Deviation/deviate grammar not checked properly

```
I think this is only an omission due to the nature of deviate statements;
grammar for deviations is not checked correctly. The deviate-stmt seems to be
the only YANG statement which “dynamically” changes substatement grammar
based on a statement's argument (all other do this based on keywords, perhaps
only the type-stmt is similar with it's restrictions).

It is not obvious from the text in RFC6020 but YANG grammar specifies the
following:
1. deviation-stmt must have at least one deviate substatement
2. if a deviate-stmt with “not-supported” argument is present within a
deviation-stmt, then no other deviate-stmt may be present as it's sibling
3. deviate-stmts with arguments “add”, “delete” and “replace” may
appear any number of times within a deviation-stmt as long as a
“not-supported” deviate-stmt is not present.
4. a deviate-stmt with a “not-supported” argument must not contain any
substatements
5. a deviate-stmt with arguments “add”, “delete” and “replace” may
or may not contain substatements, but I suggest generating a warning when such
statements are empty
6. a deviate-stmt with argument “add” may only contain the following
substatements which may appear in any order; units-stmt (0..1), must-stmt
(0..n), unique-stmt (0..n), default-stmt (0..1), config-stmt (0..1),
mandatory-stmt (0..1), min-elements-stmt (0..1) and max-elements-stmt (0..1)
7. a deviate-stmt with argument “delete” may only contain the following
substatements which may appear in any order; units-stmt (0..1), must-stmt
(0..n), unique-stmt (0..n) and default-stmt (0..1)
8. a deviate-stmt with argument “replace” may only contain the following
substatements which may appear in any order; type-stmt (0..1), units-stmt
(0..1), default-stmt (0..1), config-stmt (0..1), mandatory-stmt (0..1),
min-elements-stmt (0..1) and max-elements-stmt (0..1)

You probably wanted to check this within validation phases for deviation-stmt
and deviate-stmt, since you cannot specify such a grammar within grammar.py,
because you'd have to check a statement's argument. You should do it in a way
similar to how you check restrictions of the type-stmt.

I confirmed this in the latest svn rev739.

RFC6020 text in 7.18.3 and it's sub-chapters is extremely vague on this matter
but the grammar in chaper 12 is quite clear. I enclosed a YANG module with a
bunch of grammar errors. Pyang does report some errors with this YANG file, but
they shouldn't even be reported since a grammar check should have higher
priority. For example, attempting to add a type statement should never even be
allowed.
```

Original issue reported on code.google.com by `jernej.t...@gmail.com` on 11 Sep 2012 at 8:16

Attachments:
- [deviation-deviate-errs@2012-09-06.yang](https://storage.googleapis.com/google-code-attachments/pyang/issue-76/comment-0/deviation-deviate-errs@2012-09-06.yang)