Re: [PATCH] x86_64: fix delayed signals

[Linus Torvalds <torvalds@linux-foundation.org>]

On Thu, 10 Jul 2008, Roland McGrath wrote:
> 
> The "actual" downsides include numerous unknowns, and I always forget
> not to be surprised when you aren't scared that we have no idea what-all
> the code might actually do.

You're just grand-standing here.

The fact is, this has been "broken" for a long long time. A lot of 
applications (realistically, probably _all_ current 64-bit apps) have been 
tested with the old behaviour.

The thing is, I'm simply not the _least_ nervous about behavior that we've 
obviously had for a long time. And you shouldn't be either - nor should 
you try to make up some bogus argument to the contrary.

You should be scared about the case that IS NOT tested - which is by 
definition the new behaviour (admittedly only for 64-bit apps). Claiming 
anything else is just stupid and dishonest.

So stop making up _bad_ arguments. You do have a good one: the fact that 
x86-32 apparently works the way it works. But then you just irritate me 
with your idiotic other ones about how things are "supposed" to work.

So just be honest: you didn't have any _actual_ downsides that you knew 
about. Which was what I wondered about and asked about.

Those things make a big difference when I'm just about ready to cut a new 
release. This is _clearly_ not a regression (unless you talk about things 
from years and years ago), and apparently you don't have any actual code 
that is broken - just your expectations.

In short, it doesn't smell like something I should apply under -rc9 rules.

[ And dammit, I wish more developers would ask themselves that question: 
  "Does this patch need to go in when we're very late in the -rc 
  sequence?". Or at least not then try to grand-stand and avoid the 
  question I asked. ]

		Linus