Re: manipulating sigmask from filesystems and drivers

[Roman Zippel]

Hi

On Fri, 2 Aug 2002, Linus Torvalds wrote:

> Binary compatibility is important. As is the larger issue of generic UNIX
> compatibility. You had better have some really strong arguments for why
> you would think I'd be willing to break compatibility. So far you have had
> _no_ arguments for the question "Why?".

I never asked for breaking binary compatibility.
On the other hand I can give you an example, why I'd like to have a
choice. I expect from a good application that it recovers gracefully from
failures, so if I'm saving a file and the server goes down, I would
really, really like it if something happened when I push that stop button
or I press Esc and the application should offer me the possibility to save
the file somewhere else.

To implement this I would suggest using file flags instead of new task
flags:

	O_ATOMIC
	O_NONBLOCK
	O_SIGNALINT
	O_KILLINT
	O_DONT_BOTHER_ME

The first one might be useful for aio, it wants something like that
already anyway.
This way applications had a choice, how read/write should behave on
signals.

bye, Roman