C bindings pt3: Type table definitions and barebones files to get generated code to compile.

C bindings pt3: Type table definitions and barebones files to get generated code to compile.

With this checked in, mojoms in //mojo/public/interfaces/bindings/tests will
be generated and compiled, but not yet used.

Changes include:
- Definitions of some structs that are used to construct type tables.
- Definitions for MojomString, MojomArray, MojomMap, MojomStruct.

I don't plan to test the type tables directly, but instead test the
things that use them (New(), ComputeSerializedSize(), Encode..(),
Decode..(), Validate..()).

R=viettrungluu@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/c747b91aa53c244740e99d5b41994535cce2a658

[vardhan, 2016-06-17 20:08:04]

PTAL

[viettrungluu, 2016-06-17 20:37:19]

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/BUILD.gn (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/BUILD.gn:12: "lib/type_table.h",
Being in the lib directory means that this should go under "Implementation
details" (below), no?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/array.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/array.h:17: struct MojomArray {
Maybe this should be called MojomArrayHeader instead?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/interface.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/interface.h:19: MojoHandle handle;
Strictly speaking, isn't this either a handle or an index?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/lib/type_table.c (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.c:11: 0,  // num_elements
Did you clang-format this?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/lib/type_table.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:7: // the pointers and handles are, and
references to the type tables that further
- where all the pointers and handles are in what?
- ... and *has* references ...?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:15: // A type table for a mojom struct
is an array of entries of type
So, if I understand you correctly, the |const void* elem_table|s either point to
an array of structs (in the struct and union cases) or a single struct (in the
array case)?

Would it be clearer instead to always point to a struct, which (in the struct
and union cases) could then point to an array? (That way, you'd also have a
place to put the number of elements in the array, instead of using
|keep_going|.)

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:68: enum MojomPointerTableElemType
elem_type;
As a matter of readability, this should be immediately next to (probably
preceding) |elem_table|.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:71: // Are there more entries after
this? (the last |MojomPointerTableStructEntry|
At the very least, you're missing a closing ')'....

But also, maybe you should call this "is_last" (or "is_last_entry", or
"last_entry") instead of "keep_going"?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/map.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/map.h:20: struct MojomMap {
MojomMapHeader?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/string.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/string.h:13: struct MojomString {
MojomStringHeader?

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/union.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/union.h:7: 
This file seems awfully empty.

[vardhan, 2016-06-21 16:07:35]

vardhan@google.com changed reviewers:
+ azani@chromium.org

[vardhan, 2016-06-21 16:07:36]

PTAL

(adding azani@ to review the generator part)

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/BUILD.gn (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/BUILD.gn:12: "lib/type_table.h",
On 2016/06/17 20:37:18, viettrungluu wrote:
> Being in the lib directory means that this should go under "Implementation
> details" (below), no?

Done.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/array.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/array.h:17: struct MojomArray {
On 2016/06/17 20:37:19, viettrungluu wrote:
> Maybe this should be called MojomArrayHeader instead?

Done.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/interface.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/interface.h:19: MojoHandle handle;
On 2016/06/17 20:37:19, viettrungluu wrote:
> Strictly speaking, isn't this either a handle or an index?

true -- i'll add a comment.  I was thinking about making it a union, but it's
still awkard to use since an invalid encoded handle offset is '-1' of the
MojoHandle type (which is currently uint32_t MAX), so we would have to continue
using the |handle| value as MojoHandle..

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/lib/type_table.c (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.c:11: 0,  // num_elements
On 2016/06/17 20:37:19, viettrungluu wrote:
> Did you clang-format this?

Done.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.c:11: 0,  // num_elements
On 2016/06/17 20:37:19, viettrungluu wrote:
> Did you clang-format this?

Done.

(I only ever do "git cl format," which doesn't fix this formatting, but now I'll
start doing clang-format.  thanks)

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/lib/type_table.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:7: // the pointers and handles are, and
references to the type tables that further
On 2016/06/17 20:37:19, viettrungluu wrote:
> - where all the pointers and handles are in what?
> - ... and *has* references ...?

Done.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:15: // A type table for a mojom struct
is an array of entries of type
On 2016/06/17 20:37:19, viettrungluu wrote:
> So, if I understand you correctly, the |const void* elem_table|s either point
to
> an array of structs (in the struct and union cases) or a single struct (in the
> array case)?
Yes.

> Would it be clearer instead to always point to a struct, which (in the struct
> and union cases) could then point to an array? (That way, you'd also have a
> place to put the number of elements in the array, instead of using
> |keep_going|.)
Hm.. the trade-off being making the generator generate more structs (i.e.
generating an extra struct that only encodes a pointer to the array and array
size). i'm not sure that will be all that much easier to understand than the
|keep_going| scheme, but I understand the weirdness in encoding it in a
run-length fashion, so Done.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:68: enum MojomPointerTableElemType
elem_type;
On 2016/06/17 20:37:19, viettrungluu wrote:
> As a matter of readability, this should be immediately next to (probably
> preceding) |elem_table|.

Done.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/lib/type_table.h:71: // Are there more entries after
this? (the last |MojomPointerTableStructEntry|
On 2016/06/17 20:37:19, viettrungluu wrote:
> At the very least, you're missing a closing ')'....
> 
> But also, maybe you should call this "is_last" (or "is_last_entry", or
> "last_entry") instead of "keep_going"?

(removed all-together)

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/map.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/map.h:20: struct MojomMap {
On 2016/06/17 20:37:19, viettrungluu wrote:
> MojomMapHeader?

Done.

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
File mojo/public/c/bindings/string.h (right):

https://codereview.chromium.org/2072903002/diff/60001/mojo/public/c/bindings/...
mojo/public/c/bindings/string.h:13: struct MojomString {
On 2016/06/17 20:37:19, viettrungluu wrote:
> MojomStringHeader?

Done.

[viettrungluu, 2016-06-21 18:08:30]

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
File mojo/public/c/bindings/lib/type_table.h (right):

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:47: // Structs and unions are described
using this struct; |entries| is of type
I suggest you just make two separate structs (with suitably descriptive names --
see below) with correct types.

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:72: // |elem_table| can be
|MojomTypeTable|, MojomTypeTableArrayEntry|, or NULL.
I think you should rationalize the names of things. Think of it this way: if
these were C++ classes, instead of C structs, how would you name them?

You have two kinds of things:
1. Things that describe a type. (These stand alone, but sometimes may be in a
table contained in a thing of type 1.)
2. Things that do not stand alone, but only occur in something of type 1.

Let's call type 1 things "type descriptors". You split MojomTypeTable into
MojomTypeDescriptorStruct (or maybe it should be MojomStructTypeDescriptor?) and
MojomTypeDescriptorUnion. And MojomTypeTableArrayEntry should be renamed to
MojomTypeDescriptorArray.

Type 2 things are really "inner types" for things of type 1, so should be named
correspondingly.

E.g., MojomTypeTableStructEntry maybe becomes MojomTypeDescriptorStructEntry (or
something like that), and similarly for unions.

(Incidentally, what about maps?)

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:114: extern const struct
MojomTypeTableArrayEntry MojomTypeTableStringEntry;
It's very odd that MojomTypeTableStringEntry is not a type, but an instance of a
type.

[vardhan, 2016-06-21 22:42:59]

ptal

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
File mojo/public/c/bindings/lib/type_table.h (right):

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:47: // Structs and unions are described
using this struct; |entries| is of type
On 2016/06/21 18:08:30, viettrungluu wrote:
> I suggest you just make two separate structs (with suitably descriptive names
--
> see below) with correct types.

Done.

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:72: // |elem_table| can be
|MojomTypeTable|, MojomTypeTableArrayEntry|, or NULL.
On 2016/06/21 18:08:30, viettrungluu wrote:
> I think you should rationalize the names of things. Think of it this way: if
> these were C++ classes, instead of C structs, how would you name them?
> 
> You have two kinds of things:
> 1. Things that describe a type. (These stand alone, but sometimes may be in a
> table contained in a thing of type 1.)
> 2. Things that do not stand alone, but only occur in something of type 1.
> 
> Let's call type 1 things "type descriptors". You split MojomTypeTable into
> MojomTypeDescriptorStruct (or maybe it should be MojomStructTypeDescriptor?)
and
> MojomTypeDescriptorUnion. And MojomTypeTableArrayEntry should be renamed to
> MojomTypeDescriptorArray.
> 
> Type 2 things are really "inner types" for things of type 1, so should be
named
> correspondingly.
> 
> E.g., MojomTypeTableStructEntry maybe becomes MojomTypeDescriptorStructEntry
(or
> something like that), and similarly for unions.
Done.

> (Incidentally, what about maps?)
Maps are mojom structs with mojom arrays -- i just added a comment at the top of
the file explaining this.

https://codereview.chromium.org/2072903002/diff/160001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:114: extern const struct
MojomTypeTableArrayEntry MojomTypeTableStringEntry;
On 2016/06/21 18:08:30, viettrungluu wrote:
> It's very odd that MojomTypeTableStringEntry is not a type, but an instance of
a
> type.

Renamed to MojomStringTypeDescription.

[viettrungluu, 2016-06-21 23:06:53]

Getting close....

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
File mojo/public/c/bindings/lib/type_table.h (right):

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:5: // This file contains structs used
for constructing type tables for generated
Maybe you should call them "type descriptors" instead of "type tables"?

Also, maybe you should rename the file? (Sorry.)

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:36: // A map is a struct with 2 arrays.
Probably you should prepend "Note: " to this.

Also, maybe you should indicate here the corresponding MojomTypeDescriptorX
types here (possibly "N/A" for handle/interface/POD).

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:54: // The type that |elem_table| points
to.
(And then here you could just refer to the comment above.)

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:77: struct MojomTypeDescriptorStruct {
Note that since this is C, you can actually put this declaration before that of
|struct MojomTypeDescriptorStructEntry|, which is kind of nice.

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:118: extern const struct
MojomTypeDescriptorArray MojomStringTypeDescription;
I was wonder, w.r.t. the name, whether it shouldn't be
g_mojom_string_type_description, or something like that.

Or maybe make it obvious that it's the MojomTypeDescriptorArray for strings in
some other way? (I can't tell offhand how you're naming generated structs, but
maybe you should follow a similar pattern?)

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
File mojo/public/c/bindings/string.h (right):

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/string.h:11: // A mojom string is a mojom array of UTF8
chars that need not be
nit: "UTF-8" (with hyphen)

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/cge...
File mojom/generators/c/cgen/type_table.go (right):

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/cge...
mojom/generators/c/cgen/type_table.go:153: Name: prefix + "__TypeTable",
rename this?

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/tem...
File mojom/generators/c/templates/type_table.tmpl.go (right):

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/tem...
mojom/generators/c/templates/type_table.tmpl.go:45: {{$array.NumElements}},
{{$array.Nullable}},
Do you really want hard tabs in this file?

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/tem...
mojom/generators/c/templates/type_table.tmpl.go:55: {{$entry.Offset}},
{{$entry.MinVersion}},
...

[vardhan, 2016-06-22 15:18:48]

PTAL

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
File mojo/public/c/bindings/lib/type_table.h (right):

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:5: // This file contains structs used
for constructing type tables for generated
On 2016/06/21 23:06:52, viettrungluu wrote:
> Maybe you should call them "type descriptors" instead of "type tables"?
> 
> Also, maybe you should rename the file? (Sorry.)

Done.

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:36: // A map is a struct with 2 arrays.
On 2016/06/21 23:06:52, viettrungluu wrote:
> Probably you should prepend "Note: " to this.
> 
> Also, maybe you should indicate here the corresponding MojomTypeDescriptorX
> types here (possibly "N/A" for handle/interface/POD).

Done.

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:54: // The type that |elem_table| points
to.
On 2016/06/21 23:06:52, viettrungluu wrote:
> (And then here you could just refer to the comment above.)

Done.

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:77: struct MojomTypeDescriptorStruct {
On 2016/06/21 23:06:53, viettrungluu wrote:
> Note that since this is C, you can actually put this declaration before that
of
> |struct MojomTypeDescriptorStructEntry|, which is kind of nice.

cool, done.

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_table.h:118: extern const struct
MojomTypeDescriptorArray MojomStringTypeDescription;
On 2016/06/21 23:06:53, viettrungluu wrote:
> I was wonder, w.r.t. the name, whether it shouldn't be
> g_mojom_string_type_description, or something like that.
Done.
 
> Or maybe make it obvious that it's the MojomTypeDescriptorArray for strings in
> some other way? (I can't tell offhand how you're naming generated structs, but
> maybe you should follow a similar pattern?)
array type descriptor names are generated from a counter
(mymodule_mystruct_myarray_0, mymodule_mystruct_myarray_1, etc.), which we don't
need for MojomStrings

(i realized shortly after i'd written the naming part that we could generate
array type descriptor names by name-mangling them without a counter, which would
let us reuse common types ("array_int" for array<int>, "map_int_array_int"
"map<int,array<int>>" that we only generate once within a mojom file.  i left
that as a TODO).

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
File mojo/public/c/bindings/string.h (right):

https://codereview.chromium.org/2072903002/diff/180001/mojo/public/c/bindings...
mojo/public/c/bindings/string.h:11: // A mojom string is a mojom array of UTF8
chars that need not be
On 2016/06/21 23:06:53, viettrungluu wrote:
> nit: "UTF-8" (with hyphen)

Done.

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/cge...
File mojom/generators/c/cgen/type_table.go (right):

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/cge...
mojom/generators/c/cgen/type_table.go:153: Name: prefix + "__TypeTable",
On 2016/06/21 23:06:53, viettrungluu wrote:
> rename this?

Done.

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/tem...
File mojom/generators/c/templates/type_table.tmpl.go (right):

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/tem...
mojom/generators/c/templates/type_table.tmpl.go:45: {{$array.NumElements}},
{{$array.Nullable}},
On 2016/06/21 23:06:53, viettrungluu wrote:
> Do you really want hard tabs in this file?

oops!  thanks

https://codereview.chromium.org/2072903002/diff/180001/mojom/generators/c/tem...
mojom/generators/c/templates/type_table.tmpl.go:55: {{$entry.Offset}},
{{$entry.MinVersion}},
On 2016/06/21 23:06:53, viettrungluu wrote:
> ...

Done.

[viettrungluu, 2016-06-22 15:33:02]

lgtm w/one more renaming

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
File mojo/public/c/bindings/lib/type_descriptor.h (right):

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_descriptor.h:77: const void* elem_table;
Probably you should call it elem_descriptor instead of elem_table (also in the
various comments, obv.).

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_descriptor.h:102: const void* elem_table;
"

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_descriptor.h:114: const void* elem_table;
"

[vardhan, 2016-06-22 15:44:27]

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
File mojo/public/c/bindings/lib/type_descriptor.h (right):

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_descriptor.h:77: const void* elem_table;
On 2016/06/22 15:33:02, viettrungluu wrote:
> Probably you should call it elem_descriptor instead of elem_table (also in the
> various comments, obv.).

Done.

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_descriptor.h:102: const void* elem_table;
On 2016/06/22 15:33:02, viettrungluu wrote:
> "

Done.

https://codereview.chromium.org/2072903002/diff/200001/mojo/public/c/bindings...
mojo/public/c/bindings/lib/type_descriptor.h:114: const void* elem_table;
On 2016/06/22 15:33:02, viettrungluu wrote:
> "

Done.

[vardhan, 2016-06-22 15:49:16]

Description was changed from

==========
C bindings pt3: Type table definitions and barebones files to get generated code
to compile.

With this checked in, mojoms in //mojo/public/interfaces/bindings/tests will
be generated and compiled, but not yet used.

Changes include:
- Definitions of some structs that are used to construct type tables.
- Definitions for MojomString, MojomArray, MojomMap, MojomStruct.

I don't plan to test the type tables directly, but instead test the
things that use them (New(), ComputeSerializedSize(), Encode..(),
Decode..(), Validate..()).

R=viettrungluu@chromium.org
==========

to

==========
C bindings pt3: Type table definitions and barebones files to get generated code
to compile.

With this checked in, mojoms in //mojo/public/interfaces/bindings/tests will
be generated and compiled, but not yet used.

Changes include:
- Definitions of some structs that are used to construct type tables.
- Definitions for MojomString, MojomArray, MojomMap, MojomStruct.

I don't plan to test the type tables directly, but instead test the
things that use them (New(), ComputeSerializedSize(), Encode..(),
Decode..(), Validate..()).

R=viettrungluu@chromium.org

Committed:
https://chromium.googlesource.com/external/mojo/+/c747b91aa53c244740e99d5b419...
==========

[vardhan, 2016-06-22 15:49:20]

Committed patchset #12 (id:220001) manually as
c747b91aa53c244740e99d5b41994535cce2a658 (presubmit successful).