[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Could we fix rreebboooott?

tb@MIT.EDU (Thomas Bushnell, BSG) writes:

> Do `showtrans /tmp/console' in that state and you will see one of the
> arguments is the filename /tmp/console.

Well, "ps aux" already showed that.

> Basically, /dev/console is being set up right, but is being
> tickled to turn on in some way, and that's causing the problem.

But /dev/console is the standard location.  Why shouldn't
programs be allowed to access it?  The problem is that another
translator starts reading the console device although there is
already one.  The /dev/console accesses should be directed to
the existing translator, so that they don't start a new one.
This is why I suggested making native-install move the
/tmp/console node to /dev: to get /dev/console associated
with the translator.  Then there is still the problem that
some program may have called ttyname() and will try to reopen
/tmp/console.  Making /tmp/console a symlink to /dev/console
would take care of that.

This scheme may not work if /tmp and /dev are different file
systems, since the translator might not survive the move.
It would be better if native-install could create /dev/console
and set its passive translator like it does now, and then
somehow attach the active translator to the node with something
like "settrans --active --copy=/tmp/console /dev/console".
Is that possible?

Reply to: