Allow for Skia to be built without dwrite 1.1

Allow for Skia to be built without dwrite 1.1

We still have developers on XP and Vista, so we would like to be able to add
an SK_IGNORE_DWRITE_1_SUPPORT flag in the same style as SK_IGNORE_ETC1_SUPPORT. It
looks like the code is already set up to fallback at runtime anyway, this just enforces that
fallback at compile time.

Patch originally from https://bugzilla.mozilla.org/show_bug.cgi?id=1053652
R=bungeman
BUG=skia:

[bungeman-skia, 2014-09-08 19:38:49]

bungeman@google.com changed reviewers:
+ bungeman@google.com
- bungeman@gmail.com

[bungeman-skia, 2014-09-08 21:10:04]

https://codereview.chromium.org/537633002/diff/20001/src/ports/SkTypeface_win...
File src/ports/SkTypeface_win_dw.h (right):

https://codereview.chromium.org/537633002/diff/20001/src/ports/SkTypeface_win...
src/ports/SkTypeface_win_dw.h:20: #ifndef SK_IGNORE_DWRITE_1_SUPPORT
Ick, I'm against adding yet another magic build level define. It's a really hard
sell for XP, maybe a bit less so for Vista.

dwrite.h: SDK7.0 (0x0601)
dwrite_1.h: SDK8.0 (0x0602)
dwrite_2.h: SDK8.1 (0x0603)

It looks like we could use winsdkver.h, (the casing of which changed from
WinSDKVer.h to winsdkver.h in SDK 7.1A) but avoid the madness that is
NTDDI_MAXVER which seems to be set to something semi-random with each SDK update
(it only actually matches the max supported NTDDI_VER in SDK7.1). As a result,
it seems like the easiest correct-ish thing to do here is put something like

#define SK_HAS_DWRITE_1 (WINVER_MAXVER >= 0x0602)

in SkDWrite.h and then use it like

#if SK_HAS_DWRITE_1 

in places like this. It would also be nice to #warn that you're building with
reduced functionality if you don't have the requisite support in your SDK. Note
that sometime soon-ish we will also end up doing something like this for color
support, since that requires dwrite_2.h.



I took a look at
http://lxr.mozilla.org/mozilla-central/source/build/autoconf/winsdk.m4 and it
seems rather silly, since it might as well just be

#define WINSDK_MAXVER NTDDI_VERSION_FROM_WIN32_WINNT(_WIN32_WINNT_MAXVER)

since, except for SDK7.1 that's what it's doing anyway, and for SDK7.1 it gives
the same result.

[gw280, 2014-09-09 06:22:44]

On 2014/09/08 21:10:04, bungeman1 wrote:
> It looks like we could use winsdkver.h, (the casing of which changed from
> WinSDKVer.h to winsdkver.h in SDK 7.1A) but avoid the madness that is
> NTDDI_MAXVER which seems to be set to something semi-random with each SDK
update
> (it only actually matches the max supported NTDDI_VER in SDK7.1). As a result,
> it seems like the easiest correct-ish thing to do here is put something like
> 
> #define SK_HAS_DWRITE_1 (WINVER_MAXVER >= 0x0602)
> 
> in SkDWrite.h and then use it like
> 
> #if SK_HAS_DWRITE_1 
> 
> in places like this. It would also be nice to #warn that you're building with
> reduced functionality if you don't have the requisite support in your SDK.
Note
> that sometime soon-ish we will also end up doing something like this for color
> support, since that requires dwrite_2.h.

That looks reasonable. I will sort that out tomorrow, thanks for the hints.

[bungeman-skia, 2014-09-09 15:19:36]

On 2014/09/09 06:22:44, gw280 wrote:
> On 2014/09/08 21:10:04, bungeman1 wrote:
> > It looks like we could use winsdkver.h, (the casing of which changed from
> > WinSDKVer.h to winsdkver.h in SDK 7.1A) but avoid the madness that is
> > NTDDI_MAXVER which seems to be set to something semi-random with each SDK
> update
> > (it only actually matches the max supported NTDDI_VER in SDK7.1). As a
result,
> > it seems like the easiest correct-ish thing to do here is put something like
> > 
> > #define SK_HAS_DWRITE_1 (WINVER_MAXVER >= 0x0602)
> > 
> > in SkDWrite.h and then use it like
> > 
> > #if SK_HAS_DWRITE_1 
> > 
> > in places like this. It would also be nice to #warn that you're building
with
> > reduced functionality if you don't have the requisite support in your SDK.
> Note
> > that sometime soon-ish we will also end up doing something like this for
color
> > support, since that requires dwrite_2.h.
> 
> That looks reasonable. I will sort that out tomorrow, thanks for the hints.

I happen to have an XP vm sitting around, so I took a look into this.
See https://codereview.chromium.org/552383002/ .

[bungeman-skia, 2014-09-10 19:17:06]

On 2014/09/09 15:19:36, bungeman1 wrote:
> On 2014/09/09 06:22:44, gw280 wrote:
> > On 2014/09/08 21:10:04, bungeman1 wrote:
> > > It looks like we could use winsdkver.h, (the casing of which changed from
> > > WinSDKVer.h to winsdkver.h in SDK 7.1A) but avoid the madness that is
> > > NTDDI_MAXVER which seems to be set to something semi-random with each SDK
> > update
> > > (it only actually matches the max supported NTDDI_VER in SDK7.1). As a
> result,
> > > it seems like the easiest correct-ish thing to do here is put something
like
> > > 
> > > #define SK_HAS_DWRITE_1 (WINVER_MAXVER >= 0x0602)
> > > 
> > > in SkDWrite.h and then use it like
> > > 
> > > #if SK_HAS_DWRITE_1 
> > > 
> > > in places like this. It would also be nice to #warn that you're building
> with
> > > reduced functionality if you don't have the requisite support in your SDK.
> > Note
> > > that sometime soon-ish we will also end up doing something like this for
> color
> > > support, since that requires dwrite_2.h.
> > 
> > That looks reasonable. I will sort that out tomorrow, thanks for the hints.
> 
> I happen to have an XP vm sitting around, so I took a look into this.
> See https://codereview.chromium.org/552383002/ .

I'm closing this issue since https://codereview.chromium.org/552383002/ has
landed. If there are still issues, please let us know.