[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Heimdal 1.1 telnetd broken under Solaris 10

Doesn't the Solaris 10 telnetd do Kerberos?  That lowers the  
importance of Heimdal's considerably, I think.

On Mar 2, 2008, at 4:24 AM, Ian Delahorne wrote:

> On 25 jan 2008, at 14:21, Harald Barth wrote:
>> When tested under Solaris 10, telnetd in Heimdal 1.0.x produces
>> gibberish in the face of the user, probably some encrypted data.
>> My hope was that this was fixed in 1.1, but no, so Ian and I had
>> to fix it ourselves.
>> The problem is, that under Solaris 10, the configure test for
>> TIOCSCTTY produces "true", however the call to ioctl(t, TIOCSCTTY,
>> (char *)0) fails resulting in a call to fatalerror(). That is not
>> taking down the connection properly. Therefore the gibberish.
>> However, Solaris 10's terminal implementation is STREAMSPTY, which
>> means that ptem and ldterm are allready pushed on the stream and the
>> return value of the TIOCSCTTY ioctl can be ignored. This may even be
>> true for HPUX, which could remove the !defined(__hpux) from the  
>> #ifdef
>> hell, but I can't test that. Under Solaris 10, telnetd works with the
>> patch below. I did leave the call to TIOCSCTTY in because there _may_
>> be architectures that define STREAMSPTY and TIOCSCTTY where the call
>> to TIOCSCTTY does something usefull.
>> This still does not address the "not clean shutdown" issue when  
>> calling
>> fatalerror(), but gives a working telnetd.
> Is this going to be in the next Heimdal release? Seems that a  
> working telnetd on Solaris 10 should be in the main distribution  
> and not just floating around as a patch on a mailing list.
> /ian