On 8/27/19 12:34 PM, Max Reitz wrote:
> This does not really change anything, but it makes the code a bit easier
> to follow once we use @socket as the opaque pointer for
> aio_set_fd_handler().
>
> (Also, this change stops us from creating new CURLSocket objects when
> the cURL library just wants to stop listening on an existing socket that
> we do not recognize. With a well-behaving cURL, that should never
> happen anyway.)
>
> Cc: qemu-sta...@nongnu.org
> Signed-off-by: Max Reitz <mre...@redhat.com>
> ---
> block/curl.c | 7 ++++---
> 1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/block/curl.c b/block/curl.c
> index 92dc2f630e..8a45b371cc 100644
> --- a/block/curl.c
> +++ b/block/curl.c
> @@ -174,18 +174,16 @@ static int curl_sock_cb(CURL *curl, curl_socket_t fd,
> int action,
> if (socket->fd == fd) {
> if (action == CURL_POLL_REMOVE) {
> QLIST_REMOVE(socket, next);
> - g_free(socket);
> }
> break;
> }
> }
> - if (!socket) {
> + if (action != CURL_POLL_REMOVE && !socket) {
> socket = g_new0(CURLSocket, 1);
> socket->fd = fd;
> socket->state = state;
> QLIST_INSERT_HEAD(&state->sockets, socket, next);
> }
> - socket = NULL;
>
> trace_curl_sock_cb(action, (int)fd);
> switch (action) {
> @@ -207,6 +205,9 @@ static int curl_sock_cb(CURL *curl, curl_socket_t fd, int
> action,
> break;
> }
>
> + if (action == CURL_POLL_REMOVE) {
> + g_free(socket);
> + }
> return 0;
> }
>
>
Very naive question: why is CURL_POLL_REMOVE handled so early in the
function? Why not handle both QLIST_REMOVE and g_free under the
switch(action) construct entirely?
