Native windows stuff
Bert Belder
The good news is that we don't have to 'big bang' switch from cygwin
to using windows apis as cygwin makes it possible to mix posix and
windows api calls. It for example provides a function that converts an
FD to a windows handle and vice versa.
The first thing I'd like to try to convert is child_process.spawn().
Its performance is pretty bad currently, presumably because cygwin
doesn't do vfork() - it just does fork()s instead. Windows has the
CreateProcess function that makes it possible to connect the spawned
process' stdin/out/err to existing pipes, so we wouldn't need fork at
all.
The question is: where do I put this platform-specific stuff?
Off course I can just use '#ifdef __CYGWIN__' which would probably
work for spawn but may not be very maintainable in the long run. For
example, if we want to port asynchronous io, it's gonna be a major
problem because porting libeio is a no-go I think, so we would end up
writing #ifdefs all the time.
- Bert
r...@tinyclouds.org
> It's no secret that I'd like to have a native windows version of node.
>
> The good news is that we don't have to 'big bang' switch from cygwin
> to using windows apis as cygwin makes it possible to mix posix and
> windows api calls. It for example provides a function that converts an
> FD to a windows handle and vice versa.
Excellent! Yes, let's work slowly toward a MinGW build. There are
probably a dozen big things that need to be reimplemented.
> The first thing I'd like to try to convert is child_process.spawn().
> Its performance is pretty bad currently, presumably because cygwin
> doesn't do vfork() - it just does fork()s instead. Windows has the
> CreateProcess function that makes it possible to connect the spawned
> process' stdin/out/err to existing pipes, so we wouldn't need fork at
> all.
Good idea.
> The question is: where do I put this platform-specific stuff?
> Off course I can just use '#ifdef __CYGWIN__' which would probably
> work for spawn but may not be very maintainable in the long run. For
> example, if we want to port asynchronous io, it's gonna be a major
> problem because porting libeio is a no-go I think, so we would end up
> writing #ifdefs all the time.
Ideally we could work these into the platform files? But perhaps
windows just needs its own files. Just do a
node_child_process_windows.cc file for now.
Bert Belder
windows toolchains as well; they're talking about MinGW mostly.
I've told them about the patch Raff has made for node so maybe they
want to support cyg as well.
More info here: http://code.google.com/p/v8/issues/detail?id=64
- Bert
On Aug 13, 5:04 pm, r...@tinyclouds.org wrote:
Olly
patch? Would it not be better to just port Node.JS to MinGW which is
already supported by V8?
Surely MinGW is faster then Cygwin too?
Elijah Insua
--
You received this message because you are subscribed to the Google Groups "nodejs" group.
To post to this group, send email to nod...@googlegroups.com.
To unsubscribe from this group, send email to nodejs+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/nodejs?hl=en.
Bert Belder
> If V8 supports MinGW why are you trying to force through a Cygwin
> patch?
without much effort, but as long as the v8 team itself is not aware of
that the cygwin build will break every now and then, like the latest
version (v2.3.7).
> Would it not be better to just port Node.JS to MinGW which is
> already supported by V8?
But we need to move a few mountains before node can run on MingW.
Bert Belder
> patch?
without much effort, but as long as the v8 team itself is not aware of
that the cygwin build will break every now and then, like the latest
version (v2.3.7).
> already supported by V8?
Olly
things that needs to be done in order to have a MinGW version of node,
and people who are competent can work through different items if they
can?
Olly
compile on MSVC? I believe V8 is natively designed to compile on
windows MSVC. Using native windows constructs would be faster then
MinGW i would of thought?
Alan Gutierrez
> Why doesn't someone who knows what they're doing, write down a list of
> things that needs to be done in order to have a MinGW version of node,
> and people who are competent can work through different items if they
> can?
Because everyone who knows what they are doing is busy doing something
else. Therefore, someone who has no idea what they are doing, such as
myself, is probably going to have to create a Wiki page:
http://wiki.github.com/ry/node/windows-port-mingw
The road map was extracted from the Windows port discussions from the
mailing list found at the top of the page.
The bottom of the page has a list of suspected Windows sympathizers
based on mailing list discussions, created with the intent of fostering
a Windows hacker community. If you are mortified to find your name on
this list, please remove it. If you are crushed that your name is not on
the list, please add it.
--
Alan Gutierrez - al...@blogometer.com - http://twitter.com/bigeasy
Olly
Alan Gutierrez
> On Aug 15, 5:29 pm, Alan Gutierrez <a...@blogometer.com> wrote:
>> Olly wrote:
>>> Why doesn't someone who knows what they're doing, write down a list of
>>> things that needs to be done in order to have a MinGW version of node,
>>> and people who are competent can work through different items if they
>>> can?
>> Because everyone who knows what they are doing is busy doing something
>> else. Therefore, someone who has no idea what they are doing, such as
>> myself, is probably going to have to create a Wiki page:
>>
>> http://wiki.github.com/ry/node/windows-port-mingw
>>
>> The road map was extracted from the Windows port discussions from the
>> mailing list found at the top of the page.
>>
>> The bottom of the page has a list of suspected Windows sympathizers
>> based on mailing list discussions, created with the intent of fostering
>> a Windows hacker community. If you are mortified to find your name on
>> this list, please remove it. If you are crushed that your name is not on
>> the list, please add it.
> The trouble is, those who know what they're doing, don't use windows.
True dat, but, hey. Looks like Ryan Dahl and Elijah Insua have spoken
with Marc Lehmann, and I'm beginning to understand that...
Node.js depends on libev and libeio for Great Justice. Porting those two
libraries to Windows would port Node.js to Windows. libev already works
on Windows, with issues related to a lack of asynchronous I/O on pipes.
libeio does not work on Windows except via cygwin.
There is historic discussion about schmorp's libraries and Windows to be
found on his website.
http://www.google.com/search?q=site%3Alists.schmorp.de+windows
Does it help to read through the Cygwin code? If Node.js works on
Cygwin, is it a matter of pulling the Cygwin tricks out of Cygwin and
into eio.c? Here's Cygwin's pipe() implementation, with notes on how to
create a selectable pipe.
Or is that too optimistic on my part?