Le 18/09/2016 à 12:06, Etienne Champetier a écrit :
> (Answering from my phone with gmail so this email is in HTML and will
> get bounced by the ML)
> Le 18 sept. 2016 14:03, "John Crispin" <j...@phrozen.org
> <mailto:j...@phrozen.org>> a écrit :
>> On 18/09/2016 02:13, Florian Fainelli wrote:
>> > Le 17/09/2016 à 16:49, Etienne Champetier a écrit :
>> >> Hi Florian,
>> >> What is your use case?
>> > Pre-built toolchain which ships with kernel headers 3.8 at the moment.
>> >> If you run it on kernel less than 3.17 this will then fail at run
> time ...
>> > I run a kernel newer than 3.17, but that's a good point.
>> >> I would prefer to not compile it instead of having an unusable binary
>> > Ok, so something more like that:
> I still doesn't understand why getrandom isn't built with target kernel
> All LEDE targets are using linux 4.4 and OpenWRT targets are 3.18+, so
> this feels like the wrong fix for me
I am using an external toolchain which ships with kernel headers 3.8,
and anybody building an external toolchain in OpenWrt may run into a
Irrespective of the kernel version, since the system call is not present
in all kernel versions, testing for its presence does not hurt.
Besides that, people may end up backporting a newer user-space to run
with an older kernel that is pre 3.17, making things robust and testing
for what you use in your application is a good practice that has years
of precedence (autotols would not exist otherwise).
Lede-dev mailing list