Darren J Moffat wrote:
> Riny Qian wrote:
> 
>>
>>
>> Darren J Moffat wrote:
>>
>>> On the project page there is the following bullet:
>>>
>>> "* /dev/console serves as the current active console, that is, any 
>>> output via /dev/console will go to current active console."
>>>
>>> I don't think that is necessarily the correct behaviour.  It also 
>>> isn't the behaviour (best I can tell) that Linux has for this 
>>> functionality.
>>>
>>
>> Here we have an assumption that it is on the "workstation console". In
>> such case, /dev/console serves as the current active virtual console.
>>
>> We won't change the basic rule that "the file /dev/console refers to the
>> system  console  device". For example, in tipline case, /dev/console
>> will go to serial console.
> 
> 
> 
> I don't think I like the implication of this.
> 
> Lets consider this case.
> 
> Some part of the system starts writting messages to /dev/console, while 
> it is doing that I switch to another virtual console (lets call them VTs 
> for now since that is what people are famililary with).  What happens to 
> the output ?  If I understand what you are saying it starts getting 
> written on the VT I've just switch to.  This is not very nice in my 
> opinion and it also is not how it works on Linux.
> 
> Best I can tell how it works on Linux is /dev/console is the first real 
> console and it stays there.  The other VTs are /dev/tty1 /dev/tty2 etc 
> (or some naming like that I don't remember the exact syntax).  That 
> means that console messages always go to the same place.

I just took a quick look at Linux. On Linux, /dev/console is NOT a real
console, and all messages output via /dev/console does go to the current
active VT, it's not the same place.

> 
> This is the behaviour I want and I think it is what people would expect 
> based on their experiences with other systems that have VT functionality.
> 

There are some differences among other systems (e.g. our older Solaris
and Linux, we'll take a look at other systems).

Considering your abovementioned special case, now I prefer the policy of
our older Solaris, that is, /dev/console is a primary real console, and
VTs are just additional console terminals for users. Thus console
messages always go to the same place, and kernel messages also can
always go to the primay console (or some fixed virtual console).

I'd like to hear more people's comments on this policy/behaviour
before final design.

thanks,
Riny




Reply via email to