Attention is currently required from: arehbein, laforge, pespin.

daniel has posted comments on this change. ( https://gerrit.osmocom.org/c/libosmo-netif/+/33193 )

Change subject: Add osmo_io support to osmo_stream_cli and osmo_stream_srv ......................................................................

Patch Set 10:

(6 comments)

File src/stream.c:

https://gerrit.osmocom.org/c/libosmo-netif/+/33193/comment/5368ac86_5373c2e5 PS7, Line 598: handle_connecting(cli, res);

This split into a separate function you could do it in a previous separate preparation patch.

Ack

https://gerrit.osmocom.org/c/libosmo-netif/+/33193/comment/6c84d18b_b1b934e3 PS7, Line 620: if (msg && res <= 0) {

EAGAIN would trigger closing the socket?

This shouldn't happen since osmo_io only writes to the fd if it's writable.

https://gerrit.osmocom.org/c/libosmo-netif/+/33193/comment/bc4afef1_4581ada5 PS7, Line 881: cli->ofd.priv_nr = 0; /* XXX */

what about this XXX?

Done

https://gerrit.osmocom.org/c/libosmo-netif/+/33193/comment/b6c5350a_6c95c2c9 PS7, Line 1046: cli->iofd = osmo_iofd_setup(cli, -1, cli->name, OSMO_IO_FD_MODE_READ_WRITE, &osmo_stream_cli_ioops, cli);

shouldn't this -1 be "fd"?

The relevant fd is in osmo_iofd_register(). In the reconnect case when osmo_stream_cli_open is called cli->iofd will already be set and osmo_iofd_register() registers a new fd.

https://gerrit.osmocom.org/c/libosmo-netif/+/33193/comment/7a08bcd5_9d49e29b PS7, Line 1540: osmo_stream_srv_destroy(conn);

just wondering whether you may need to return -EBADF here?

Not in a void function :-)

osmo_stream_srv_destroy() calls osmo_iofd_free() which does the right thing™ It'll close the fd, defer free() until after the callback has returned and not call the other callback.

https://gerrit.osmocom.org/c/libosmo-netif/+/33193/comment/b6e69d6f_6b476914 PS7, Line 1714: conn->iofd_read_cb = read_cb;

See, these can perfectly be separate setter APIs like we do for osmo_stream_cli.

I never argued that they couldn't be