3 additions to common.gypi to control FPO and general cflags for Windows tool builds

3 additions to common.gypi:
1) re-instate r113657 from crbug/106711:
   win_release_OmitFramePointers=0 must explicitly pass /Oy-
   since /O2 implies /Oy
2) add win_debug_OmitFramePointers as tri-state where blank
   value (the default) avoids perturbing VS defaults
3) add win_release_extra_cflags and win_debug_extra_cflags
   to make it easier to add custom local flags for
   tool builds

BUG=106711
TEST=Tweaked these vars via GYP_DEFINES and include.gypi and confirmed the changes in build/All.vcproj

For example:

I have "'win_release_InlineFunctionExpansion': 0" in my include.gypi:

% GYP_DEFINES='win_debug_extra_cflags="/QIfist /nologo"' gclient runhooks
% grep -E -o 'Name="[^"|]*\|[^"]*"|Omit[^ ]*|include" AdditionalOptions="[^"]*"' build/All.vcproj 
Name="Release|Win32"
include" AdditionalOptions="/MP /Ob0"
OmitFramePointers="true"
Name="Debug|Win32"
include" AdditionalOptions="/MP /QIfist /nologo"
Name="Release|x64"
include" AdditionalOptions="/MP /Ob0"
OmitFramePointers="true"
Name="Debug|x64"
include" AdditionalOptions="/MP /QIfist /nologo"

% GYP_DEFINES='win_release_extra_cflags="/QIfist /nologo" win_release_OmitFramePointers=0 win_debug_OmitFramePointers=1' gclient runhooks
% grep -E -o 'Name="[^"|]*\|[^"]*"|Omit[^ ]*|include" AdditionalOptions="[^"]*"' build/All.vcproj
Name="Release|Win32"
include" AdditionalOptions="/MP /QIfist /nologo /Ob0 /Oy-"
OmitFramePointers="false"
Name="Debug|Win32"
include" AdditionalOptions="/MP"
OmitFramePointers="true"
Name="Release|x64"
include" AdditionalOptions="/MP /QIfist /nologo /Ob0 /Oy-"
OmitFramePointers="false"
Name="Debug|x64"
include" AdditionalOptions="/MP"
OmitFramePointers="true"


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=114893

[Derek Bruening, 2011-12-16 21:25:20]

We're waiting on this to make our bots green...

[brettw, 2011-12-16 21:32:42]

http://codereview.chromium.org/8966025/diff/5/build/common.gypi
File build/common.gypi (right):

http://codereview.chromium.org/8966025/diff/5/build/common.gypi#newcode1455
build/common.gypi:1455: 'OmitFramePointers': 'false',
Just curious: would it be better to change OmitFramePonters = false to mean
"/Oy-" ?

[M-A Ruel, 2011-12-16 21:33:01]

http://codereview.chromium.org/8966025/diff/5/build/common.gypi
File build/common.gypi (right):

http://codereview.chromium.org/8966025/diff/5/build/common.gypi#newcode1450
build/common.gypi:1450: # if win_debug_OmitFramePointers is blank, leave as
default
There are cases where we enable optimizations in debug configurations?

[Derek Bruening, 2011-12-16 22:01:14]

http://codereview.chromium.org/8966025/diff/5/build/common.gypi
File build/common.gypi (right):

http://codereview.chromium.org/8966025/diff/5/build/common.gypi#newcode1450
build/common.gypi:1450: # if win_debug_OmitFramePointers is blank, leave as
default
On 2011/12/16 21:33:01, Marc-Antoine Ruel wrote:
> There are cases where we enable optimizations in debug configurations?

We are now building Debug chrome with /O1 when we make a build targeted to run
under Dr. Memory (for improved performance) which does perform some
optimizations

http://codereview.chromium.org/8966025/diff/5/build/common.gypi#newcode1455
build/common.gypi:1455: 'OmitFramePointers': 'false',
On 2011/12/16 21:32:42, brettw wrote:
> Just curious: would it be better to change OmitFramePonters = false to mean
> "/Oy-" ?

Yes, if we could, but we don't control that: Microsoft does.  It is rather
strange that OmitFramePointers=false still performs FPO with /O2 or /Ox but
that's the way it is (for VS2008 at least).

[brettw, 2011-12-16 22:05:21]

I see, LGTM if maruel is also OK.

[M-A Ruel, 2011-12-16 22:44:32]

lgtm

http://codereview.chromium.org/8966025/diff/5/build/common.gypi
File build/common.gypi (right):

http://codereview.chromium.org/8966025/diff/5/build/common.gypi#newcode1450
build/common.gypi:1450: # if win_debug_OmitFramePointers is blank, leave as
default
On 2011/12/16 22:01:14, Derek Bruening wrote:
> On 2011/12/16 21:33:01, Marc-Antoine Ruel wrote:
> > There are cases where we enable optimizations in debug configurations?
> 
> We are now building Debug chrome with /O1 when we make a build targeted to run
> under Dr. Memory (for improved performance) which does perform some
> optimizations

This is a bit sad, I would have preferred to assume debug is debug. Oh well.

[commit-bot: I haz the power, 2011-12-16 22:47:15]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/bruening@chromium.org/8966025/5

[Derek Bruening, 2011-12-16 22:51:43]

On 2011/12/16 22:44:32, Marc-Antoine Ruel wrote:
> lgtm

Thank you for the quick review!

> > We are now building Debug chrome with /O1 when we make a build targeted to
run
> > under Dr. Memory (for improved performance) which does perform some
> > optimizations
> 
> This is a bit sad, I would have preferred to assume debug is debug. Oh well.

Actually, using /O1 minus opts was already being done for the TSan/Win bot on
the Memory FYI waterfall.  We may be able to share that builder by using /O1 for
Dr. Memory.

[commit-bot: I haz the power, 2011-12-17 02:16:27]

Change committed as 114893