On Monday 02 September 2002 3:28 pm, John Levon wrote:
> On Mon, Sep 02, 2002 at 02:55:59PM +0100, Angus Leeming wrote:
> > if (rerrno == EAGAIN) {
> > - errno = 0;
> > - return;
> > - }
> > - if (rerrno != 0) {
> > + break;
> > +
> > + } else if (rerrno != 0) {
> > lyxerr << "LyXComm: " << strerror(rerrno) <<
> > endl; if (!lsbuf.empty()) {
> > lyxerr << "LyXComm: truncated command: "
> > << lsbuf << endl;
> > lsbuf.erase();
> > }
> > - break; // reset connection
> > +
> > + // reset connection
> > + closeConnection();
> > + openConnection();
> > + break;
> > }
>
> Hmm, surely EAGAIN is a perfectly reasonable return value, not an error
> worthy of a printf
My point is that it's ridiculous to
closeConnection();
openConnection();
in this case.
I don't give a monkey about the print statement at the moment, although this
(below) is semantically equivalent to the current code and far cleaner.
However, I prefer a softly, softly approach to code that others may be more
familiar with than me.
Regards,
Angus
void LyXComm::read_alert()
{
read_buffer_.erase();
int const CMDBUFLEN = 100;
char charbuf[CMDBUFLEN];
errno = 0;
int status;
// the single = is intended here.
while ((status = ::read(infd, charbuf, CMDBUFLEN - 1))) {
if (status > 0) {
charbuf[status] = '\0'; // turn it into a c string
read_buffer_ += rtrim(charbuf, "\r");
} else if (errno != EAGAIN) {
if (!read_buffer_.empty()) {
cerr << "Comm: truncated command: "
<< read_buffer_ << endl;
read_buffer_.erase();
}
// reset connection
closeConnection();
openConnection();
break;
} else {
// errno == EAGAIN
// Nothing new has arrived, so now's the time
// to tell the outside world if there's anything
// in the read buffer.
break;
}
}
if (!read_buffer_.empty()) {
cerr << "Comm: Received from fd "
<< infd << '\n'
<< '\"' << read_buffer_ << '\"' << endl;
// emit signal
message_received();
}
errno = 0;
}
