On Mon, 16 Feb 2004, Grigory Entin wrote:

> > Each curl handle has its own cache and they are not shared unless
> > specificly made so.
>
> Yes, that is what I thought and was afraid of - it crashes for for FTP
> connection, and there I guess only one hostname.
>
> Just in case, below is what's relevant:
>
> * Re-using existing connection! (#0)
> >>
> #0 0x900abfe4 in getnameinfo ()
> #1 0x014fe158 in verboseconnect (conn=0x28b1c00, dns=0x16010a0) at
> #url.c:1844
> #2 0x014ff774 in SetupConnection (conn=0x28b1c00, hostaddr=0x0) at
> #url.c:3155

Eeek. This is badness. What I don't understand in this sequence is that
'hostaddr' comes in to SetupConnection() as a NULL, but is later passed on to
verboseconnect() as a pointer... Can you figure out why this happens? AFAICS,
it should be passed on untouch as a NULL.

> If remember it correctly, it can be getnameinfo (where it crashed right now
> in the above case), or (from my memory) infof call that follows getnameinfo
> call, both crashes due to messed up data in the area pointed by ai.

I believe this verboseconnect() function is a bit stupid in the ipv6-part when
we reuse connections (and the pointer is NULL as above). Can you see if the
attached patch makes life happier for your test-case? Thinking about it, it
will only do so if verboseconnect() get that NULL pointer passed in, not
otherwise.

-- 
    Daniel Stenberg -- http://curl.haxx.se/ -- http://daniel.haxx.se/
   [[ Do not send mails to this email address. They won't reach me. ]]