Comment #4 on issue 302 by liuj...@google.com: Skip(count) fails when count
offset >2^32 count type should be off_t, not int
http://code.google.com/p/protobuf/issues/detail?id=302
Well, for the java version, I think you mean the LimitedInputStream. That
class has to use long to override its super class java.io.FilterInputStream;
If you take a look at the c.g.protobuf.CodedInputStream, it uses int32 for
size type.
public void skipRawBytes(final int size) throws IOException
We cannot simply change the Skip() to accept int64 because:
1> In protobuf, we use "int" to describe buffer size, e.g. the constructor
of CodedInputStream, PushLimit(), GetDirectBufferPointer(), etc. We cannot
change Skip() to accept int64, while leaving others still expecting int32.
Users may be confused by the inconsistent size type. Even if we change
Skip() to int64, the rest of code will still assumes the underlying buffer
size is within int32, which can cause potential problems.
2> ZeroCopyStream interface uses int32. Changing the interface function
signature means we need to update all the implementation classes. We could
do a large code refactoring inside google, but we cannot change code
outside the company. Changing the interface may break existing opensource
protobuf users. I believe there are many clients who implement
ZeroCopyStream for their IO layer.
For your case, I think there are many workarounds. e.g. ensuring the chunk
doesn't exceed int32max, and use ZeroCopyStream to operate on each chunk.
Or maybe implement your own ZeroCopyStream, which ports large int operation
to protobuf API.
--
You received this message because you are subscribed to the Google Groups "Protocol
Buffers" group.
To post to this group, send email to protobuf@googlegroups.com.
To unsubscribe from this group, send email to
protobuf+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/protobuf?hl=en.
