On Wednesday 04 April 2012 11:27:44 Joe Hershberger wrote:
On Tue, Apr 3, 2012 at 3:54 PM, Sebastian Andrzej Siewior wrote:
On 04/03/2012 10:42 PM, Mike Frysinger wrote:
On Friday 23 March 2012 16:11:19 Sebastian Andrzej Siewior wrote:
--- a/drivers/net/tsec.c +++ b/drivers/net/tsec.c
if (strcmp(getenv("stdout"), "nc"))i really don't like special casing devices like this
In this case I think it is better to check if stdout is nc, but not to explicitly write to serial_printf(). The nc device is the reason to avoid printing this since it uses the network. The serial_printf seems like the special case to avoid. Consider the case of using a SPI UART. There is no reason these traces should not go to it via the normal printf routing.
my point is that this doesn't scale ... not even close. either drop the printf calls in the core net case (when it'd be a problem with the netconsole), or figure out a solution that does actually scale. maybe a net_printf(). -mike