Closed wallento closed 8 years ago
The timeout is 0 ==unlimited in glip_write_b, so it should block forever.
I notice, however, two problems:
This means the tcp implementation should never return a timeout when timeout == 0, correct?
Does this workaround hurt for the moment? Its not a problem to capture partial writes anyways I suppose.
exactly, timeout == 0 means no -ETIMEDOUT. The actual bug is in the tcp implementation.
The workaround doesn't hurt, however, so let's add it for now. But please create another ticket so we don't forget about the other two issues.
Thanks, I have started already and will have a PR later today
The loopback_measure should also track the sub-index to the data for partial blocking writes. When we hit a timeout, the size_written progresses the amount of overall written data. We also need to offset into the data then to not transmit unexpected data.