bug in `tty: BKL pushdown'?

bug in `tty: BKL pushdown'?

[Johannes Weiner]

Hello!

git-bisect thinks that 04f378b198da233ca0aca341b113dc6579d46123 broke my
screen setup on the console in the following way:

I run ncurses applications like less and emacs in a screen session.

Since the above mentioned commit, a detach from screen followed by a
reattach (screen -r or by unlocking the screen session after lock),
these programs behave weird.  Emacs does display mode-line updates in
the next line each time which makes the screen scroll off.  This
behaviour is with vga=normal.

On vga=791, the breakage _toggles_ between detach-reattach cycles.
I.e. emacs breaks after the first detach-reattach and works properly
again after the second one.

I have yet to test a current kernel version with this patch reverted
because I am too sleepy right now to solve the conflicts.

If you need further information to debug this, please let me know.

	Hannes

Re: bug in `tty: BKL pushdown'?

[Alan Cox]

> I have yet to test a current kernel version with this patch reverted
> because I am too sleepy right now to solve the conflicts.

That wouldn't actually be a useful test if you did as the locking changes
would then bite you. The patch in question should just be a locking
change so short of compiler bugs ought to be fairly easy to audit. I will
dig into it next week.

Alan