On Wed, Oct 5, 2022 at 5:34 PM Kevin Wolf <kw...@redhat.com> wrote: > The coroutine_fn annotation for qemu_coroutine_self() is wrong, but I > think it already works outside of coroutine context, and consistently in > all three backends, by returning &leader. > > Changing that to NULL makes me kind of nervous because the callers might > actually access the leader Coroutine object, and after this change they > would crash. (And even if they didn't crash, they wouldn't be able to > distinguish the leader coroutines of different threads any more.) > > Do we have an actual reason to make this chance? That is, do we have any > case that was broken before?
No, I just wasn't sure if the current implementations would return a meaningful value when called from outside coroutine context, but it seems they do. I'll send a patch only dropping the coroutine_fn annotation. Thanks, Alberto
