Issues with BufferClaim and ideas

[iceberglet]

Hi, while using aeron & aeron cluster, I came across the following pain points around BufferClaim:

1. It is very hard to test, because most likely the actual code will invoke the aeron API tryClaim with its own BufferClaim instance, which is hard to mock behaviours even if captured.

I don't have good ideas which wouldn't involving API breaking changes : ( Just raising a point here in case someone has good ideas

2. The tryClaim methods in various aeron cluster components will insert extra headers. Example being ClusteredServiceAgent which adds a SESSION_HEADER_LENGTH. When invoking on such tryClaim the programmer has to be very careful and apply a correct offset, as the BufferClaim's offset is hardcoded to be only aeron transport header frame size.

Suggestion:

• add a new field msgOffset into BufferClaim
• add two more methods advanceMsgOffset(int) and msgOffset()
• when calling wrap on BufferClaim, initialize msgOffset to HEADER_LENGTH (which is aeron transport's header)
• every stage of tryClaim where an inner header is added, advanceMsgOffset should be called to increment msgOffset. In the ClusteredServiceAgent example, it should increment by SESSION_HEADER_LENGTH
• programmer to switch to using msgOffset() rather than offset() when wrapping encoders, avoiding confusions