Re: [evolution-patches] fix for "Error: Success" with imap

[Not Zed <notzed ximian com>]

I'd really like to know if there's some test case/data for this.  Unfortunately this patch is only a bandaid solution and is probably hiding the real cause of the problem.

If for example we are ot expect 0 reads on any stream then it should sit at a totally different level and not just in the imap code.

On Fri, 2004-10-15 at 09:09 +0800, Not Zed wrote:

Well is that a bogus zero-read?  Was there more data but for some strange reason we read 0?  What about all the select returns leading up to it?

This stuff really should be in a bug report with the traces attached :-/

On Thu, 2004-10-14 at 20:50 -0400, Jeffrey Stedfast wrote:

someone on IRC posted a subsection of an strace that I took a look at.

whatever IMAP server he was using, we sent STATUS <mailbox> (...)\r\n,
got a successful write, then tried to read and got:

read (36, "", 1024) == 0
fcntl64...
close (36) == 0

Jeff

On Thu, 2004-10-14 at 20:35, Not Zed wrote:
> Hmm, do you have some test data or something that demonstrates this?
> 
> I'm concerned that if we're getting 0 reads when we shouldn't be, that
> something else is going on, and this is a very small bandaid on the
> problem.
> 
> Is the problem easily repeatable?
> 
> This patch can probably go in for now with a TODO comment about why
> it's there.
> 
> On Thu, 2004-10-14 at 13:29 -0400, Jeffrey Stedfast wrote: 
> > Seems the problem happens when we get a 0-byte read from the socket.
> > 
> > the attached patch fixes the error (or, at least, one of the places
> > where error:success can happen)
> > 
> > Jeff
> > 
> 

--

Michael Zucchi <notzed ximian com> "born to die, live to work, it's all downhill from here" Novell's Evolution and Free Software Developer

--

Michael Zucchi <notzed ximian com> "born to die, live to work, it's all downhill from here" Novell's Evolution and Free Software Developer