Build pyauto on win

- Update README.chromium to mention change to swig libs in third_party
- Provide platform-specific macros to swig.
- Update DEPS to use right swig updates

To be committed _after_ accompanying CL: http://codereview.chromium.org/661334

BUG=32285
TEST=None


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=41048

[John Grabowski, 2010-03-02 06:00:06]

http://codereview.chromium.org/661342/diff/32/1018
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/661342/diff/32/1018#newcode1866
chrome/chrome_tests.gypi:1866: ['(OS!="linux" or target_arch!="x64" or
linux_fpic==1) and '
Ugh.  Make more obvious.  E.g.
OS=="mac" or OS=="win" or (linux_fpic==1 and target_arch!="arm"):

http://codereview.chromium.org/661342/diff/32/1018#newcode1937
chrome/chrome_tests.gypi:1937: ['OS=="win"', {
Mixing py2.4 and py2.5 may be problematic.  Can't we settle on py2.5?

http://codereview.chromium.org/661342/diff/32/1013
File tools/swig/swig.py (right):

http://codereview.chromium.org/661342/diff/32/1013#newcode36
tools/swig/swig.py:36: 'cygwin':  '-DSWIGWIN',
The presence of cygwin implies it is supported.
Is it?  Do we need to support both win py and cygwin py?
Life is simpler if only one.  win py is probably better since it allows access
to win stuff like the registry.

[Nirnimesh, 2010-03-02 23:50:22]

The try bots have an issue with patching this CL. I won't commit until I get
them green.

http://codereview.chromium.org/661342/diff/32/1018
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/661342/diff/32/1018#newcode1866
chrome/chrome_tests.gypi:1866: ['(OS!="linux" or target_arch!="x64" or
linux_fpic==1) and '
On 2010/03/02 06:00:06, John Grabowski wrote:
> Ugh.  Make more obvious.  E.g.
> OS=="mac" or OS=="win" or (linux_fpic==1 and target_arch!="arm"):
> 

Done.
Also updated this line to fix a conflict with trunk.

http://codereview.chromium.org/661342/diff/32/1018#newcode1937
chrome/chrome_tests.gypi:1937: ['OS=="win"', {
On 2010/03/02 06:00:06, John Grabowski wrote:
> Mixing py2.4 and py2.5 may be problematic.  Can't we settle on py2.5?
> 

The py2.4 and py2.5 story runs till py2.6 too (chromeos uses 2.6) :)
I'm pretty sure whenever a new platform will come up (say, arm), it'll use yet
another version.

mac and linux use system py2.5 but win uses py2.4 (actually, py2.4 binaries are
checked inside the src tree for windows).

http://codereview.chromium.org/661342/diff/32/1013
File tools/swig/swig.py (right):

http://codereview.chromium.org/661342/diff/32/1013#newcode36
tools/swig/swig.py:36: 'cygwin':  '-DSWIGWIN',
On 2010/03/02 06:00:06, John Grabowski wrote:
> The presence of cygwin implies it is supported.
> Is it?  Do we need to support both win py and cygwin py?
> Life is simpler if only one.  win py is probably better since it allows access
> to win stuff like the registry.

I agree. Don't know if anyone builds on cygwin.
Removing.

[John Grabowski, 2010-03-03 00:14:07]

LGTM if you promise to deal with the Windows py2.4 --> py2.5 thing.

http://codereview.chromium.org/661342/diff/32/1018
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/661342/diff/32/1018#newcode1937
chrome/chrome_tests.gypi:1937: ['OS=="win"', {
On 2010/03/02 23:50:22, Nirnimesh wrote:
> On 2010/03/02 06:00:06, John Grabowski wrote:
> > Mixing py2.4 and py2.5 may be problematic.  Can't we settle on py2.5?
> > 
> 
> The py2.4 and py2.5 story runs till py2.6 too (chromeos uses 2.6) :)
> I'm pretty sure whenever a new platform will come up (say, arm), it'll use yet
> another version.
> 
> mac and linux use system py2.5 but win uses py2.4 (actually, py2.4 binaries
are
> checked inside the src tree for windows).

This sucks and will mean Windows will break here and there (c.f. exception
syntax, subprocess, and a few other common things.)

http://codereview.chromium.org/661342/diff/1029/41#newcode1867
chrome/chrome_tests.gypi:1867: 'target_arch!="arm"', {
I think you can place this on two lines.  Python will concatenate like C.  E.g.

'OS=="mac" or OS=="win" or chromeos==1 or '
'(OS....    )'

http://codereview.chromium.org/661342/diff/1029/41#newcode1937
chrome/chrome_tests.gypi:1937: ['OS=="win"', {
In a follow-up CL can you land new binaries for py2.5 then use them?  3
Different pythons sucks.  2.5 is modern enough I think.

[Nirnimesh, 2010-03-08 22:40:46]

John, I didn't check this in following breakages with the try-bot and me
fighting with my windows machine, and essentially wasting a good part of last
week figuring it out.

Replacing os.execv with os.system in this update since that seems to be the norm
in other scripts as well.

http://codereview.chromium.org/661342/diff/1029/41
File chrome/chrome_tests.gypi (right):

http://codereview.chromium.org/661342/diff/1029/41#newcode1867
chrome/chrome_tests.gypi:1867: ['OS=="mac" or OS=="win" or chromeos==1 or
(OS=="linux" and target_arch==host_arch and (target_arch!="x64" or
linux_fpic==1))', {
On 2010/03/03 00:14:07, John Grabowski wrote:
> I think you can place this on two lines.  Python will concatenate like C. 
E.g.
> 
> 'OS=="mac" or OS=="win" or chromeos==1 or '
> '(OS....    )'
> 
> 

Done.

http://codereview.chromium.org/661342/diff/1029/41#newcode1937
chrome/chrome_tests.gypi:1937: ['OS=="win"', {
On 2010/03/03 00:14:07, John Grabowski wrote:
> In a follow-up CL can you land new binaries for py2.5 then use them?  3
> Different pythons sucks.  2.5 is modern enough I think.
> 

Filed http://code.google.com/p/chromium/issues/detail?id=37258 since it involves
sign off from several parties using python. I'll follow up.

[John Grabowski, 2010-03-09 00:27:39]

http://codereview.chromium.org/661342/diff/5001/5002
File tools/swig/swig.py (right):

http://codereview.chromium.org/661342/diff/5001/5002#newcode40
tools/swig/swig.py:40: os.system(cmd)
subprocess module replaces os.system

[Nirnimesh, 2010-03-09 07:56:33]

Thanks for reviewing even while you're away, John.

http://codereview.chromium.org/661342/diff/5001/5002
File tools/swig/swig.py (right):

http://codereview.chromium.org/661342/diff/5001/5002#newcode40
tools/swig/swig.py:40: os.system(cmd)
On 2010/03/09 00:27:39, John Grabowski wrote:
> subprocess module replaces os.system

Done.