Server Problem with operation length > J_STRIPE_SIZE about julea

[oliver-pola]

Fixed the write part, oliver-pola@7e14ac8, oliver-pola@1874f8b

That was about partial writes for a long message. Read will be about partial read and sending multiple replies and is a little more challenging.

Will send pull request when I fixed read.

[michaelkuhn]

This is not documented behavior, just expected behavior. :-)
It therefore makes sense to fix this, especially since not all clients have a concept of stripes (the distributed object client splits the data into stripes, while the object client does not).

Thanks for working on this!

[oliver-pola]

Pull request #27 fixes the write.

Fixing read did not go as expected. But read does not crash and return bytes_read = 0.

If leaving the FIXME comment is enough to tell that there could be further improvements, yo can close this after PR merge.

[michaelkuhn]

Let's leave this open until the read part is fixed, too.

[michaelkuhn]

This has been fixed in 92839e3 and be79089 for clients and servers configured with the same maximum operation size. Error reporting/handling still has to be improved for the case where clients have a larger maximum operation size, though.