This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: rsync over ssh hang issue understood


On Sat, Jul 01, 2006 at 03:11:48PM +0100, Darryl Miles wrote:
>Christopher Faylor wrote:
>>I just want to be clear here.  I really do want to be convinced that
>>the current implementation in cygwin can't be fixed before we scrap
>>that and implement something new.
>
>Yes I intend to "convince" the list of my findings.
>
>I have simple cut down unix code that works on unix for not under
>cywin, over the next few days I should have a lid on the exact rsync
>and ssh interactions at the time of the lockup and be able to describe
>the problem better.

You are misunderstanding me.  There is code in cygwin which is designed
to work around the problem of select() not working correctly when it is
used to test the write side of pipes.  It is currently turned off
because it wasn't working perfectly but the person responsible for
fixing it seemed to be confident that the problem could be fixed.  Before
we throw out that code, I want to know that it can't be made to work since
it involved a substantial change to pipe handling.

We don't need proof that there is a problem.  No one doubts that there
are problems in this area (have you tried a snapshot, btw?  There have
been some changes in pipe handling in the snapshots).  How could we?
This topic shows up in the mailing list all of the time.

>When I announce that situation it will be for the original implementer 
>or current maintainer to then speak up an defend their design.  As I 
>should be able to point out the exact nature of the problem, where the 
>POSIX semantics got broke and maybe where misunderstandings in how WIN32 
>kernel behaves in some scenarios.

This is an open source project.  We don't need the "original implementer"
to defend anything.  I guess my intent isn't clear so let me try to state
it once again.

I don't want you to go to a lot of work implementing something new
around the guts of an almost working implementation which may just need
some extra work to fix.  I'm trying to avoid the scenario of having you
submit something to cygwin-patches which is guaranteed to be immediately
rejected, or at least delayed.

cgf

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]