Re: [PATCH 1/2] condition_variable: Report early wakeup of wait_until as no_timeout

[Jonathan Wakely <jwakely at redhat dot com>]

On 24/07/18 21:44 +0200, François Dumont wrote:
> On 24/07/2018 13:57, Jonathan Wakely wrote:
> > On 24/07/18 12:04 +0200, François Dumont wrote:
> > > This patch claims that it just replaces a cv_status::timeout into 
> > > cv_status::no_timeout if we didn't wait enough. But it also 
> > > replaces any cv_status returned by __wait_until_impl into a 
> > > cv_status::no_timeout, is that intentional ? IMHO it should store 
> > > the returned status and return it unchanged if not timeout.
> >
> > But __wait_until_impl can only return timeout or no_timeout:
> >
> > enum class cv_status { no_timeout, timeout };
> I didn't want to check __wait_until_impl cause my remark was more general.
>
> But if enum only has 2 entries it sounds safe enough indeed.

It's a scoped enumeration, so all the values of its underlying type
are valid values of the enumeration type i.e. somebody could do
cv_status{413} and it would be valid. But __wait_until_impl doesn't do
that, and never will, and wait_until is required to only return
timeout or no_timeout, so Mike's patch would still be correct anyway
even if __wait_until_impl did something dumb :-)