On 2017-08-11 20:56:22 -0700, Noah Misch wrote:
> > > If nobody volunteers, you could always resolve this by reverting 1e8a850 and
> > > successors.
> >
> > I think you're blaming the victim. Our current theory about the cause
> > of this is that on Windows, WaitLatchOrSocket cannot be used to wait for
> > completion of a nonblocking connect() call. That seems pretty broken
> > independently of whether libpqwalreceiver needs the capability.
>
> Yes, the theorized defect lies in APIs commit 1e8a850 used, not in the commit
> itself. Nonetheless, commit 1e8a850 promoted the defect from one reachable
> only by writing C code to one reachable by merely configuring replication on
> Windows according to the documentation. For that, its committer owns this
> open item. Besides the one approach I mentioned, there exist several other
> fine ways to implement said ownership.
FWIW, I'm personally quite demotivated by this style of handling
issues. You're essentially saying that any code change, even if it just
increases exposure of a preexisting bug, needs to be handled by the
committer of the exposing change. And even if that bug is on a platform
the committer doesn't have. And all that despite the issue getting
attention.
I'm not ok with this.