VMIN/VTIME and read expectations

[Joel Sherrill]

On 12/17/2009 06:05 PM, Eric Norum wrote:
> ???
> How did you make the TERMIOS stuff work on a socket?
> I use SO_SNDTIMEOand SO_RCVTIMEO options for sockets on RTEMS.  This causes read to return -1 with errno=EWOULDBLOCK on timeout.
I think this is the root of the problem.  The code behaves differently 
on the
head from the last release branch.  I need to see what has changed.
>
> We should probably take this correspondence off the newlib
>    
Yep.  Sorry folks.  Moving...

--joel
> On Dec 17, 2009, at 3:27 PM, Joel Sherrill wrote:
>
>    
>> On 12/17/2009 04:30 PM, Eric Norum wrote:
>>      
>>> When you're using VMIN=0, VTIME=255 you have ICANON=0 so 'end-of-file' doesn't really exist.
>>>
>>>        
>> I was kind of guessing that.  Unfortunately  this is in the RTEMS login code.
>> I can make it work OK for serial ports but when attached to a socket, I
>> have no idea how to detect when the socket is dropped.
>>
>> I switched to calling read() and as documented, it always returns 0
>> when there is no data.  When the connection is dropped, it returns 0
>> VERY quickly.  Do you think this part is a bug in the read on a socket?
>>
>> If this turns into an RTEMS specific issue, we need to switch lists.
>> But I would really like to know what is correct. :)
>>
>> --joel
>>      
>>> On Dec 17, 2009, at 2:16 PM, Joel Sherrill wrote:
>>>
>>>
>>>        
>>>> Hi,
>>>>
>>>> I have been looking into a program on RTEMS that
>>>> uses VMIN=0, VTIME=255. The program calls
>>>> fgetc() which calls read().
>>>>
>>>> fp->_flags == 0x2806 before the fgetc() call.
>>>> read() times out and returns 0.
>>>> fgetc() returns -1 and fp->_flags == 0x2826
>>>>
>>>>
>>>> The 0x0020 being set is _SEOF.
>>>>
>>>> My questions are:
>>>>
>>>> + Is this correct behaviour?
>>>>
>>>> + How does the application programmer distinguish
>>>> between a real EOF and an unsatisfied read?  Even the
>>>> Fedora read(2) man page says:
>>>>
>>>>        On success, the number of bytes read is returned (zero indicates end of
>>>>        file), and the file position is advanced by this number.
>>>>
>>>> So even if I dropped down to read(2), I still don't know
>>>> how to distinguish EOF from a read timeout.
>>>>
>>>> Thanks.
>>>>
>>>> -- 
>>>> Joel Sherrill, Ph.D.             Director of Research&    Development
>>>> joel.sherrill@OARcorp.com        On-Line Applications Research
>>>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>>>>    Support Available             (256) 722-9985
>>>>
>>>>
>>>>
>>>>          
>>>
>>>        
>>
>> -- 
>> Joel Sherrill, Ph.D.             Director of Research&   Development
>> joel.sherrill@OARcorp.com        On-Line Applications Research
>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>>    Support Available             (256) 722-9985
>>
>>
>>      
>