Add -pedantic to Makefile.standalone to match production.

Add -pedantic to Makefile.standalone to match production.

This also requires modifying the ICE_CACHELINE_BOUNDARY macro to avoid a warning about anonymous structs:

src/IceUtils.h:132:3: warning: anonymous structs are a GNU extension [-Wgnu-anonymous-struct]

BUG= none
R=jfb@chromium.org

Committed: https://gerrit.chromium.org/gerrit/gitweb?p=native_client/pnacl-subzero.git;a=commit;h=26f9433e4e033d66574fa0d74392a504f8e491c0

[Jim Stichnoth, 2015-01-28 00:57:04]

stichnot@chromium.org changed reviewers:
+ jfb@chromium.org, jvoung@chromium.org

[JF, 2015-01-28 01:28:46]

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h
File src/IceDefs.h (right):

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h#newcode139
src/IceDefs.h:139: } TOKENPASTE2(Boundary, __LINE__)
You don't need the fake usage as long as you instantiate the struct.

But all of this is in vain because the struct must occupy some storage, it isn't
zero-sized... So we're back to:

  __attribute__((aligned(Align))) int : 0;

[Jim Stichnoth, 2015-01-28 03:26:37]

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h
File src/IceDefs.h (right):

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h#newcode139
src/IceDefs.h:139: } TOKENPASTE2(Boundary, __LINE__)
On 2015/01/28 01:28:45, JF wrote:
> You don't need the fake usage as long as you instantiate the struct.
> 
> But all of this is in vain because the struct must occupy some storage, it
isn't
> zero-sized... So we're back to:
> 
>   __attribute__((aligned(Align))) int : 0;

Done.

[JF, 2015-01-28 07:14:46]

lgtm, up to you if you want to take the suggestion into account.

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h
File src/IceDefs.h (right):

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h#newcode139
src/IceDefs.h:139: } TOKENPASTE2(Boundary, __LINE__)
On 2015/01/28 03:26:37, stichnot wrote:
> On 2015/01/28 01:28:45, JF wrote:
> > You don't need the fake usage as long as you instantiate the struct.
> > 
> > But all of this is in vain because the struct must occupy some storage, it
> isn't
> > zero-sized... So we're back to:
> > 
> >   __attribute__((aligned(Align))) int : 0;
> 
> Done.

Richard makes a good point that you can wrap each field you want to be together
in an anonymous struct, and make that struct alignas.

That would be standard-friendly and more readable.

[Jim Stichnoth, 2015-01-28 15:39:18]

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h
File src/IceDefs.h (right):

https://codereview.chromium.org/883983002/diff/1/src/IceDefs.h#newcode139
src/IceDefs.h:139: } TOKENPASTE2(Boundary, __LINE__)
On 2015/01/28 07:14:45, JF wrote:
> On 2015/01/28 03:26:37, stichnot wrote:
> > On 2015/01/28 01:28:45, JF wrote:
> > > You don't need the fake usage as long as you instantiate the struct.
> > > 
> > > But all of this is in vain because the struct must occupy some storage, it
> > isn't
> > > zero-sized... So we're back to:
> > > 
> > >   __attribute__((aligned(Align))) int : 0;
> > 
> > Done.
> 
> Richard makes a good point that you can wrap each field you want to be
together
> in an anonymous struct, and make that struct alignas.
> 
> That would be standard-friendly and more readable.

Yeah, that is pretty nice and clear.  Except is there a good way to initialize
the fields that are now wrapped inside an anonymous struct?  E.g. this
conversion from MyClass1 to MyClass2:

class MyClass1 {
  std::mutex MyFieldLock;
  MyType MyField;
public:
  MyClass1(ArgType Arg) : MyField(Arg) {}
};

class MyClass2 {
  alignas(64) struct {
    std::mutex MyFieldLock;
    MyType MyField;
  } F;
public:
  // This doesn't work:
  // MyClass2(ArgType Arg) : F.MyField(Arg) {}
  // This may work, depending on MyType and ArgType:
  MyClass2(ArgType Arg) { F.MyField = Arg; }
};

The other issue is that if you have N fields needing cache alignment, you
generally want N+1 instances of alignas(), so that fields after the explicitly
aligned ones start on a new cache line.  So even with the aligned (anonymous)
struct solution, you still need one more boundary marker.

[Jim Stichnoth, 2015-01-28 16:37:25]

Committed patchset #2 (id:20001) manually as
26f9433e4e033d66574fa0d74392a504f8e491c0 (presubmit successful).