Martin Sustrik
Martin Sustrik
More info here: https://github.com/sustrik/libdill/blob/master/rfc/bsd-socket-api-revamp.md#protocol-termination
No, there should be a dedicated btls_close() function to close the socket down cleanly.
The way to close hastily is hclose(). The way to close cleanly is <protocol-name>_close(). The reason there is no unified way to do the latter is that protocol initialization/termination is...
Ack. btls doesn't fit the spec at the moment. We should fix it.
Yes. It would be welcome. The text is generated from .md files in gh-pages branch, e.g.: https://github.com/sustrik/libdill/blob/gh-pages/tutorial.md
Fallback options that are almost never used are a maintenance liability. Having to support DILL_POLL and DILL_ARCH_FALLBACK is bad enough. Why add more?
An alternative is to have prioritized choose(), i.e. if there are multiple active clauses, always choose the one which comes first in the array. To make prioritized queue with three...
Some thoughts in no specific order: 1. In general, I prefer one-coroutine-one-input-channel model. The reason is that multiple input channels lead to heavy use of choose and from there it's...
It would be nice to see some use cases. The most prioritized queue solutions I've seen were dealing with cancelation events (e.g. OOB in TCP). However, we already have a...
What about this approach? ``` coroutine worker(int input, int oob) { while(1) { msg_t msg; chrecv(input, &msg, sizeof(msg), -1); process_input(msg); int rc = chrecv(oob, &msg, sizeof(msg), 0); if(rc == 0)...