Parser erroneously allows whitespace in package declaration.

[JakeWharton]

We had a case where protoc allowed a package declaration that had newlines in it but other tooling based on the spec failed to parse the file. I believe this should have failed to parse inside protoc as well.

Relevant parts of the spec:

package = "package" fullIdent ";"
fullIdent = ident { "." ident }
ident = letter { letter | decimalDigit | "_" }
letter = "A" … "Z" | "a" … "z"
decimalDigit = "0" … "9"

Example showing protoc allowing newlines:

$ echo "package example.

foo

;

message Foo {}" > test.proto

$ protoc --version
libprotoc 3.0.0

$ protoc --java_out=. test.proto
[libprotobuf WARNING google/protobuf/compiler/parser.cc:547] No syntax specified for the proto file: test.proto. Please use 'syntax = "proto2";' or 'syntax = "proto3";' to specify a syntax version. (Defaulted to proto2 syntax.)

$ tree
.
├── example
│   └── foo
│       └── Test.java
└── test.proto

[haberman]

Yes this seems like a bug.

[xfxyjwf]

@anandolee It's the syntax spec that needs a fix here. A package name spreading across multiple lines is needed when the name is very long. For example:

message Foo {
  optional a.very.very.long.name.that.can.not.fit.into.one.line.without.exceeding
      .80.char.limit.Bar value = 1;
}

[haberman]

@xfxyjwf Is the syntax spec checked in anywhere? What actually needs to be fixed here?

[xfxyjwf]

The syntax spec here:
https://developers.google.com/protocol-buffers/docs/reference/proto2-spec

The spec needs to specify where spaces are allowed and where not.