Re: Re: [ECOS] Termios and waiting for character input

Re: Re: [ECOS] Termios and waiting for character input

[Andre-John Mas]

Where do I change the VMIN value? I'll change the value to see
if this helps.

BTW I had worked around the issue by simply doing a loop until
the input value was not 0x0. Not very elegant when CPU cycles
matter. I also made the same observation as Gary that the code
worked on other systems. I had tested the same code on MacOS X.

Andre

Jonathan Larmour wrote:
> 
> Gary Thomas wrote:
> > Jonathan Larmour wrote:
> >> Gary Thomas wrote:
> >>> Jonathan Larmour wrote:
> >>>> Non-canonical mode processing was only implemented to support 
> >>>> VMIN=0,TIME=0. Support for VMIN>0 is rudimentary and I wouldn't 
> >>>> trust it. But VMIN defaults to 0, and that sample code posted 
> >>>> doesn't attempt to set it to anything else.
> >>>>
> >>>> I imagine it defaults to something else on Linux, hence the 
> >>>> difference in behaviour. <fx: goes and checks> Aha, in fact it 
> >>>> defaults to 6. The POSIX standard does not set any requirements for 
> >>>> the default value, so eCos is fine on this count.
> >>>
> >>> Thanks for looking into this.
> >>>
> >>> Perhaps we should make the default values for VMIN/VMAX configurable,
> >>> possibly with settings that match other systems?
> >>
> >> Maybe so if I had more confidence it worked! If Andre-John (or 
> >> yourself) can play around with VMIN>0 and don't find any problems, 
> >> then we could change the default for VMIN. NB there is no VMAX, but if 
> >> you meant TIME, then TIME>0 is definitely unsupported in the current 
> >> code.
> > 
> > It does work correctly (or at least the same as Linux) when VMIN=1.
> > If I set VMIN=6, then I have to type 6 characters before the first one 
> > is seen.
> > Perhaps that's where a setting of TIME>0 would matter...
> 
> Yes indeed, that's the correct behaviour for TIME=0. So maybe it does work, 
> and we could set default VMIN to 1.
> 
> > At least we could have a way to default the value of VMIN and document
> > the fact that TIME>0 is not implemented (if not already in the docs)
> 
> Since VMIN can be set in a well-known and standard way at run-time, I'm not 
> sure there's much value in a configuration option for the default.
> 
> Termios is unfortunately undocumented. This was in the days when making doc 
> edits happen was like pulling teeth. And it became one of those things that 
> just never happened.
> 
> Jifl
> 


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Termios and waiting for character input

[Gary Thomas]

Andre-John Mas wrote:
> Where do I change the VMIN value? I'll change the value to see
> if this helps.
> 
> BTW I had worked around the issue by simply doing a loop until
> the input value was not 0x0. Not very elegant when CPU cycles
> matter. I also made the same observation as Gary that the code
> worked on other systems. I had tested the same code on MacOS X.

It's just another parameter to the termios structure.  Simply
add this line:
   term = termsave;            /* termsave contains old termios */
   term.c_lflag &= ~ECHO;      /* turn off echo if it's set */
   term.c_lflag &= ~ECHOE;     /* turn off echoe if it's set */
   term.c_lflag &= ~ICANON;    /* turn off echoe if it's set */
   term.c_cc[VMIN] = 1;

> Andre
> 
> Jonathan Larmour wrote:
>> Gary Thomas wrote:
>>> Jonathan Larmour wrote:
>>>> Gary Thomas wrote:
>>>>> Jonathan Larmour wrote:
>>>>>> Non-canonical mode processing was only implemented to support 
>>>>>> VMIN=0,TIME=0. Support for VMIN>0 is rudimentary and I wouldn't 
>>>>>> trust it. But VMIN defaults to 0, and that sample code posted 
>>>>>> doesn't attempt to set it to anything else.
>>>>>>
>>>>>> I imagine it defaults to something else on Linux, hence the 
>>>>>> difference in behaviour. <fx: goes and checks> Aha, in fact it 
>>>>>> defaults to 6. The POSIX standard does not set any requirements for 
>>>>>> the default value, so eCos is fine on this count.
>>>>> Thanks for looking into this.
>>>>>
>>>>> Perhaps we should make the default values for VMIN/VMAX configurable,
>>>>> possibly with settings that match other systems?
>>>> Maybe so if I had more confidence it worked! If Andre-John (or 
>>>> yourself) can play around with VMIN>0 and don't find any problems, 
>>>> then we could change the default for VMIN. NB there is no VMAX, but if 
>>>> you meant TIME, then TIME>0 is definitely unsupported in the current 
>>>> code.
>>> It does work correctly (or at least the same as Linux) when VMIN=1.
>>> If I set VMIN=6, then I have to type 6 characters before the first one 
>>> is seen.
>>> Perhaps that's where a setting of TIME>0 would matter...
>> Yes indeed, that's the correct behaviour for TIME=0. So maybe it does work, 
>> and we could set default VMIN to 1.
>>
>>> At least we could have a way to default the value of VMIN and document
>>> the fact that TIME>0 is not implemented (if not already in the docs)
>> Since VMIN can be set in a well-known and standard way at run-time, I'm not 
>> sure there's much value in a configuration option for the default.
>>
>> Termios is unfortunately undocumented. This was in the days when making doc 
>> edits happen was like pulling teeth. And it became one of those things that 
>> just never happened.
>>
>> Jifl
>>


-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss