On Tue, 2026-03-31 at 15:41 +0200, Nam Cao wrote:
> Gabriele Monaco <[email protected]> writes:
> 
> > On Tue, 2026-03-31 at 12:49 +0200, Nam Cao wrote:
> > > Since commit 0c43094f8cc9 ("eventpoll: Replace rwlock with spinlock"),
> > > epoll_wait is real-time-safe syscall for sleeping.
> > > 
> > > Add epoll_wait to the list of rt-safe sleeping APIs.
> > > 
> > > Signed-off-by: Nam Cao <[email protected]>
> > 
> > Thanks for the patch, looks reasonable.
> > I tried re-generating the header (sleep.h) with rvgen based on the new
> > specification and I'm getting a different order.
> > 
> > Is what you're committing the result of rvgen on your computer?
> > We probably still have some unpredictable result in the rvgen's output if
> > that's
> > the case (no big deal then, though it triggers me a bit).
> 
> Right, fixing this is in my list. The script uses set and set's order is
> not deterministic. You get different (but equivalent) results every
> time. I should start working on that..

Reasonable, no rush. I just noticed this behaviour when trying to write some
selftests for rvgen and didn't manage to make it deterministic with trivial
changes.

> 
> > I would still like to run some tests on this, how urgently would you like
> > this
> > patch through? I was really about to send Steve a PR with the other changes
> > so
> > this might need to wait for the next merge window.
> 
> The earlier the better, but no one will die because it misses a merge
> window.

Alright I'll see if I can squeeze it in within this week, if not, it'll have to
wait. For now:

Reviewed-by: Gabriele Monaco <[email protected]>

Thanks,
Gabriele


Reply via email to