I've been testing against Dispatch and I'm seeing lots of timeouts on
It turns out that pn_read_frame is used to parse text that is not a
frame header. It's actually the AMQP header so it interprets the string
"AMQP" as the frame length. Now that we're testing frames against the
max-frame-size, this causes a framing error which drops the connection.
It's interesting to note that this mis-parse is not a regression. It's
been happening all along but was not noticed because "AMQP" is much
larger than the typical number of available bytes. Before RC2,
pn_read_frame would simply return zero for this non-frame and things
muddled along more or less correctly.
On 08/10/2015 02:34 PM, Robbie Gemmell wrote:
I have put up a second cut for 0.10, please test it and vote accordingly.
This fixes a crash (PROTON-976) reported against the previous spin,
and adds a known issue (PROTON-975) of possible failure using the
DIGEST-MD5 mechanism against non-proton servers.
The release archive and sig/checksums can be grabbed from:
Maven artifacts for the Java bits can be found in a temporary staging repo at: