On Windows, when node or io.js attempts to dynamically load a compiled
addon, the compiled addon tries to load node.exe or iojs.exe again -
depending on which import library the module used when it was linked.
This causes many compiled addons to break when node.exe or iojs.exe are
renamed, because when the binary has been renamed the addon DLL can't
find the (right) .exe file to load its imports from.
This patch gives compiled addon developers an option to overcome this
restriction by compiling a delay-load hook into their binary. The
delay-load hook ensures that whenever a module tries to load imports
from node.exe/iojs.exe, it'll just look at the process image, thereby
making the addon work regardless of what name the node/iojs binary has.
To enable this feature, the addon developer must set the
'win_delay_load_hook' option to 'true' in their binding.gyp file, like
this:
```
{
'targets': [
{
'target_name': 'ernie',
'win_delay_load_hook': 'true',
...
```
Bug: https://github.com/iojs/io.js/issues/751
Bug: https://github.com/iojs/io.js/issues/965
Upstream PR: https://github.com/TooTallNate/node-gyp/pull/599
PR-URL: https://github.com/iojs/io.js/pull/1251
Reviewed-By: Rod Vagg <rod@vagg.org>
PR-URL: https://github.com/iojs/io.js/pull/1266
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Partially revert 776b73b243.
Following code crashes after backported timer leak fixes:
```javascript
var timer = setInterval(function() {
clearInterval(timer);
}, 10);
timer.unref();
```
Note that this is actually tested in a `test-timers-unref.js`, and is
crashing only with 776b73b243.
Calling `clearInterval` leads to the crashes in case of `.unref()`ed
timers, and might lead to a extra timer spin in case of regular
intervals that was closed during the interval callback. All of these
happens because `.unref()`ed timer has it's own `_handle` and was used
after the `.close()`.
PR-URL: https://github.com/iojs/io.js/pull/1330
Reviewed-by: Trevor Norris <trev.norris@gmail.com>
Drop the homegrown thread pool that was introduced in commit 50839a0
("v8_platform: provide default v8::Platform impl") and use one from
V8's libplatform library. Performance is comparable and it removes
a few hundred lines of code.
The calls to v8::platform::PumpMessageLoop() are currently no-ops
because V8 does not (yet?) use v8::Platform::CallOnForegroundThread().
Packagers that link against a shared libv8 now also need to make
libv8_platform available.
PR-URL: https://github.com/iojs/io.js/pull/1329
Reviewed-By: Fedor Indutny <fedor@indutny.com>
This commit fixes a few things for this benchmark:
1. Ensures the temporary directory for the unix socket exists.
2. Prevents the client code from being run directly because the
server script is the one that calls out the client code.
3. Ensures the server is closed once the client benchmarks have
finished.
4. Since this is an http benchmark, it should be moved to the http
benchmarks subdirectory.
PR-URL: https://github.com/iojs/io.js/pull/1257
Reviewed-By: Roman Reiss <me@silverwind.io>
Minor convenience for platforms that doesn't have gmake installed
but prefer the habit of writing make instead of gmake.
test needs to live in .PHONY to get passed on to gmake.
PR-URL: https://github.com/iojs/io.js/pull/1298
Reviewed-By: Fedor Indutny <fedor@indutny.com>
Notable changes:
* fs: corruption can be caused by fs.writeFileSync() and append-mode
fs.writeFile() and fs.writeFileSync() under certain circumstances,
reported in #1058, fixed in #1063 (Olov Lassus).
* iojs: an "internal modules" API has been introduced to allow core
code to share JavaScript modules internally only without having to
expose them as a public API, this feature is for core-only #848
(Vladimir Kurchatkin).
* timers: two minor problems with timers have been fixed:
- Timer#close() is now properly idempotent #1288 (Petka Antonov).
- setTimeout() will only run the callback once now after an
unref() during the callback #1231 (Roman Reiss).
* Windows: a "delay-load hook" has been added for compiled add-ons
on Windows that should alleviate some of the problems that Windows
users may be experiencing with add-ons in io.js #1251
(Bert Belder).
* V8: minor bug-fix upgrade for V8 to 4.1.0.27.
* npm: upgrade npm to 2.7.4. See npm CHANGELOG.md for details.
On Windows, when node or io.js attempts to dynamically load a compiled
addon, the compiled addon tries to load node.exe or iojs.exe again -
depending on which import library the module used when it was linked.
This causes many compiled addons to break when node.exe or iojs.exe are
renamed, because when the binary has been renamed the addon DLL can't
find the (right) .exe file to load its imports from.
This patch gives compiled addon developers an option to overcome this
restriction by compiling a delay-load hook into their binary. The
delay-load hook ensures that whenever a module tries to load imports
from node.exe/iojs.exe, it'll just look at the process image, thereby
making the addon work regardless of what name the node/iojs binary has.
To enable this feature, the addon developer must set the
'win_delay_load_hook' option to 'true' in their binding.gyp file, like
this:
```
{
'targets': [
{
'target_name': 'ernie',
'win_delay_load_hook': 'true',
...
```
Bug: https://github.com/iojs/io.js/issues/751
Bug: https://github.com/iojs/io.js/issues/965
Upstream PR: https://github.com/TooTallNate/node-gyp/pull/599
PR-URL: https://github.com/iojs/io.js/pull/1251
Reviewed-By: Rod Vagg <rod@vagg.org>
PR-URL: https://github.com/iojs/io.js/pull/1266
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Fix a race condition in parallel/test-vm-debug-context where the 'exit'
event for the child process is emitted before the first and only 'data'
event for the child process's stderr stream.
I considered deferring the 'exit' event in lib/child_process.js until
all stdio streams have been closed but I realized that's not going to
work when the child process spins off grandchildren that keep the stdio
file descriptors alive.
Fixes: https://github.com/iojs/io.js/issues/1291
PR-URL: https://github.com/iojs/io.js/pull/1294
Reviewed-By: Brendan Ashworth <brendan.ashworth@me.com>
On AIX, OS X and the BSDs, calling shutdown() on one end of a pipe
when the other end has closed the connection fails with ENOTCONN.
The sequential/test-child-process-execsync test failed sporadically
because of a race between the parent and the child where one closed
its end of the pipe before the other got around to calling shutdown()
on its end of the pipe.
Libuv is not the right place to handle that because it can't tell if
the ENOTCONN error is genuine but io.js can.
Refs: https://github.com/libuv/libuv/pull/268
PR-URL: https://github.com/iojs/io.js/pull/1214
Reviewed-By: Bert Belder <bertbelder@gmail.com>
When debug in remote mode with host:port or pid, the interface
spawn child process also. If the debugger agent is running, will
get following output:
```
< Error: listen EADDRINUSE :::5858
< at Object.exports._errnoException (util.js:734:11)
< at exports._exceptionWithHostPort (util.js:757:20)
< at Agent.Server._listen2 (net.js:1155:14)
< at listen (net.js:1181:10)
< at Agent.Server.listen (net.js:1268:5)
< at Object.start (_debug_agent.js:21:9)
< at startup (node.js:68:9)
< at node.js:799:3
```
This fix won't spawn child process and no more error message was
shown.
When use `iojs debug`, the tip information just like this:
```
Usage: iojs debug script.js
```
This fix will display the advance usage also:
```
Usage: iojs debug script.js
iojs debug <host>:<port>
iojs debug -p <pid>
```
Fixes: https://github.com/iojs/io.js/issues/889
PR-URL: https://github.com/iojs/io.js/pull/1282
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
The fs.watch test's write events sometimes aren't produced on OS X,
possibly because of a fsevents race condition. This patch gives delays
the writing a total of 20ms, which makes the test pass consistently.
PR-URL: https://github.com/iojs/io.js/pull/1275
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Fix a regression that was introduced in commit 2db758c ("iojs: introduce
internal modules") where the computed id for "config.gypi" on Windows
was not "config" but an empty string.
With an empty string, the build succeeds but the binary is unusable:
startup.processConfig() in src/node.js chokes on the missing .config
property.
PR-URL: https://github.com/iojs/io.js/pull/1281
Reviewed-By: Vladimir Kurchatkin <vladimir.kurchatkin@gmail.com>
Don't lazy-load setInterval(), setTimeout(), etc. Most applications are
going to need them and routing every call through NativeModule.require()
and Function#apply() is not exactly efficient.
PR-URL: https://github.com/iojs/io.js/pull/1280
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
The delay-load hook that was landed in 3d46fef to make compiled addons
work on Windows regardless of the iojs.exe/node.exe filename causes
issues with a small amount of compiled addons.
Therefore this patch makes it an opt-in feature. An addon may set the
'win_delay_load_hook' option to 'true' in its binding.gyp to enable this
feature.
Example:
```
{
'targets': [
{
'target_name': 'ernie',
'win_delay_load_hook': 'true',
...
```
Refs: https://github.com/iojs/io.js/pull/1251
PR-URL: https://github.com/iojs/io.js/pull/1266
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
Calling this.unref() during the callback of SetTimeout caused the
callback to get executed twice because unref() didn't expect to be
called during that time and did not stop the ref()ed Timeout but
did start a new timer. This commit prevents the new timer creation
when the callback was already called.
Fixes: https://github.com/iojs/io.js/issues/1191
Reviewed-by: Trevor Norris <trev.norris@gmail.com>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
PR-URL: https://github.com/iojs/io.js/pull/1231
On Windows, when node or io.js attempts to dynamically load a compiled
addon, the compiled addon tries to load node.exe or iojs.exe again -
depending on which import library the module used when it was linked.
This makes it impossible to rename node.exe or iojs.exe, because when
that happens the module can't find its dependencies.
With this patch, a delay-load hook is added to all modules that are
compiled with node-gyp. The delay-load hook ensures that whenever a
module tries to load imports from node.exe/iojs.exe, it'll just refer
back to the process image, thus making it possible to rename the
iojs/node binary.
Bug: https://github.com/iojs/io.js/issues/751
Bug: https://github.com/iojs/io.js/issues/965
Upstream PR: https://github.com/TooTallNate/node-gyp/pull/599
PR-URL: https://github.com/iojs/io.js/pull/1251
Reviewed-By: Rod Vagg <rod@vagg.org>
Bert Belder