might be allowing to opt-out this at runtime from
javascript-land, but unfortunately OpenSSL doesn't support it right now.
see #5119
Since _tickCallback and _tickDomainCallback were both called from
MakeCallback, it was possible for a callback to be called that required
a domain directly to _tickCallback.
The fix was to implement process.usingDomains(). This will set all
applicable functions to their domain counterparts, and set a flag in cc
to let MakeCallback know domain callbacks always need to be checked.
Added test in own file. It's important that the test remains isolated.
_charsWritten is an internal property that was constantly written to,
but never read from. So it has been removed.
Removed documentation reference as well.
Add the `sessionTimeout` integral value to the list of options
recognized by `tls.createServer`.
This option will be useful for applications which need frequently
establish short-lived TLS connections to the same endpoint. The TLS
tickets RFC is an ideal option to reduce the socket setup overhead
for such scenarios, but the default ticket timeout value (5
minutes) is too low to be useful.
It's possible to read multiple messages off the parent/child channel.
When that happens, make sure that recvHandle is cleared after emitting
the first message so it doesn't get emitted twice.
Commit f53441a added crypto.getCiphers() as a function that returns the
names of SSL ciphers.
Commit 14a6c4e then added crypto.getHashes(), which returns the names of
digest algorithms, but that creates a subtle inconsistency: the return
values of crypto.getHashes() are valid arguments to crypto.createHash()
but that is not true for crypto.getCiphers() - the returned values are
only valid for SSL/TLS functions.
Rectify that by adding tls.getCiphers() and making crypto.getCiphers()
return proper cipher names.
In process#send() and child_process.ChildProcess#send(), use 'utf8' as
the encoding instead of 'ascii' because 'ascii' mutilates non-ASCII
input. Correctly handle partial character sequences by introducing
a StringDecoder.
Sending over UTF-8 no longer works in v0.10 because the high bit of
each byte is now cleared when converting a Buffer to ASCII. See
commit 96a314b for details.
Fixes#4999 and #5011.
Commit 8632af3 ("tools: update gyp to r1601") broke the Windows build.
Older versions of GYP link to kernel32.lib, user32.lib, etc. but that
was changed in r1584. See https://codereview.chromium.org/12256017
Fix the build by explicitly linking to the required libraries.
While libuv supports reporting subsecond stat resolution across
platforms, to actually get that resolution your platform and filesystem
must support it (not HFS, ext[23], fat), otherwise the nsecs are 0
The EncIn, EncOut, ClearIn & ClearOut functions are victims of some code
copy + pasting. A common line copied to all of them is:
`if (off >= buffer_length) { ...`
448e0f43 corrected ClearIn's check from `>=` to `>`, but left the others
unchanged (with an incorrect bounds check). However, if you look down at
the next very next bounds check you'll see:
`if (off + len > buffer_length) { ...`
So the check is actually obviated by the next line, and should be
removed.
This fixes an issue where writing a zero-length buffer to an encrypted
pair's *encrypted* stream you would get a crash.
The EncIn, EncOut, ClearIn & ClearOut functions are victims of some code
copy + pasting. A common line copied to all of them is:
`if (off >= buffer_length) { ...`
448e0f43 corrected ClearIn's check from `>=` to `>`, but left the others
unchanged (with an incorrect bounds check). However, if you look down at
the next very next bounds check you'll see:
`if (off + len > buffer_length) { ...`
So the check is actually obviated by the next line, and should be
removed.
This fixes an issue where writing a zero-length buffer to an encrypted
pair's *encrypted* stream you would get a crash.
Increase the number of bits by 1 by making Flags unsigned.
BUG=chromium:211741
Review URL: https://chromiumcodereview.appspot.com/12886008
This is a back-port of commits 13964 and 13988 addressing CVE-2013-2632.
Throw a TypeError if size > 0x3fffffff. Avoids the following V8 fatal
error:
FATAL ERROR: v8::Object::SetIndexedPropertiesToExternalArrayData()
length exceeds max acceptable value
Fixes#5126.
The stall is exposed in the test, though the test itself asserts before
it stalls.
The test is constructed to replicate the stalling state of a complex
Passthrough usecase since I was not able to reliable trigger the stall.
Some of the preconditions for triggering the stall are:
* rs.length >= rs.highWaterMark
* !rs.needReadable
* _transform() handler that can return empty transforms
* multiple sync write() calls
Combined this can trigger a case where rs.reading is not cleared when
further progress requires this. The fix is to always clear rs.reading.
Before this patch calling `socket.setTimeout(0xffffffff)` will result in
signed int32 overflow in C++ which resulted in assertion error:
Assertion failed: (timeout >= -1), function uv__io_poll, file
../deps/uv/src/unix/kqueue.c, line 121.
see #5101
Since WriteBuffer has been replaced with WriteOneByte, writing ascii
will no longer automatically convert 0x0 to 0x20. So removed mention of
this special case from docs.
Every constant is certainly 4 bytes now, but freebsd's objdump utility
prints out odd byte sequences (5-bytes, 6-bytes and even 9-bytes long)
for v8's data section. We can safely ignore all upper bytes, because all
constants that we're using are just `int`s. Since on all supported
platforms `int` is 32bit long (and anyway v8's constants are 32bit too),
we ignore all higher bits if they were read.
All compile time warnings about using deprecated APIs have been
suppressed by updating node's API. Though there are still many function
calls that can accept Isolate, and still need to be updated.
node_isolate had to be added as an extern variable in node.h and
node_object_wrap.h
Also a couple small fixes for Error handling.
Before v8 3.16.6 the error stack message was lazily written when it was
needed, which allowed you to change the message after instantiation.
Then the stack would be written with the new message the first time it
was accessed. Though that has changed. Now it creates the stack message
on instantiation. So setting a different message afterwards won't be
displayed.
This is not a complete fix for the problem. Getting error without any
message isn't very useful.
Fixes#5071, #5073.
* Normalize capitalization of drive letter
* Fix `exit()` typo in failure path
* Ignore symlink tests (Windows) if not elevated
The `test_relative_input_cwd()` test was failing on Windows when
`skipSymlinks` was `true`. So we won't run it if `skipSymlinks` is
`true`.
When it failed, the unhandled error caused Node to die before
having a chance to clean up, which resulted in two files missing
in subsequent unit tests:
* `test/fixtures/nested-index/one/hello.js`
* `test/fixtures/nested-index/one/index.js`
We should probably find a way to isolate this test from the other
test (`simple/test-module-loading`) that was failing when this test
poluted the disk state.
From OpenSSL's documentation:
"If BIO_free() is called on a BIO chain it will only free one BIO
resulting in a memory leak."
and
"BIO_free_all() frees up an entire BIO chain, it does not halt if an
error occurs freeing up an individual BIO in the chain"
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
