Function cmd_getc in command.c contains this code (2.7.3):
retval = select(num_fds, &readfds, NULL, NULL,
(quick_timeout ? &value : NULL));
/* See if we can read from the application */
if (FD_ISSET(cmd_fd, &readfds)) {
int count, n;
cmdbuf_ptr = cmdbuf_endp = cmdbuf_base;
for (count = BUFSIZ; count; count -= n, cmdbuf_endp += n)
if ((n = read(cmd_fd, cmdbuf_endp, count)) <= 0)
break;
if (count != BUFSIZ) /* some characters read in */
return (*cmdbuf_ptr++);
}
/* select statement timed out - we're not hard and fast scrolling */
if (retval == 0) {
refresh_count = 0;
refresh_limit = 1;
}
I'm definitely NO expert on the select function, but going by
'info libc 'Waiting for I/O'' there is a more efficient code.
if (select(num_fds, &readfds, NULL, NULL,
(quick_timeout ? &value : NULL)) == 0) {
/* select statement timed out - we're not hard and fast scrolling */
refresh_count = 0;
refresh_limit = 1;
}
/* See if we can read from the application */
else if (FD_ISSET(cmd_fd, &readfds)) {
int count, n;
cmdbuf_ptr = cmdbuf_endp = cmdbuf_base;
for (count = BUFSIZ; count; count -= n, cmdbuf_endp += n)
if ((n = read(cmd_fd, cmdbuf_endp, count)) <= 0)
break;
if (count != BUFSIZ) /* some characters read in */
return (*cmdbuf_ptr++);
}
libc.info states:
"If select returns because the timeout period expires,
it returns a value of zero."
So if select returns zero, there is no point in testing
FD_ISSET(cmd_fd, &readfds).
Since select will return zero most of the time (it only
returns non-zero in case of "input"), skipping the
if FD_ISSET(cmd_fd, &readfds)
part in case select returns zero makes sense.
The code proposed also does away with the 'retval' variable.
Now I probably miss out on some tricky detail, in which case
I would be happy to hear (perhaps from the person who coded
the cmd_getc function) why the retval value is not used
straightaway.
I have another question pertaining to cmd_getc, which I
will post in a separate message shortly.
Cheers,
Tijs
