Alex Rousskov wrote:
On Sun, 2008-11-02 at 13:46 +1300, Amos Jeffries wrote:
Robert Collins wrote:
On Sat, 2008-11-01 at 10:45 -0600, Alex Rousskov wrote:
Moreover, if we must read until EOF before we close, then the cleanup
code would have to be rewritten in many places, not just the deferred
read-related ones!
src/comm.cc line 353:
/**
* Empty the read buffers
*
* This is a magical routine that empties the read buffers.
* Under some platforms (Linux) if a buffer has data in it before
* you call close(), the socket will hang and take quite a while
* to timeout.
*/
static void
comm_empty_os_read_buffers(int fd)
{
#ifdef _SQUID_LINUX_
/* prevent those nasty RST packets */
char buf[SQUID_TCP_SO_RCVBUF];
if (fd_table[fd].flags.nonblocking == 1) {
while (FD_READ_METHOD(fd, buf, SQUID_TCP_SO_RCVBUF) > 0) {};
}
#endif
}
Aha, thanks Robert that was it exactly.
Two seconds more thought and this API seems to be what the destructor
should be doing instead of kicking off its own internal read.
But we are already doing the above for all sockets, including sockets
with deferred reads. If I am not misreading the code, it still looks
like we can just ignore deferred read handlers when their manager is
being destructed, right?
I haven't checked the whole chain. All that matters is that its called
once before socket close. If you have found a second spot, great some
code reduction :)
Amos
--
Please be using
Current Stable Squid 2.7.STABLE5 or 3.0.STABLE10
Current Beta Squid 3.1.0.1
