Add msvs version option to gyptest.

This makes it easier to test combinations of msvs version without having to
muck with environment variables
Committed: https://code.google.com/p/gyp/source/detail?r=1467

[iannucci, 2012-08-09 00:34:29]

Pretty straightforward (I hope). I've been using this change to test my other
gyp+msvs changes :). Works independently of -f, so you can do -f ninja -m 2008
types of things too.

[scottmg, 2012-08-09 15:22:13]

On 2012/08/09 00:34:29, iannucci wrote:
> Pretty straightforward (I hope). I've been using this change to test my other
> gyp+msvs changes :). Works independently of -f, so you can do -f ninja -m 2008
> types of things too.

Maybe it'd be better to make this match the other command line gyp format? i.e.
be able to pass generator flags as -Gmsvs_version=2008?

[iannucci, 2012-08-09 18:12:44]

On 2012/08/09 15:22:13, scottmg wrote:
> On 2012/08/09 00:34:29, iannucci wrote:
> > Pretty straightforward (I hope). I've been using this change to test my
other
> > gyp+msvs changes :). Works independently of -f, so you can do -f ninja -m
2008
> > types of things too.
> 
> Maybe it'd be better to make this match the other command line gyp format?
i.e.
> be able to pass generator flags as -Gmsvs_version=2008?

Like a generic -G flag which takes <option>=<val>, and then does
`os.environ["GYP_%s" % option.toupper()] = val`? Ok, sounds good, I'll do that.

[scottmg, 2012-08-09 18:22:05]

On 2012/08/09 18:12:44, iannucci wrote:
> On 2012/08/09 15:22:13, scottmg wrote:
> > On 2012/08/09 00:34:29, iannucci wrote:
> > > Pretty straightforward (I hope). I've been using this change to test my
> other
> > > gyp+msvs changes :). Works independently of -f, so you can do -f ninja -m
> 2008
> > > types of things too.
> > 
> > Maybe it'd be better to make this match the other command line gyp format?
> i.e.
> > be able to pass generator flags as -Gmsvs_version=2008?
> 
> Like a generic -G flag which takes <option>=<val>, and then does
> `os.environ["GYP_%s" % option.toupper()] = val`? Ok, sounds good, I'll do
that.

I'm actually not sure how it works. :) I was just looking at
pylib/gyp/__init__.py for where msvs_version is set. All I was thinking is that
it'd be best to keep the gyp-the-tool command line options close to gyptest.py's
command line options, if possible.

[iannucci, 2012-08-10 01:50:40]

Ok, modified to pass -G options through to gyp, as well as use msvs_version in
gyptest, if it's present. This is a little more complicated, but it allows
arbitrary flags to pass to gyp, and it re-uses the arg names that gyp uses.

I think the red trybot was flake.

[scottmg, 2012-08-10 16:11:46]

lgtm

https://chromiumcodereview.appspot.com/10830232/diff/8002/gyptest.py
File gyptest.py (right):

https://chromiumcodereview.appspot.com/10830232/diff/8002/gyptest.py#newcode232
gyptest.py:232: status = cr.run([sys.executable, test]+gyp_options,
spacing

https://chromiumcodereview.appspot.com/10830232/diff/8002/test/lib/TestGyp.py
File test/lib/TestGyp.py (right):

https://chromiumcodereview.appspot.com/10830232/diff/8002/test/lib/TestGyp.py...
test/lib/TestGyp.py:625: msvs_version = os.environ.get('GYP_MSVS_VERSION',
msvs_version)
Hmm, sorry. I sort of assumed we passed generator flags through here already. I
guess it makes sense not to so that you're isolated from the environment when
running tests though.

[iannucci, 2012-08-10 18:30:32]

On 2012/08/10 16:11:46, scottmg wrote:
> lgtm
> 
> https://chromiumcodereview.appspot.com/10830232/diff/8002/gyptest.py
> File gyptest.py (right):
> 
>
https://chromiumcodereview.appspot.com/10830232/diff/8002/gyptest.py#newcode232
> gyptest.py:232: status = cr.run([sys.executable, test]+gyp_options,
> spacing
> 
> https://chromiumcodereview.appspot.com/10830232/diff/8002/test/lib/TestGyp.py
> File test/lib/TestGyp.py (right):
> 
>
https://chromiumcodereview.appspot.com/10830232/diff/8002/test/lib/TestGyp.py...
> test/lib/TestGyp.py:625: msvs_version = os.environ.get('GYP_MSVS_VERSION',
> msvs_version)
> Hmm, sorry. I sort of assumed we passed generator flags through here already.
I
> guess it makes sense not to so that you're isolated from the environment when
> running tests though.

Fixed nit. Yeah it's a little messy, I guess, but since it's test scaffolding,
it seems better to have the scaffolding be a little more complex to keep the
tested thing simple/consistent (since we want to exercise the -G flag on gyp). 

Might it be worth passing the format parameter in this fashion as well?

[scottmg, 2012-08-10 19:09:52]

On 2012/08/10 18:30:32, iannucci wrote:
> On 2012/08/10 16:11:46, scottmg wrote:
> > lgtm
> > 
> > https://chromiumcodereview.appspot.com/10830232/diff/8002/gyptest.py
> > File gyptest.py (right):
> > 
> >
>
https://chromiumcodereview.appspot.com/10830232/diff/8002/gyptest.py#newcode232
> > gyptest.py:232: status = cr.run([sys.executable, test]+gyp_options,
> > spacing
> > 
> >
https://chromiumcodereview.appspot.com/10830232/diff/8002/test/lib/TestGyp.py
> > File test/lib/TestGyp.py (right):
> > 
> >
>
https://chromiumcodereview.appspot.com/10830232/diff/8002/test/lib/TestGyp.py...
> > test/lib/TestGyp.py:625: msvs_version = os.environ.get('GYP_MSVS_VERSION',
> > msvs_version)
> > Hmm, sorry. I sort of assumed we passed generator flags through here
already.
> I
> > guess it makes sense not to so that you're isolated from the environment
when
> > running tests though.
> 
> Fixed nit. Yeah it's a little messy, I guess, but since it's test scaffolding,
> it seems better to have the scaffolding be a little more complex to keep the
> tested thing simple/consistent (since we want to exercise the -G flag on gyp).


Yeah, seems fine.

> 
> Might it be worth passing the format parameter in this fashion as well?

Shrug, not too worried about it. Either way.