Invalid UTF-8 in strings are not rejected

[dsnet]

The proto2 and proto3 specifications state that the string type must be encoded as valid UTF-8. Currently the Go implementation neither checks that the string is valid when marshaling, nor checks that the string is valid when unmarshaling. This is contrary to the stated specification and the C++ implementation, which does reject these cases.

However, making this change is likely to break a number of users.

[dsnet]

Change made on dev branch. See #499

[orian]

For a people googling, in some cases a protobuf message created in Go is causing a Java exception: com.google.protobuf.InvalidProtocolBufferException: Protocol message had invalid UTF-8

This bug was a reason for that.

[glasser]

We're interested in using this soon. What's the status of the dev branch? Is it likely to be released to master at some point?

[dsnet]

#479. It's happening on April 30th.