Re: 1.5.12 release announcement

["Bob Byrnes" <byrnes at curl dot com>]

On Oct 15,  9:35am, Christopher Faylor wrote:
-- Subject: Re: 1.5.12 release announcement
>
> | We should disable the NtQueryInformation stuff in peek_pipe for now
> | and treat the write side as always writable.  Then we have time up to
> | the next release to fix this in a more satisfactory way.
> 
> Bob Byrnes, do you agree with this assessment?  Or are you close to
> solving this problem?
> 
-- End of excerpt from Christopher Faylor

I feel like I'm making progress, but it's hard to quantify or provide
time estimates, because I'm playing around with mostly undocumented
native API calls.

Treating the write side of pipes as always writable is a regression
that trades one set of problems for another.  With the status quo,
programs like scp run slowly, but there are work-arounds (like running
with a smaller buffer size).  With the old "always writable" strategy,
rsync-over-ssh hangs will be more frequent, and I don't know of any
good work-arounds for that.

I'll keep pounding on a real solution, and I therefore don't have strong
feelings about how to cope in the short term, but I guess I'd recommend
keeping the NtQueryInformation stuff in place, since the rsync hangs
seem worse (at least for us).

--
Bob