match chromium's approach for locating the most appropriate macos SDK to use

match chromium's approach for locating the most appropriate macos SDK to use

By doing this, the ninja generator for gyp works great on macos, resulting in faster (I think) and much, much terser builds.

BUG=

Committed: https://code.google.com/p/skia/source/detail?r=7684

[Humper, 2013-01-16 23:28:15]

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py
File tools/find_mac_sdk.py (right):

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py#newcode1
tools/find_mac_sdk.py:1: #!/usr/bin/env python
I yanked this file verbatim from the chromium tree src/build/mac/find_sdk.py

[DerekS, 2013-01-17 13:47:08]

I'm fine with it, but I'll let those more familiar with the inner workings of
python and gyp make the call.

[Humper, 2013-01-29 13:58:26]

On 2013/01/17 13:47:08, DerekS wrote:
> I'm fine with it, but I'll let those more familiar with the inner workings of
> python and gyp make the call.

ping

[epoger, 2013-02-05 18:59:19]

On 2013/01/29 13:58:26, Humper wrote:
> On 2013/01/17 13:47:08, DerekS wrote:
> > I'm fine with it, but I'll let those more familiar with the inner workings
of
> > python and gyp make the call.
> 
> ping

I nominate bungeman to be sole reviewer

[bungeman, 2013-02-05 20:03:46]

Not against the idea after taking a look at and understanding the python code,
but there's quite a bit in here that isn't very clear or needed by Skia.

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py
File tools/find_mac_sdk.py (right):

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py#newcode18
tools/find_mac_sdk.py:18: from optparse import OptionParser
See below, but after removing options we can get rid of this as well.

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py#newcode28
tools/find_mac_sdk.py:28: parser.add_option("--verify",
We aren't using --verify, and using it would probably cause problems with the
skia bots. I think the intention was to ensure that the min_sdk_version passed
was the one actually used, so that there wouldn't be any issues running on the
intended version. I doubt all of the skia bots even have 10.6 sdks on them, and
we already have actual 10.6 bots anyway.

tl;dr: I'd prefer to get rid of this verify thing.

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py#newcode31
tools/find_mac_sdk.py:31: parser.add_option("--sdk_path",
I'm not even sure what this was for, given the name. Maybe the idea was to have
a way to override the whole sdk_dir thing, but currently it just turns off
verification when set to anything. Skia already has the skia_osx_sdkroot which
can be set to point to anything, so I see no need for any of this. If we wanted
to keep it we should implement it as a (if set) replacement for calculating
sdk_dir on lines 37-53.

tl;dr: lets get rid of this as well.

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py#newcode37
tools/find_mac_sdk.py:37: job = subprocess.Popen(['xcode-select',
'-print-path'],
Lines 37-53 need to be their own function, it's rather messy code but does
something very separable.

https://codereview.appspot.com/7142047/diff/1/tools/find_mac_sdk.py#newcode62
tools/find_mac_sdk.py:62: if options.verify and best_sdk != min_sdk_version and
not options.sdk_path:
Removing parameters means removing this block as well. Skia has a 10.6 bot, so
we're already covered as far as backward compatibility.

[bungeman, 2013-02-08 21:41:48]

lgtm