On Sat, 14 Mar 1998, Charles M. Hannum wrote:

> So, as part of the ongoing saga of improving the tty interface, I'm
> about to make the following change:
> 
> Add a new field to struct tty -- ts_wopen.  This will be used
> whereever the TS_WOPEN bit is currently used, but will instead keep a
> *count* of the processes waiting to open the device.  This can be used
> when an open fails to determine whether any other processes are
> waiting, and thus whether the device should be shut down.  (Currently
> there is no way to do this without device-specific hacks, which
> defeats the point of having TS_WOPEN in the first place.  In fact, we
> mostly get this wrong.)

Sounds very useful. :-)

> Although it would be easy to be backward-compatible here, on the
> general theory that cruft tends to never go away, I will *not* do so.
> This will simply make drivers that haven't been updated fail to
> compile, which isn't particularly harmful.

I missed something here. How will adding a field to struct tty make
non-updated drivers fail to compile?

Take care,

Bill