Disable Win32k renderer lockdown if DirectWrite is disabled.

Disable Win32k renderer lockdown if DirectWrite is disabled.

Warn the user if --disable-direct-write is set and Win32k renderer lockdown has been disabled as a result of this unsupported flag.

BUG=442157

[Will Harris, 2015-01-21 03:54:04]

wfh@chromium.org changed reviewers:
+ cpu@chromium.org

[Will Harris, 2015-01-21 03:55:40]

cpu@ PTAL first pass

I feel this is a bit ugly since I have to pull ui/gfx/switches.cc out into
content, but it's what other projects have done...?

Alternative would be to move kDisableDirectWrite into
/src/content/public/common/content_switches.cc

WDYT?

[cpu_(ooo_6.6-7.5), 2015-01-22 01:21:34]

cpu@chromium.org changed reviewers:
+ jam@chromium.org

[cpu_(ooo_6.6-7.5), 2015-01-22 01:22:03]

adding John for the layering with content and gfx

https://codereview.chromium.org/857123002/diff/40001/content/public/common/co...
File content/public/common/content_switches.cc (right):

https://codereview.chromium.org/857123002/diff/40001/content/public/common/co...
content/public/common/content_switches.cc:968: bool
IsWin32kRendererLockdownEnabled(bool ignore_directwrite) {
not loving this contract, I mean interface.

https://codereview.chromium.org/857123002/diff/40001/ui/gfx/gfx_export.h
File ui/gfx/gfx_export.h (right):

https://codereview.chromium.org/857123002/diff/40001/ui/gfx/gfx_export.h#newc...
ui/gfx/gfx_export.h:9: #if defined(WIN32)
no idea what this is.

[Will Harris, 2015-01-22 01:26:01]

wfh@chromium.org changed reviewers:
+ scottmg@chromium.org

[Will Harris, 2015-01-22 01:26:59]

Also added scottmg for the gfx_export.h fu

What is safe to access from where to avoid a layering violation - e.g. can I
access content/public from ui/gfx?

If so, I could just move the directwrite switch definitions to content/public

[cpu_(ooo_6.6-7.5), 2015-01-22 02:25:52]

cpu@chromium.org changed reviewers:
+ ananta@chromium.org

[cpu_(ooo_6.6-7.5), 2015-01-22 02:25:53]

ananta had some ideas.

[jam, 2015-01-22 17:00:16]

On 2015/01/22 01:26:59, Will Harris wrote:
> Also added scottmg for the gfx_export.h fu
> 
> What is safe to access from where to avoid a layering violation - e.g. can I
> access content/public from ui/gfx?

no, ui doesn't know about content 

> 
> If so, I could just move the directwrite switch definitions to content/public

https://codereview.chromium.org/857123002/diff/40001/chrome/browser/ui/startu...
File chrome/browser/ui/startup/bad_flags_prompt.cc (right):

https://codereview.chromium.org/857123002/diff/40001/chrome/browser/ui/startu...
chrome/browser/ui/startup/bad_flags_prompt.cc:30: #if defined(OS_WIN)
nit: here and in other places, you don't need the ifdef. just include it for all
platforms

[Will Harris, 2015-01-22 21:07:53]

On 2015/01/22 17:00:16, jam wrote:
> On 2015/01/22 01:26:59, Will Harris wrote:
> > Also added scottmg for the gfx_export.h fu
> > 
> > What is safe to access from where to avoid a layering violation - e.g. can I
> > access content/public from ui/gfx?
> 
> no, ui doesn't know about content 

So just to confirm - can content know about ui?  If so, the general approach of
the CL is fine?  I can address comments/nits if so, otherwise please explain an
alternative approach.

[ananta, 2015-01-22 22:03:43]

On 2015/01/22 21:07:53, Will Harris wrote:
> On 2015/01/22 17:00:16, jam wrote:
> > On 2015/01/22 01:26:59, Will Harris wrote:
> > > Also added scottmg for the gfx_export.h fu
> > > 
> > > What is safe to access from where to avoid a layering violation - e.g. can
I
> > > access content/public from ui/gfx?
> > 
> > no, ui doesn't know about content 
> 
> So just to confirm - can content know about ui?  If so, the general approach
of
> the CL is fine?  I can address comments/nits if so, otherwise please explain
an
> alternative approach.

content does know about ui. Wouldn't changing the
IsWin32kRendererLockdownEnabled function to return
false if the ShouldUseDirectWrite function returns false be good enough?

[jam, 2015-01-22 23:29:07]

On 2015/01/22 21:07:53, Will Harris wrote:
> On 2015/01/22 17:00:16, jam wrote:
> > On 2015/01/22 01:26:59, Will Harris wrote:
> > > Also added scottmg for the gfx_export.h fu
> > > 
> > > What is safe to access from where to avoid a layering violation - e.g. can
I
> > > access content/public from ui/gfx?
> > 
> > no, ui doesn't know about content 
> 
> So just to confirm - can content know about ui?  

https://code.google.com/p/chromium/codesearch#search/&q=%22%5C%22ui/%22%20fil...

yes :)

> If so, the general approach of
> the CL is fine?  I can address comments/nits if so, otherwise please explain
an
> alternative approach.

[Will Harris, 2015-01-22 23:35:58]

> content does know about ui. Wouldn't changing the
> IsWin32kRendererLockdownEnabled function to return
> false if the ShouldUseDirectWrite function returns false be good enough?

ShouldUseDirectWrite calls into skia and other files within ui.  This works fine
except for the crazy 64-bit on 32-bit parts of gyp where we pull in a very
limited set of dependencies with 64-bit overrides - and gfx wasn't one of them. 
See patch set 1 and the try bot results for what I mean:

crash_service_win64.content_switches.obj :error LNK2019: unresolved external
symbol "bool __cdecl gfx::win::ShouldUseDirectWrite(void)

I really don't want to have to create a 64-bit skia_win64 target just to get
this working...

I'll continue on this path, and fix up the nits and see where that gets things
to.

[Will Harris, 2015-01-22 23:59:29]

I spoke to ananta I'm going to simplify this CL and remove the infobar (push
that to another CL)