Adds an IPython kernel running in NaCl, and a wrapper as a Chrome extension.

Adds an IPython kernel running in NaCl, and a wrapper as a Chrome extension.

This requires CL 138913004 to work, and some further minor modifications to 138913004 not included here.

BUG=

[Kester Tong, 2014-02-14 16:00:21]

Hey,

This CL adds an IPython kernel.  While the IPython kernel uses zmq for
communicating with the front end, this port uses pepper messaging.

Because this CL depends on the port of IPython and other python modules, it is
incomplete, but it gives a general idea of the structure of the new kernel.

Kester

[Kester Tong, 2014-02-18 21:11:39]

Hi Sam,

I've added you as a reviewer on this CL.  The CL isn't in its
final form since it also requires changes that can only be
made once Matthew Turk's CL is checked in.  But the main
code is here.  It would be great if you can take a look at
some point.

Kester

[Sam Clegg, 2014-02-18 23:39:31]

lgtm.

This looks very cool.

Would you mind running 'git cl upload' again as the first two files in the CL
didn't seem to upload correctly.

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/back...
File ports/ipython_extension/background.js (right):

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/back...
ports/ipython_extension/background.js:1: 
Please add copyright to all files in naclports.

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/buil...
File ports/ipython_extension/build.sh (right):

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/buil...
ports/ipython_extension/build.sh:2: # Copyright (c) 2013 The Native Client
Authors. All rights reserved.
2014

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/buil...
ports/ipython_extension/build.sh:41: DefaultBuildStep
You should be able to just call BuildStep_SDKBuildSystem
here.  It was recently added.

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/mani...
File ports/ipython_extension/manifest.json (right):

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/mani...
ports/ipython_extension/manifest.json:10: "https://colab.corp.google.com/*",
Do you mind making this URL public?

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/pyth...
File ports/ipython_extension/python.cc (right):

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/pyth...
ports/ipython_extension/python.cc:1: #include <python2.7/Python.h>
Copyright

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/simp...
File ports/ipython_extension/simple_shell_main.py (right):

https://codereview.chromium.org/165473002/diff/1/ports/ipython_extension/simp...
ports/ipython_extension/simple_shell_main.py:1: """A simple shell that uses the
IPython messaging system."""
Copyright

[Kester Tong, 2014-05-02 18:52:22]

Hi, I've fixed the copyright notices, and upgraded to the new build system.

One difference is that I no longer include any files for a Chrome App, as this
Native Client code is designed to be part of a larger web/Chrome App.

I'm not sure if I have the dependencies in pkg_info 100% correct, and I'm also
unsure about whether they are guaranteed to build in order, which is necessary
since python-static needs to be called after numpy etc.

[Sam Clegg, 2014-05-02 18:59:27]

On 2014/05/02 18:52:22, Kester Tong wrote:
> Hi, I've fixed the copyright notices, and upgraded to the new build system.
> 
> One difference is that I no longer include any files for a Chrome App, as this
> Native Client code is designed to be part of a larger web/Chrome App.
> 
> I'm not sure if I have the dependencies in pkg_info 100% correct, and I'm also
> unsure about whether they are guaranteed to build in order, which is necessary
> since python-static needs to be called after numpy etc.

Right now dependencies are built in order but I was planning on parallelizing
this.

Perhaps we could add numpy as a dependency of python-static instead?

Anyway lgtm.

[Sam Clegg, 2014-05-02 18:59:33]

https://codereview.chromium.org/165473002/diff/290001/ports/ipython-ppapi/REA...
File ports/ipython-ppapi/README.nacl (right):

https://codereview.chromium.org/165473002/diff/290001/ports/ipython-ppapi/REA...
ports/ipython-ppapi/README.nacl:2: ===============================
typo (Python)

[Matthew Turk, 2014-05-02 19:01:18]

On 2014/05/02 18:52:22, Kester Tong wrote:
> Hi, I've fixed the copyright notices, and upgraded to the new build system.
> 
> One difference is that I no longer include any files for a Chrome App, as this
> Native Client code is designed to be part of a larger web/Chrome App.
> 
> I'm not sure if I have the dependencies in pkg_info 100% correct, and I'm also
> unsure about whether they are guaranteed to build in order, which is necessary
> since python-static needs to be called after numpy etc.

Hi Kester, since python-static will always have pyppapi, do you think there's
any possibility in the future of using that for message passing?

[Kester Tong, 2014-05-02 20:17:00]

On 2014/05/02 18:59:27, Sam Clegg wrote:
> On 2014/05/02 18:52:22, Kester Tong wrote:
> > Hi, I've fixed the copyright notices, and upgraded to the new build system.
> > 
> > One difference is that I no longer include any files for a Chrome App, as
this
> > Native Client code is designed to be part of a larger web/Chrome App.
> > 
> > I'm not sure if I have the dependencies in pkg_info 100% correct, and I'm
also
> > unsure about whether they are guaranteed to build in order, which is
necessary
> > since python-static needs to be called after numpy etc.
> 
> Right now dependencies are built in order but I was planning on parallelizing
> this.
> 
> Perhaps we could add numpy as a dependency of python-static instead?
> 
> Anyway lgtm.


I'm happy for python-static to contain all libraries, (and the user could always
manually
remove these dependencies if they wanted).  How does that sound, Matt?

[Kester Tong, 2014-05-02 20:17:27]

On 2014/05/02 18:59:33, Sam Clegg wrote:
>
https://codereview.chromium.org/165473002/diff/290001/ports/ipython-ppapi/REA...
> File ports/ipython-ppapi/README.nacl (right):
> 
>
https://codereview.chromium.org/165473002/diff/290001/ports/ipython-ppapi/REA...
> ports/ipython-ppapi/README.nacl:2: ===============================
> typo (Python)

thanks, will fix in next patch

[Kester Tong, 2014-05-02 20:18:49]

On 2014/05/02 19:01:18, Matthew Turk wrote:
> On 2014/05/02 18:52:22, Kester Tong wrote:
> > Hi, I've fixed the copyright notices, and upgraded to the new build system.
> > 
> > One difference is that I no longer include any files for a Chrome App, as
this
> > Native Client code is designed to be part of a larger web/Chrome App.
> > 
> > I'm not sure if I have the dependencies in pkg_info 100% correct, and I'm
also
> > unsure about whether they are guaranteed to build in order, which is
necessary
> > since python-static needs to be called after numpy etc.
> 
> Hi Kester, since python-static will always have pyppapi, do you think there's
> any possibility in the future of using that for message passing?

This is a good idea.  Right now pyppapi isn't working inside the IPython kernel.
My error message is

Replacing gethostname()
Either ppb_messaging or pp_instance is NULL.
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-4-565374b0d5f1> in <module>()
----> 1 import pyppapi

/lib/python2.7/site-packages/pyppapi/__init__.py in <module>()
     29 
     30 from .ppapi_simple import NaClInstance
---> 31 nacl_instance = NaClInstance()
     32 nacl_instance.pylab()
     33 

/lib/python2.7/site-packages/pyppapi/ppapi_simple.a in
pyppapi.ppapi_simple.NaClInstance.__init__ (pyppapi/ppapi_simple.cpp:1680)()

RuntimeError: 

But I will keep looking into it.

[Matthew Turk, 2014-05-03 18:49:04]

On 2014/05/02 20:18:49, Kester Tong wrote:
> On 2014/05/02 19:01:18, Matthew Turk wrote:
> > On 2014/05/02 18:52:22, Kester Tong wrote:
> > > Hi, I've fixed the copyright notices, and upgraded to the new build
system.
> > > 
> > > One difference is that I no longer include any files for a Chrome App, as
> this
> > > Native Client code is designed to be part of a larger web/Chrome App.
> > > 
> > > I'm not sure if I have the dependencies in pkg_info 100% correct, and I'm
> also
> > > unsure about whether they are guaranteed to build in order, which is
> necessary
> > > since python-static needs to be called after numpy etc.
> > 
> > Hi Kester, since python-static will always have pyppapi, do you think
there's
> > any possibility in the future of using that for message passing?
> 
> This is a good idea.  Right now pyppapi isn't working inside the IPython
kernel.
> My error message is
> 
> Replacing gethostname()
> Either ppb_messaging or pp_instance is NULL.
> ---------------------------------------------------------------------------
> RuntimeError                              Traceback (most recent call last)
> <ipython-input-4-565374b0d5f1> in <module>()
> ----> 1 import pyppapi
> 
> /lib/python2.7/site-packages/pyppapi/__init__.py in <module>()
>      29 
>      30 from .ppapi_simple import NaClInstance
> ---> 31 nacl_instance = NaClInstance()
>      32 nacl_instance.pylab()
>      33 
> 
> /lib/python2.7/site-packages/pyppapi/ppapi_simple.a in
> pyppapi.ppapi_simple.NaClInstance.__init__ (pyppapi/ppapi_simple.cpp:1680)()
> 
> RuntimeError: 
> 
> But I will keep looking into it.

Ah, I know what the issue is.  It's because NaClInstance assumes it can access
an extern'd pointer to the PPAPI instance.  This isn't crucial, so don't worry
about looking into it.

[Kester Tong, 2014-05-05 20:21:58]

On 2014/05/03 18:49:04, Matthew Turk wrote:
> On 2014/05/02 20:18:49, Kester Tong wrote:
> > On 2014/05/02 19:01:18, Matthew Turk wrote:
> > > On 2014/05/02 18:52:22, Kester Tong wrote:
> > > > Hi, I've fixed the copyright notices, and upgraded to the new build
> system.
> > > > 
> > > > One difference is that I no longer include any files for a Chrome App,
as
> > this
> > > > Native Client code is designed to be part of a larger web/Chrome App.
> > > > 
> > > > I'm not sure if I have the dependencies in pkg_info 100% correct, and
I'm
> > also
> > > > unsure about whether they are guaranteed to build in order, which is
> > necessary
> > > > since python-static needs to be called after numpy etc.
> > > 
> > > Hi Kester, since python-static will always have pyppapi, do you think
> there's
> > > any possibility in the future of using that for message passing?
> > 
> > This is a good idea.  Right now pyppapi isn't working inside the IPython
> kernel.
> > My error message is
> > 
> > Replacing gethostname()
> > Either ppb_messaging or pp_instance is NULL.
> > ---------------------------------------------------------------------------
> > RuntimeError                              Traceback (most recent call last)
> > <ipython-input-4-565374b0d5f1> in <module>()
> > ----> 1 import pyppapi
> > 
> > /lib/python2.7/site-packages/pyppapi/__init__.py in <module>()
> >      29 
> >      30 from .ppapi_simple import NaClInstance
> > ---> 31 nacl_instance = NaClInstance()
> >      32 nacl_instance.pylab()
> >      33 
> > 
> > /lib/python2.7/site-packages/pyppapi/ppapi_simple.a in
> > pyppapi.ppapi_simple.NaClInstance.__init__ (pyppapi/ppapi_simple.cpp:1680)()
> > 
> > RuntimeError: 
> > 
> > But I will keep looking into it.
> 
> Ah, I know what the issue is.  It's because NaClInstance assumes it can access
> an extern'd pointer to the PPAPI instance.  This isn't crucial, so don't worry
> about looking into it.

Ok, sounds good.  What do do you think about including all python libs as
dependencies of python-static?  This way we guarantee they are build properly,
and don't have to worry about python-static getting built too early.

[Matthew Turk, 2014-05-13 19:08:44]

lgtm

[Kester Tong, 2014-05-13 22:23:24]

The CQ bit was checked by kestert@google.com

[Kester Tong, 2014-05-13 22:23:35]

The CQ bit was unchecked by kestert@google.com

[Kester Tong, 2014-05-13 22:25:53]

The CQ bit was checked by kestert@google.com

[Kester Tong, 2014-05-13 22:26:03]

The CQ bit was unchecked by kestert@google.com

[Kester Tong, 2014-05-13 23:43:50]

The CQ bit was checked by kestert@google.com

[Kester Tong, 2014-05-13 23:45:03]

On 2014/05/13 22:26:03, Kester Tong wrote:
> The CQ bit was unchecked by mailto:kestert@google.com

I'm having trouble committing this patch.  I ran git cl dcommit, but I got a
"tree is closed" error.

Then I realized that maybe all I had to do was check the "commit" checkbox in
the patch, and it would be added to the commit queue.  Is that right?

[commit-bot: I haz the power, 2014-05-13 23:45:07]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2014-05-13 23:45:07]

Commit queue rejected this change because it did not recognize the base URL.
Please commit your change manually.

[Sam Clegg, 2014-05-13 23:50:46]

On 2014/05/13 23:45:03, Kester Tong wrote:
> On 2014/05/13 22:26:03, Kester Tong wrote:
> > The CQ bit was unchecked by mailto:kestert@google.com
> 
> I'm having trouble committing this patch.  I ran git cl dcommit, but I got a
> "tree is closed" error.
> 
> Then I realized that maybe all I had to do was check the "commit" checkbox in
> the patch, and it would be added to the commit queue.  Is that right?

Sorry, the commit queues doesn't work for this repo.  You need to commit using
dcommit.

I'll open the tree.

[Matthew Turk, 2014-05-15 10:28:18]

On 2014/05/13 23:50:46, Sam Clegg wrote:
> On 2014/05/13 23:45:03, Kester Tong wrote:
> > On 2014/05/13 22:26:03, Kester Tong wrote:
> > > The CQ bit was unchecked by mailto:kestert@google.com
> > 
> > I'm having trouble committing this patch.  I ran git cl dcommit, but I got a
> > "tree is closed" error.
> > 
> > Then I realized that maybe all I had to do was check the "commit" checkbox
in
> > the patch, and it would be added to the commit queue.  Is that right?
> 
> Sorry, the commit queues doesn't work for this repo.  You need to commit using
> dcommit.
> 
> I'll open the tree.

Hi Sam,

Kester asked me to try committing, so I did so, although codereview wouldn't
authorize the closure of the issue.  This was committed as svn revision 1272.

[sbc, 2016-09-07 22:16:58]

sbc@google.com changed reviewers:
- sbc@google.com