| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In my testing, this appears to fix #1168, an intermittent test
failure. The underlying cause of the failure appears to be
a surprise blocking return old async-io versions:
> The problem was in that random yielding inside
> AsyncRead/AsyncWrite (which is supposed to prevent task starvation
> when I/O never returns the WouldBlock error) doesn't play nice
> with the TLS protocol. It seems that TLS handshake fails if a
> write errors with WouldBlock.
-- https://github.com/smol-rs/async-io/issues/31
I think the problem here might actually be something more subtle
than _that_, since WouldBlock on an underlying write is totally
legit. But this appears to fix the testing issue at least.
|
