On Mon, Sep 28, 2020 at 10:31 AM Siddhesh Poyarekar via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> On 28/09/20 19:05, Florian Weimer via Libc-alpha wrote:
> > I'm not sure if this change is philosophically correct as far as the
> > tunables framework is concerned.  I had thought the limits should be
> > something static, so that they are consistent across systems.
>
> It seems like a good idea to support dynamic limits if they will always
> be more restrictive than the most restrictive static limit one could
> come up with for the tunable.  I didn't exclude dynamic limits from a
> design perspective; it's just that the tunables implemented at that time
> didn't need them.
>
> There is a case to always have static bounds (at the minimum to ensure
> that values don't overflow the underlying types) but that shouldn't
> preclude more restrictive dynamic limits IMO.
>
> Bikeshed: maybe the macro should be called TUNABLE_SET_WITH_BOUNDS()
> instead of TUNABLE_SET_ALL.
>

Here is the updated patch with TUNABLE_SET_WITH_BOUNDS.

OK for master?

Thanks.

-- 
H.J.