Termios: Non-blocking read() working? .-

[Sebastian E. Garcia]

On 06/11/2012 05:55 AM, Joris van Rantwijk wrote:
> On Monday 11 June 2012 10:09 Sebastian E. Garcia wrote:
>> 1. Gaisler's UART ISR and rtems_termios_enqueue_raw_characters function
>> seems OK at first glance.
>
> That stuff is inside "#if CONSOLE_USE_INTERRUPTS".
> Are you sure it is enabled?

(This applies to Gaisler's port). The last RCC compiler enables the use 
of APBUART individual configuration of interrupt or polling mode without 
needing to rebuild the kernel, configuring this struct in the 
application code:

/* APBUART1 */
struct drvmgr_key grlib_drv_res_apbuart1[] =
{
  {"mode", KEY_TYPE_INT, {(unsigned int)1}}, /* IRQ mode */
  {"dbgcon", KEY_TYPE_INT, {(unsigned int)0}}, /* debug out disabled */
  {"syscon", KEY_TYPE_INT, {(unsigned int)0}}, /* sys console disabled */
  KEY_EMPTY
};


>> Without having so much experience with Termios, the behavior I expect
>> when VTIME==VMIN==0, is to read at once (non-blocking):
>>    Min{No. of bytes requested by read(); No. of bytes in the raw buffer}
>> , and return ASAP.
>
> That is also what I would expect.
>
>> But in this code, when VMIN==0, only one char from the raw buffer gets
>> copied.
>
> I don't see that.
> The line marked "SG" causes the outer loop to exit ASAP, i.e. no waiting for more data to come in.
 > In the mean time, the inner loop can still collect multiple bytes if 
 > they are available in the queue. Right?
 >
 > Joris.


You are right, I've overlooked that. It was the perfect bug for need :)

Thanks again for your reply. Regarding leonuart, it seems that many data 
objects have changed since then (not only their name but structs fields, 
etc.).

Best regards,
Sebastian.