Attention is currently required from: osmith, Hoernchen, neels, laforge, pespin, fixeria.
11 comments:
Patchset:
In general I think we need to think about naming / namespace pollution. […]
Are you talking about a map file for all of libosmocore or just a libosmoio map file?
File src/osmo_io.c:
Patch Set #3, Line 50: struct iofd_backend_ops g_iofd_ops;
then better name it with an osmo_ prefix.
This should be resolved if we end up introducing a .map file, right?
File src/osmo_io.c:
Patch Set #4, Line 228: msg_pending = iofd_msgb_alloc(iofd);
This function now sets pending_out to the remaining msgb and iofd_handle_segmented_read is responsible for setting this. This also saves us one msgb_pending call in the handle_more case below.
Patch Set #4, Line 299: * \returns 0 in case of success; a negative value in case of error
Even though almost no code uses it I guess you're right.
struct os
Done
Patch Set #4, Line 367: iofd->io_ops = *ioops;
I agree, the name should be mandatory.
Patch Set #4, Line 414: * \param[in] headroom the headroom of the msgb when receiving data
Ack
File src/osmo_io_internal.h:
I'm wondering if it had any advantage to have only a single function with an argument stating the "o […]
I'll keep it as it is
struct {
bool read_enabled;
bool read_pending;
bool write_pending;
bool write_enabled;
/* TODO: index into array of registered fd's? */
} uring;
Ack
Done
File src/osmo_io_poll.c:
Patch Set #4, Line 104: struct osmo_sockaddr saddr;
Done
Patch Set #4, Line 128: if (rc < 0) {
Does the kernel use addrlen for more than just checking if it's long enough for the variant? In any case I guess we only want to copy to the kernel what we really need.
I think should be a helper function in libosmocore like osmo_sockaddr_size() that returns the size depending on the family that way we can reuse it elsewhere.
To view, visit change 30934. To unsubscribe, or for help writing mail filters, visit settings.