[syslinux] extlinux 6.03 serial console not responding to input

[Dany St-Amant]

> Le 2015-05-19 à 18:14, Ady via Syslinux <syslinux at zytor.com> a écrit :
> 
> 
>> Hi,
>> 
>>> On Tue, May 19, 2015 at 05:13:25PM -0400, Dany St-Amant wrote:
>>> Replying to the original e-mail as the thread was containing little extra useful information (beside the fact that the issue was similar as to when the flow control values were not provided at all).
>>> 
>>> Based a quick look at the code I had on hand 2.06 and 6.01. I would suggest that you use a decimal number instead of an hexadecimal one for the flow control.
>>> 
>>> The assembly parseint() of 2.06 supported decimal,octal,hexadecimal and some val+K, val+M. The new C parsing code uses plain atoi() which appears to be quite limited.
>>> 
>>> So both your 0x003 and 0x00b are handled as 0, causing you to go back to before you introduced these values.
>> 
>> Yes!  That's it -- the fix was as simple as replacing "0x003" with "3".
>> I guess the code should use strtol or sscanf instead of atoi.  Or the docs need
>> to be fixed.
>> 
>> Thank you!
>> 
>> Regards,
>> Andy
> 
> Excellent! After all, testing with alternative notations and with 
> different values indeed (would have) triggered a different behavior :).
> 
> This is not the first time we see something not working with the SERIAL 
> directive since v.5.xx. For instance, see:
> 
> _Start of the email thread ("syslinux 5.01 serial regression"):
> http://www.syslinux.org/archives/2013-May/020019.html 
> _GeneC:
> http://www.syslinux.org/archives/2013-May/020023.html 
> 
> Note the similarities to the current case.
> 
> This was the email thread that triggered the prior (latest) changes in 
> the source code I mentioned before.
> 
> Please note that one month later, during June 2013, there was an 
> additional report about the TIMEOUT not working correctly when also 
> using the SERIAL directive. It might be worth re-testing such case.
> 
> 
> Now, instead of expecting users to only use decimal values, could 
> someone with the adequate knowledge *please* consider sending a patch 
> (to this Syslinux mailing list), so older configurations would work as 
> expected?

How far such a backward compatibility fix should go?
Based on my limited Intel assembly skill, I suspect that the following was valid:

timeout --0xAk

To get a  +1024 seconds timeout (in a cryptic way)

> 
> TIA,
> Ady.

Dany