Previously, we were attempting to read the entire message in one call to fillBuffer(). This was doomed to fail if the message had more than Integer.MAX_VALUE bytes. After this diff, we will call fillBuffer() separately for each segment. This approach turns out to be simpler, too. It might imply a small performance hit for messages with many small segments, but such messages are discouraged anyway. This diff also adds more overflow checking in the surrounding logic. |
||
---|---|---|
.. | ||
src | ||
pom.xml |