Dart bindings: support struct versioning.

Dart bindings: support struct versioning.

BUG=461397
TEST=the previously-skipped validation tests pass
R=zra@google.com

Committed: https://chromium.googlesource.com/external/mojo/+/10c55d8ba49fc2d55be0e5a377450beda4f287ef

[yzshen1, 2015-03-30 22:03:37]

yzshen@chromium.org changed reviewers:
+ zra@chromium.org

[yzshen1, 2015-03-30 22:03:38]

Hi, Zachary.

Would you please take a look? Thanks!

[zra, 2015-03-31 14:46:41]

zra@google.com changed reviewers:
+ zra@google.com

[zra, 2015-03-31 14:46:41]

lgtm with question

https://codereview.chromium.org/1043903004/diff/20001/mojo/public/tools/bindi...
File mojo/public/tools/bindings/generators/dart_templates/struct_definition.tmpl
(right):

https://codereview.chromium.org/1043903004/diff/20001/mojo/public/tools/bindi...
mojo/public/tools/bindings/generators/dart_templates/struct_definition.tmpl:113:
if (mainDataHeader.version <= kVersions.last.version) {
I may not understand what the intended semantics are, but it looks like an
incoming message can have a crazy version number (> kVersions.last.version) so
long as the message size matches the size of a known version. I guess that would
allow inc'ing the version number without changing the message size. Is that the
intention here?

[yzshen1, 2015-03-31 15:26:26]

Thanks Zachary!

https://codereview.chromium.org/1043903004/diff/20001/mojo/public/tools/bindi...
File mojo/public/tools/bindings/generators/dart_templates/struct_definition.tmpl
(right):

https://codereview.chromium.org/1043903004/diff/20001/mojo/public/tools/bindi...
mojo/public/tools/bindings/generators/dart_templates/struct_definition.tmpl:113:
if (mainDataHeader.version <= kVersions.last.version) {
On 2015/03/31 14:46:41, zra wrote:
> I may not understand what the intended semantics are, but it looks like an
> incoming message can have a crazy version number (> kVersions.last.version) so
> long as the message size matches the size of a known version. I guess that
would
> allow inc'ing the version number without changing the message size. Is that
the
> intention here?

Encountering a version number bigger than the latest version number that we know
about is possible. It means the other side is built with a newer version of the
API definition. In this case, those unknown fields are simply ignored by us.

Also, it is possible that in some cases we increase the version number (adding
more fields) without increasing the message size, because our packing algorithm
places a newly-added field in the first big-enough "hole" in the layout, which
may be some padding area in the middle.

[yzshen1, 2015-03-31 15:28:47]

Committed patchset #2 (id:20001) manually as
10c55d8ba49fc2d55be0e5a377450beda4f287ef (presubmit successful).