Some FC 2 Test / rawhide results

[Mark Heslep]

I pulled  the latest from yesterdays rawhide FC development tree  and 
ran an install of  FC 2 Test  (early?).  Created a syslinux USB stick 
based on the boot.iso image to use as a boot medium   Install method was 
NFS from a tree w/ kickstart to JFS partitions.


1. 1/2 dozen gnome(ish) package installs ( gdm, evolution, file-roller, 
etc ) report a "...parser error: Input is not proper UTF-8...."  to the  
anaconda /root/install.log but install ok otherwise.

2.  Bus device name attribuition caused headaches with conflicts betwenn 
the  usb storage device and internal SCSI adapter.  When the usb stick 
is the boot device  it takes /dev/sda away from the internal scsi 
adapter / disks. [1] The install can be made to go ok but on reboot from 
the internal drive is atrributed sda again causing  havoc.  Passing  
"nousbstorage" to anaconda  clears the way of course for the internal 
SCSI but  makes the usb stick unusable for any further use during the 
installation.  I thought filesystem  LABELs  were invented to avoid this 
kind of problem but they dont help either:   the kernel  can't find 
'init' on root  and panics. Had to fall back to  rescue mode and mangle 
/etc/fstab back to  /dev/sda.

3. Graphical install fails on  the attempt to start the X server on this 
ATI FireGL X1 card connected to a DVI LCD and a VGA LCD.  The failure 
locks everything up completely.  This appears to be the same radeon 
problem reported on FC 1 installs.

4. The often reported Gnome process-hang-after-log-outs continue to 
leave several processes running after X goes down, usually nautilus and 
bonobo.

5.  Unsupported category:  jfs  based installation crashes Anaconda due 
to a bad path to jfs_tune.  bug# 115101.  Kick start  %pre sym link 
provies a work around for now.


-Mark

[1]  I  vaguely recall that a post to lkml on scsi name attribution said 
a recent 2.6 patch solves this problem.