pserver login problem (strange)

[Graeme . Vetterlein]

This is exactly the behaviour I got which lead to the above
debate about CR-LF sequences. We have mixed development on doze & Unix.
The same 'snadboxes' are visible on all platforms. The CVS repository
is locked away in nice safe datacentre and only accesed via pserver.

If a doze user does a checkout, cvs will create a file CVS/Root in
their 'sandbox' .. if that is visible (e.g. NFS, samba (RFS:-) ) then
then when you do say cvs commit ... it will use the $CVSROOT value found
in there and NOT the one in your environment. The doze created file contains
a (from a Unix standpoint) spurious CR (shown as ^M) so it tries to use
a repository called (e.g) "mycvs^M" since it can't find this it can't 
do the security check so it says "I HATE YOU" .. which it did to me for
around
eight solid hours in the wee small hours one night... gurrr. Check the
CVS/Root
and .cvspass files of the failing users (hex dump if necessary).

As an aside. I spent many years with IBM029 card punches and fixed format
records.
The ONLY reason we have the <CR><LF>  sequence is that the KSR33/ASR33 (33
char/sec
teletypes) took quite a time to return the carriage to the LHS of the paper.
If you send <LF><CR> the subsequent printing would take place while the
print
head was still moving Right to left. Nowdays there is realy no reason we
need
two characters to mark a record boundary ;-}

> Message: 10
> Date: Wed, 24 Jan 2001 13:19:39 +0000 (GMT)
> From: James Greenhalgh <address@hidden>
> To: address@hidden
> Subject: pserver login problem (strange)
> 
> Hi - maybe someone can help with this.
> 
> We have a CVS server, using the pserver inetd.conf setup 
> (with the correct
> allow_root set).
> 
> Some users are able to log in to cvs from any host (be it 
> linux, solaris,
> or windows), and a few are unable to log in from the solaris
> hosts.  Comparing a working user and a nonworking one, I 
> can't find any
> difference between them - but obviously there must be something.
> 
> Does anyone have any ideas on this?  Is there a way of getting cvs to
> actually log what it's doing - I'd love to know what it tries to
> authenticate against, and what values its trying to use.
> 
> thanks,
> james