platform/upstream/nodejs.git
11 years ago2013.05.14, Version 0.10.6 (Stable) v0.10.6
isaacs [Tue, 14 May 2013 18:42:16 +0000 (11:42 -0700)]
2013.05.14, Version 0.10.6 (Stable)

* module: Deprecate require.extensions (isaacs)

* stream: make Readable.wrap support objectMode, empty streams (Daniel Moore)

* child_process: fix handle delivery (Ben Noordhuis)

* crypto: Fix performance regression (isaacs)

* src: DRY string encoding/decoding (isaacs)

11 years agocrypto: Pass encodings for Hmac digest
isaacs [Fri, 3 May 2013 01:38:04 +0000 (18:38 -0700)]
crypto: Pass encodings for Hmac digest

11 years agocrypto: Pass encodings to C++ for Sign/Verify
isaacs [Fri, 3 May 2013 01:34:25 +0000 (18:34 -0700)]
crypto: Pass encodings to C++ for Sign/Verify

11 years agocrypto: use StringBytes::Encode
isaacs [Thu, 2 May 2013 22:32:32 +0000 (15:32 -0700)]
crypto: use StringBytes::Encode

11 years agocrypto: Pass strings to binding layer directly
isaacs [Thu, 2 May 2013 19:28:48 +0000 (12:28 -0700)]
crypto: Pass strings to binding layer directly

11 years agobuffer: DRY string encoding using StringBytes
isaacs [Thu, 2 May 2013 17:44:45 +0000 (10:44 -0700)]
buffer: DRY string encoding using StringBytes

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.

11 years agosrc: use StringBytes for DecodeWrite/DecodeBytes/Encode
isaacs [Thu, 2 May 2013 17:49:20 +0000 (10:49 -0700)]
src: use StringBytes for DecodeWrite/DecodeBytes/Encode

Bonus: this makes node::Encode actually work properly with base64,
ucs2, hex, etc.

11 years agosrc: Use StringBytes in StreamWrap
isaacs [Wed, 1 May 2013 22:40:13 +0000 (15:40 -0700)]
src: Use StringBytes in StreamWrap

11 years agosrc: Add StringBytes static util class
isaacs [Wed, 1 May 2013 22:40:43 +0000 (15:40 -0700)]
src: Add StringBytes static util class

Four functions:

- StringBytes::StorageSize()
- StringBytes::Size()
- StringBytes::Write()
- StringBytes::Encode()

11 years agotools: remove unnecessary cpplint rules
isaacs [Wed, 1 May 2013 22:36:09 +0000 (15:36 -0700)]
tools: remove unnecessary cpplint rules

We don't actually care about header order much, and since we never use
stl classes, 'string' isn't an issue for node ever.

11 years agostream: Make default encoding configurable
isaacs [Thu, 2 May 2013 19:25:50 +0000 (12:25 -0700)]
stream: Make default encoding configurable

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.

11 years agostream: don't create unnecessary buffers in Readable
isaacs [Tue, 30 Apr 2013 22:09:54 +0000 (15:09 -0700)]
stream: don't create unnecessary buffers in Readable

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.

11 years agolint
isaacs [Tue, 30 Apr 2013 22:02:19 +0000 (15:02 -0700)]
lint

11 years agobenchmark: hash stream
isaacs [Tue, 30 Apr 2013 21:50:21 +0000 (14:50 -0700)]
benchmark: hash stream

11 years agoblog: Release v0.11.2
isaacs [Tue, 14 May 2013 18:35:43 +0000 (11:35 -0700)]
blog: Release v0.11.2

11 years agotest: increase workers to 8 in cluster-disconnect
Benoit Vallée [Tue, 30 Apr 2013 04:25:23 +0000 (12:25 +0800)]
test: increase workers to 8 in cluster-disconnect

Increasing the number of workers from 2 to 8 makes this test
more likely to trigger race conditions. See #5330 for background.

11 years agodoc: clarify subsystems in CONTRIBUTING.md
Ben Noordhuis [Tue, 14 May 2013 10:31:38 +0000 (12:31 +0200)]
doc: clarify subsystems in CONTRIBUTING.md

11 years agochild_process: fix handle delivery
Ben Noordhuis [Fri, 10 May 2013 18:19:17 +0000 (11:19 -0700)]
child_process: fix handle delivery

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.

11 years agodoc: s/search.npmjs.org/npmjs.org/
isaacs [Thu, 9 May 2013 22:15:39 +0000 (15:15 -0700)]
doc: s/search.npmjs.org/npmjs.org/

11 years agotest: fix pummel/test-net-many-clients.js
Timothy J Fontaine [Mon, 6 May 2013 23:37:03 +0000 (16:37 -0700)]
test: fix pummel/test-net-many-clients.js

client sockets no longer emit 'connect' event inside the
requestListener, update test-net-many-clients to reflect that

11 years agodoc: document stream.Writable 'error' event
Robert Kowalski [Mon, 6 May 2013 19:53:35 +0000 (21:53 +0200)]
doc: document stream.Writable 'error' event

Fixes #5255.

11 years agostream: make Readable.wrap support empty streams
Daniel Moore [Wed, 1 May 2013 17:46:31 +0000 (13:46 -0400)]
stream: make Readable.wrap support empty streams

This makes Readable.wrap behave properly when the wrapped stream ends
before emitting any data events.

11 years agostream: make Readable.wrap support objectMode
Daniel Moore [Mon, 29 Apr 2013 13:25:08 +0000 (09:25 -0400)]
stream: make Readable.wrap support objectMode

Added a check to see if the stream is in objectMode before deciding
whether to include or exclude data from an old-style wrapped stream.

11 years agotest: don't use total_seconds() because of py2.6
Timothy J Fontaine [Tue, 7 May 2013 22:17:33 +0000 (15:17 -0700)]
test: don't use total_seconds() because of py2.6

11 years agotest: report test duration in TAP progress
Timothy J Fontaine [Tue, 7 May 2013 18:59:22 +0000 (11:59 -0700)]
test: report test duration in TAP progress

11 years agobuild: only use DESTDIR instead of PREFIX for pkg
Timothy J Fontaine [Thu, 2 May 2013 20:54:37 +0000 (13:54 -0700)]
build: only use DESTDIR instead of PREFIX for pkg

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

11 years agodoc: update options for exec and execFile
Kevin Locke [Fri, 26 Apr 2013 23:14:39 +0000 (17:14 -0600)]
doc: update options for exec and execFile

The stdio and customFds options are never used by exec or execFile,
remove them from the documentation for these functions.

11 years agobuild: vcbuild.bat fix for Visual Studio 2012
Miroslav Bajtoš [Fri, 3 May 2013 06:51:51 +0000 (08:51 +0200)]
build: vcbuild.bat fix for Visual Studio 2012

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.

11 years agodoc: document return values of EventEmitter methods
Sam Roberts [Tue, 30 Apr 2013 22:24:48 +0000 (15:24 -0700)]
doc: document return values of EventEmitter methods

11 years agodoc: link joyent logo in website footer
isaacs [Thu, 2 May 2013 16:47:46 +0000 (09:47 -0700)]
doc: link joyent logo in website footer

11 years agodebugger: breakpoints in scripts not loaded yet
Miroslav Bajtoš [Fri, 26 Apr 2013 06:05:33 +0000 (08:05 +0200)]
debugger: breakpoints in scripts not loaded yet

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.

11 years agoRevert "deps: downgrade openssl to v1.0.0f"
Ben Noordhuis [Wed, 1 May 2013 14:21:21 +0000 (16:21 +0200)]
Revert "deps: downgrade openssl to v1.0.0f"

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 4fdb8acdaef4c3cb1d855e992ada0e63fee520a6.

11 years agodoc: Fix require.extensions documentation
isaacs [Tue, 30 Apr 2013 14:40:43 +0000 (07:40 -0700)]
doc: Fix require.extensions documentation

1. The stability index must come first, or it messes up the markdown
2. require.extensions is an Object, not an Array.

Close #5387

11 years agodeps: downgrade openssl to v1.0.0f
Ben Noordhuis [Fri, 26 Apr 2013 12:49:54 +0000 (14:49 +0200)]
deps: downgrade openssl to v1.0.0f

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).

11 years agodoc: cluster: s/server.destroy/server.close/
Ben Noordhuis [Mon, 29 Apr 2013 10:06:30 +0000 (12:06 +0200)]
doc: cluster: s/server.destroy/server.close/

Fixes #5379.

11 years agodoc: Deprecate require.extensions
isaacs [Mon, 29 Apr 2013 05:10:24 +0000 (22:10 -0700)]
doc: Deprecate require.extensions

11 years agodebugger: `restart` with custom debug port
Miroslav Bajtoš [Fri, 26 Apr 2013 19:07:56 +0000 (21:07 +0200)]
debugger: `restart` with custom debug port

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

11 years agoChangeLog: Merge in v0.8
isaacs [Tue, 23 Apr 2013 21:43:46 +0000 (14:43 -0700)]
ChangeLog: Merge in v0.8

Close #5355

11 years agoblog: Post for v0.10.5
isaacs [Tue, 23 Apr 2013 21:08:01 +0000 (14:08 -0700)]
blog: Post for v0.10.5

11 years agoNow working on 0.10.6
isaacs [Tue, 23 Apr 2013 21:07:23 +0000 (14:07 -0700)]
Now working on 0.10.6

11 years agoMerge branch 'v0.10.5-release' into v0.10
isaacs [Tue, 23 Apr 2013 21:07:09 +0000 (14:07 -0700)]
Merge branch 'v0.10.5-release' into v0.10

11 years ago2013.04.23, Version 0.10.5 (Stable) v0.10.5
isaacs [Tue, 23 Apr 2013 01:00:41 +0000 (18:00 -0700)]
2013.04.23, Version 0.10.5 (Stable)

* uv: Upgrade to 0.10.5 (isaacs)

* build: added support for Visual Studio 2012 (Miroslav Bajtoš)

* http: Don't try to destroy nonexistent sockets (isaacs)

* crypto: LazyTransform on properties, not methods (isaacs)

* assert: put info in err.message, not err.name (Ryan Doenges)

* dgram: fix no address bind() (Ben Noordhuis)

* handle_wrap: fix NULL pointer dereference (Ben Noordhuis)

* os: fix unlikely buffer overflow in os.type() (Ben Noordhuis)

* stream: Fix unshift() race conditions (isaacs)

11 years agobuild: added support for Visual Studio 2012
Miroslav Bajtoš [Tue, 23 Apr 2013 07:38:17 +0000 (09:38 +0200)]
build: added support for Visual Studio 2012

11 years agouv: Upgrade to 0.10.5
isaacs [Tue, 23 Apr 2013 00:39:05 +0000 (17:39 -0700)]
uv: Upgrade to 0.10.5

11 years agoos: Fix uname() error handling on sunos
isaacs [Mon, 22 Apr 2013 17:18:46 +0000 (10:18 -0700)]
os: Fix uname() error handling on sunos

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.

11 years agohttp: Don't try to destroy nonexistent sockets
isaacs [Mon, 22 Apr 2013 15:52:42 +0000 (08:52 -0700)]
http: Don't try to destroy nonexistent sockets

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.

11 years agotest: fix dgram-bind-default-address on osx
isaacs [Fri, 19 Apr 2013 15:32:24 +0000 (08:32 -0700)]
test: fix dgram-bind-default-address on osx

Allow the IPv4-mapped-as-IPv6 style address.

11 years agocrypto: LazyTransform on properties, not methods
isaacs [Mon, 8 Apr 2013 15:43:20 +0000 (08:43 -0700)]
crypto: LazyTransform on properties, not methods

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.

11 years agosrc: fix potential memory leak on early return
mscdex [Sun, 21 Apr 2013 03:21:13 +0000 (23:21 -0400)]
src: fix potential memory leak on early return

11 years agosrc: don't initialize variable before assignment
mscdex [Sun, 21 Apr 2013 03:21:05 +0000 (23:21 -0400)]
src: don't initialize variable before assignment

11 years agoblog: v0.11.1 does not work on windows x64
isaacs [Fri, 19 Apr 2013 16:21:29 +0000 (09:21 -0700)]
blog: v0.11.1 does not work on windows x64

11 years agoblog: Post about 0.11.1
isaacs [Fri, 19 Apr 2013 16:12:34 +0000 (09:12 -0700)]
blog: Post about 0.11.1

11 years agoblog: Post about 0.10.4
isaacs [Fri, 19 Apr 2013 16:12:22 +0000 (09:12 -0700)]
blog: Post about 0.10.4

11 years agodoc: document value of `this` inside listeners
Sean Silva [Thu, 18 Apr 2013 22:34:12 +0000 (18:34 -0400)]
doc: document value of `this` inside listeners

Fixes #5326.

11 years agoassert: put info in err.message, not err.name
Ryan Doenges [Sun, 14 Apr 2013 01:48:00 +0000 (18:48 -0700)]
assert: put info in err.message, not err.name

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.

11 years agowebsite: add link to nightlies on download page
Ben Noordhuis [Thu, 18 Apr 2013 20:06:03 +0000 (22:06 +0200)]
website: add link to nightlies on download page

11 years agocrypto: fix return Local Handle w/o scope.Close()
Trevor Norris [Wed, 17 Apr 2013 18:45:13 +0000 (11:45 -0700)]
crypto: fix return Local Handle w/o scope.Close()

A new String was being created and returned, but was not sent through
the scope.Close(), which caused it to be cleaned up before being
returned.

11 years agodgram: fix no address bind()
Ben Noordhuis [Wed, 17 Apr 2013 22:44:53 +0000 (00:44 +0200)]
dgram: fix no address bind()

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.)

11 years agotest: make stdout-close-unref work in test runner
Ben Noordhuis [Wed, 17 Apr 2013 21:59:49 +0000 (23:59 +0200)]
test: make stdout-close-unref work in test runner

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().

11 years agodocs: update path.join() example for v0.10
Kelly Gerber [Tue, 9 Apr 2013 20:36:34 +0000 (13:36 -0700)]
docs: update path.join() example for v0.10

The current example shows the behavior of v0.8. In v0.10 arguments
to path.join() must be strings; otherwise, an exception is thrown.

11 years agodoc: note a gotcha with http.Server sockets
Ryan Graham [Tue, 16 Apr 2013 20:52:52 +0000 (13:52 -0700)]
doc: note a gotcha with http.Server sockets

11 years agohandle_wrap: fix NULL pointer dereference
Ben Noordhuis [Tue, 16 Apr 2013 20:59:35 +0000 (22:59 +0200)]
handle_wrap: fix NULL pointer dereference

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.)

11 years agotest: preserve process.env after test-init exec
Stanislav Ochotnicky [Tue, 16 Apr 2013 07:51:56 +0000 (09:51 +0200)]
test: preserve process.env after test-init exec

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.

11 years agotest: preserve process.env in forked child_process
Stanislav Ochotnicky [Wed, 10 Apr 2013 13:38:02 +0000 (15:38 +0200)]
test: preserve process.env in forked child_process

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.

11 years agoos: unbreak windows build
Ben Noordhuis [Mon, 15 Apr 2013 20:36:49 +0000 (22:36 +0200)]
os: unbreak windows build

Windows doesn't have MAXHOSTNAMELEN. Introduced in afbadde.

11 years agoos: handle 256 character hostnames
Ben Noordhuis [Mon, 15 Apr 2013 19:05:20 +0000 (21:05 +0200)]
os: handle 256 character hostnames

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.

11 years agoos: fix unlikely buffer overflow in os.type()
Ben Noordhuis [Mon, 15 Apr 2013 18:51:28 +0000 (20:51 +0200)]
os: fix unlikely buffer overflow in os.type()

* Fix a buffer overflow that happens iff strlen(info.sysname) > 255.
* Check the return value of uname().

11 years agobuild: Typo in tools/msvs/msi/product.wxs
isaacs [Fri, 12 Apr 2013 03:40:05 +0000 (20:40 -0700)]
build: Typo in tools/msvs/msi/product.wxs

11 years agostream: Fix unshift() race conditions
isaacs [Thu, 11 Apr 2013 22:01:26 +0000 (15:01 -0700)]
stream: Fix unshift() race conditions

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.

11 years agoNow working on v0.10.5
isaacs [Thu, 11 Apr 2013 18:07:08 +0000 (11:07 -0700)]
Now working on v0.10.5

11 years agoMerge branch 'v0.10.4-release' into v0.10
isaacs [Thu, 11 Apr 2013 18:06:37 +0000 (11:06 -0700)]
Merge branch 'v0.10.4-release' into v0.10

11 years agolint
isaacs [Thu, 11 Apr 2013 18:06:07 +0000 (11:06 -0700)]
lint

11 years agoblog: Fix title for v0.8.23 release
isaacs [Thu, 11 Apr 2013 18:05:53 +0000 (11:05 -0700)]
blog: Fix title for v0.8.23 release

11 years ago2013.04.11, Version 0.10.4 (Stable) v0.10.4
isaacs [Thu, 11 Apr 2013 16:39:16 +0000 (09:39 -0700)]
2013.04.11, Version 0.10.4 (Stable)

* uv: Upgrade to 0.10.4

* npm: Upgrade to 1.2.18

* v8: Avoid excessive memory growth in JSON.parse (Fedor Indutny)

* child_process, cluster: fix O(n*m) scan of cmd string (Ben Noordhuis)

* net: fix socket.bytesWritten Buffers support (Fedor Indutny)

* buffer: fix offset checks (Łukasz Walukiewicz)

* stream: call write cb before finish event (isaacs)

* http: Support write(data, 'hex') (isaacs)

* crypto: dh secret should be left-padded (Fedor Indutny)

* process: expose NODE_MODULE_VERSION in process.versions (Rod Vagg)

* crypto: fix constructor call in crypto streams (Andreas Madsen)

* net: account for encoding in .byteLength (Fedor Indutny)

* net: fix buffer iteration in bytesWritten (Fedor Indutny)

* crypto: zero is not an error if writing 0 bytes (Fedor Indutny)

* tls: Re-enable check of CN-ID in cert verification (Tobias Müllerleile)

11 years agonpm: Upgrade to 1.2.18
isaacs [Thu, 11 Apr 2013 16:16:47 +0000 (09:16 -0700)]
npm: Upgrade to 1.2.18

11 years agouv: Upgrade to v0.10.4
isaacs [Thu, 11 Apr 2013 16:07:22 +0000 (09:07 -0700)]
uv: Upgrade to v0.10.4

11 years agochild_process: fix O(n*m) scan of cmd string
Ben Noordhuis [Thu, 11 Apr 2013 11:50:45 +0000 (13:50 +0200)]
child_process: fix O(n*m) scan of cmd string

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.

11 years agocluster: fix O(n*m) scan of cmd string
Ben Noordhuis [Thu, 11 Apr 2013 11:18:02 +0000 (13:18 +0200)]
cluster: fix O(n*m) scan of cmd string

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.

11 years agosrc: don't SetInternalField() in ObjectWrap dtor
Ben Noordhuis [Wed, 10 Apr 2013 13:25:40 +0000 (15:25 +0200)]
src: don't SetInternalField() in ObjectWrap dtor

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 ()

11 years agoRevert "crypto: use better memory BIO implementation"
Ben Noordhuis [Wed, 10 Apr 2013 11:50:59 +0000 (13:50 +0200)]
Revert "crypto: use better memory BIO implementation"

This change shouldn't have landed in the stable branch. It's a feature,
not a bug fix.

This reverts commit 58f93ffc4a23aa7240808288acf8cf9f3022abea.
This reverts commit 8c8ebe49b62c47b0ac87d697d0bc3515604667c3.
This reverts commit ba0f7b8066cb4dc8cb3cd4931e52aa94af40a709.
This reverts commit 21f3c5c3670861436392f30da2033cccbc137309.

11 years agonet: fix socket.bytesWritten Buffers support
Fedor Indutny [Wed, 10 Apr 2013 10:21:41 +0000 (14:21 +0400)]
net: fix socket.bytesWritten Buffers support

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.

11 years agocrypto: remove unused ClientHelloParser field
Ben Noordhuis [Tue, 9 Apr 2013 23:38:59 +0000 (01:38 +0200)]
crypto: remove unused ClientHelloParser field

11 years agoblog: Post for v0.8.23
isaacs [Tue, 9 Apr 2013 00:41:49 +0000 (17:41 -0700)]
blog: Post for v0.8.23

11 years agobuffer: fix offset checks
Łukasz Walukiewicz [Fri, 5 Apr 2013 14:48:18 +0000 (16:48 +0200)]
buffer: fix offset checks

Fixed offset checks in Buffer.readInt32LE() and Buffer.readInt32BE()
functions.

11 years agostream: call write cb before finish event
isaacs [Mon, 8 Apr 2013 18:00:26 +0000 (11:00 -0700)]
stream: call write cb before finish event

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

Fixes felixge/node-formidable#209
Fixes #5215

11 years agohttp: Support write(data, 'hex')
isaacs [Mon, 8 Apr 2013 15:46:51 +0000 (08:46 -0700)]
http: Support write(data, 'hex')

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.

11 years agocrypto: dh secret should be left-padded
Fedor Indutny [Mon, 8 Apr 2013 14:00:18 +0000 (18:00 +0400)]
crypto: dh secret should be left-padded

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

11 years agoprocess: expose NODE_MODULE_VERSION in process.versions
Rod Vagg [Mon, 8 Apr 2013 05:13:23 +0000 (15:13 +1000)]
process: expose NODE_MODULE_VERSION in process.versions

11 years agocrypto: move write_head in bio's Reset() method
Fedor Indutny [Sat, 6 Apr 2013 19:26:24 +0000 (23:26 +0400)]
crypto: move write_head in bio's Reset() method

11 years agocrypto: fix changing buffers in bio
Fedor Indutny [Sat, 6 Apr 2013 19:26:00 +0000 (23:26 +0400)]
crypto: fix changing buffers in bio

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.

11 years agocrypto: fix style issues in bio
Fedor Indutny [Sat, 6 Apr 2013 18:47:53 +0000 (22:47 +0400)]
crypto: fix style issues in bio

Stop changing arguments, use local variables for things that change.

11 years agocrypto: fix constructor call in crypto streams
Andreas Madsen [Mon, 8 Apr 2013 08:32:53 +0000 (10:32 +0200)]
crypto: fix constructor call in crypto streams

When using some stream method on a lazy crypto stream, the transform
constructor wasn't called. This caused the internal state object to
be undefined.

11 years agonet: account encoding in .byteLength
Fedor Indutny [Mon, 8 Apr 2013 07:48:46 +0000 (11:48 +0400)]
net: account encoding in .byteLength

11 years agodoc: document linux pwrite() bug
Ben Noordhuis [Sun, 7 Apr 2013 22:41:33 +0000 (00:41 +0200)]
doc: document linux pwrite() bug

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.

11 years agonet: fix buffer iteration in bytesWritten
Fedor Indutny [Sun, 7 Apr 2013 20:50:40 +0000 (00:50 +0400)]
net: fix buffer iteration in bytesWritten

11 years agocrypto: zero is not an error if writing 0 bytes
Fedor Indutny [Sun, 7 Apr 2013 14:03:27 +0000 (18:03 +0400)]
crypto: zero is not an error if writing 0 bytes

fix #5128

11 years agotls: Re-enable check of CN-ID in cert verification
Tobias Müllerleile [Thu, 4 Apr 2013 19:10:53 +0000 (22:10 +0300)]
tls: Re-enable check of CN-ID in cert verification

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.

11 years agostream: unused variable
Rafael Garcia [Thu, 4 Apr 2013 19:54:57 +0000 (12:54 -0700)]
stream: unused variable

11 years agostream: remove vestiges of previous _transform API
Rafael Garcia [Thu, 4 Apr 2013 19:18:21 +0000 (12:18 -0700)]
stream: remove vestiges of previous _transform API

11 years agodoc: Correct caveats for http Readables
isaacs [Fri, 5 Apr 2013 18:07:53 +0000 (11:07 -0700)]
doc: Correct caveats for http Readables