>Number:         16021
>Category:       port-amiga
>Synopsis:       installing via miniroot cannot extract sets, claiming crc errors
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    port-amiga-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Mar 22 18:27:01 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     Klaus Heinz
>Release:        NetBSD 1.5ZA
>Organization:
	none
>Environment:
	Amiga 3000, Cyberstorm MK2/060, CV64, A2065, 2x IBM DORS 2GB (only
	one used), 80 MB RAM

	System: NetBSD 1.5.2 NetBSD 1.5.2 (GENERIC) #0: Sat Aug 18 20:43:49
	MDT 2001 mhitch@amiga.msu.montana.edu:/work/src/sys/arch/amiga/compile/
	GENERIC amiga


>Description:
	I had installed the last snapshot 20020222-1.5ZA_crossbuilt about
	2 weeks ago, but not via the miniroot approach but manually from
	my running 1.5.2 system. This worked and I could boot 1.5ZA.

	Today I tried reinstalling with the miniroot and it didn't work.
        I had prepared the necessary partitions (NBR, NBU and the same swap
	I use for 1.5.2) and installed the miniroot into a small unused
	partition (not into my swap partition for 1.5.2 :-).

	I booted via the boot menu and the install process began.
	When it came to extracting the sets, I choose 'local disk' and let
	the installer mount the 1.5.2 partition where I had stored the sets.
	Then the trouble started (typed from the screen): 
	  Extracting the base.tgz set
	  pax: Invalid header, starting valid header search
	  pax: Invalid header, starting valid header search
	  pax: Invalid header, starting valid header search
	  pax: Invalid header, starting valid header search

	  gzip: stdin: invalid compressed data--crc error
	  Extraction complete
	  Extracting the comp.tgz set
	  pax: Invalid header, starting valid header search
	  pax: Invalid header, starting valid header search
	  ... [more of the same]
	  pax: unable to link onefilename from anotherfilename
	  ... [many more of the same]
	  
	etc.tgz, misc.tgz, text.tgz worked, man.tgz and games.tgz showed the
	same problems.

	I verified the sets under 1.5.2 after that and gzip didn't see
	any problem with them. I repeated the install of the miniroot
	and aborted the install program. Mounting the partition and
	inspecting the sets with gzip showed the same kind of problems but
	not with the same sets. etc.tgz, games.tgz and text.tgz was tested ok,
	but base.tgz, comp.tgz, man.tgz and misc.tgz showed crc errors or
	'format violated'.

	Again starting 1.5.2 and testing the sets showed no errors. I even
	tested them (via NFS) on an i386 box with 1.5.2, no errors.

	I also copied the statically linked a.out /usr/bin/gzip from my 1.5.2
	installation to /tmp (mfs filesystem) and prepended /tmp to PATH.
	This produced some strange results:

	  # for file in *.tgz; do echo $file; gzip -t $file; done
	  base.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")
	  comp.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")
	  etc.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")
	  games.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")
	  kern-GENERIC.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")
	  man.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")
	  misc.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")
	  text.tgz
	  מ-צנא: not found
	  gzip: 4: Syntax error: word unexpected (expecting ")")

	When I put the a.out gzip in /sbin it reverted to 'normal'
	behaviour, i.e. only crc errors and format violation.

	I think, this shows, that it is _not_ a problem with the 'instbin'
	program created by crunchgen.

	Unmounting and remounting the partition and then reversing the
	testing order (text.tgz first, base.tgz last) showed crc errors
	in text.tgz, misc.tgz, man.tgz and base.tgz. The others were ok.
	This was different than shown above.
	
	The sets where errors occur change sometimes _after_ unmounts/remounts
	but stay the _same_ in several tests _without_ unmounting/remounting.
	During the whole testing process with gzip the HD-LED was hardly ever
	blinking. 
	Could this point to a UVM/UBC problem ?
	
>How-To-Repeat:
	- prepare the partitions (NBR, NBS, NBU)
	- install the miniroot
	- start this partition and try to install the snapshot
>Fix:
	none :-/	
>Release-Note:
>Audit-Trail:
>Unformatted: