Bug 214325

Summary:	SLES 10 autoinstall not working with serial console
Product:	[openSUSE] SUSE LINUX 10.0	Reporter:	Stefan Hinker <stefan.hinker>
Component:	YaST2	Assignee:	Katarina Machalkova <kmachalkova>
Status:	RESOLVED DUPLICATE	QA Contact:	Klaus Kämpf <kkaempf>
Severity:	Normal
Priority:	P5 - None	CC:	werner
Version:	unspecified	Keywords:	Install
Target Milestone:	---
Hardware:	x86-64
OS:	SLES 10
Whiteboard:
Found By:	Other	Services Priority:
Business Priority:		Blocker:	---
Marketing QA Status:	---	IT Deployment:	---
Attachments:	syslog output of the boot/autoinstall process console output of the install process linuxrc file used to boot via PXE autoyast config file

Description Stefan Hinker 2006-10-23 09:17:00 UTC

I have a *working* configuration of autoyast/autoinstall for SLES10.
This configuration breaks (stops with the "An error occurred during the
installation" message) if I include the option "console=ttyS0" in the
linuxrc parameters.

The last error message I see on syslog is:

yast2[3711]: <5> [ui-ncurses] YNCursesUI.cc(YNCursesUI):101 NCurses:
(-1) newterm() failed

This seems to be a bug in the yast2 ncurses frontend.

Specifying different options for term (vt100, xterm) is recognized by yast,
but does *not* change the behaviour.

Comment 1 Stefan Hinker 2006-10-23 09:18:45 UTC

Created attachment 102251 [details]
syslog output of the boot/autoinstall process

This is the output of syslog (via loghost=...) of the
install process until it stops with the error message.

Comment 2 Stefan Hinker 2006-10-23 09:19:58 UTC

Created attachment 102252 [details]
console output of the install process

This is the console output (as captured via xterm "log to file".

Comment 3 Stefan Hinker 2006-10-23 09:20:50 UTC

Created attachment 102253 [details]
linuxrc file used to boot via PXE

This is the PXE configuration to configure the booting
kernel for autoinstall.

Comment 4 Stefan Hinker 2006-10-23 09:21:38 UTC

Created attachment 102254 [details]
autoyast config file

This is the autoyast config file for the autoinstall.
It is never used, since the process stops before this.

Comment 6 Katarina Machalkova 2006-10-24 13:42:14 UTC

Stefan, first of all - is this SUSE Linux 10.0, or SLES10 ?

Secondly, I've seen a crash very much like this when the content of $TERM environment variable was something, that was actually not defined in /usr/share/terminfo. 

However, newterm() is a function that comes from ncurses library (adding our ncurses maintainer to cc.). I'm not really sure how all these (i.e. ncurses, terminfo, console, TERM variable,...) work together at the end and why newterm() segfaults with wrong TERM content. Werner, could you please comment ?

Comment 7 Olaf Hering 2006-10-24 14:23:30 UTC

similar to bug #192052

Comment 8 Olaf Hering 2006-10-31 15:17:23 UTC

are there any details about it? how can we reproduce it?

Comment 9 Stefan Hinker 2006-10-31 15:30:36 UTC

(In reply to comment #8)
> are there any details about it? how can we reproduce it?
> 
The details (logfiles, console output, config files etc.) are attached to the
bug description.  What else do you require ?  

To reproduce, take a suitable machine, set it up to install using serial console, 
and try it.

stefan

Comment 10 Dr. Werner Fink 2006-11-02 12:37:28 UTC

First of all, changing the value of the environment variable TERM does
never re-program the terminal of terminal emulator its self. Such a change
only switches from one terminal description to an other maybe wrong terminal
description. The function newterm(3) returns a NULL if a not existing TERM(inal)
type was choosen. It is on the caller to catch this error.

Comment 11 Katarina Machalkova 2006-11-03 09:02:44 UTC

OK, this really seems to be dup of bug #192052

*** This bug has been marked as a duplicate of bug 192052 ***