
	SHIFT BUG LIST  :       @(#)BUGLIST	1.6 09/23/92

BUG ids are allocated as follows

	<CATEGORY><SHIFT VERSION><PACKAGE><ID>

where
	<CATEGORY> is   A for SHIFT software
			B for third-party network problem (e.g ULTRA)
			C for OS problem (e.g. SGI)
			D for Hardware problem

	<SHIFT VERSION> is      100 for first SHIFT versions
				xxx starting from 1.2.1

	<PACKAGEID> is D for DPM
			R for RFIO
			C for RTCOPY
			T for TAPE
			O for other
			M for msg
			L for library
			S for TMS/SYSREQ
			I for installation
			Z for documentation
	<ID> is a sequential id in the <PACKAGE>


BUG reports are in the form :

	BUGNO:  CxxxPnnnn       OPEN|CLOSED
	DATE opened : DD-MON-YY
	DATE closed : DD-MON-YY
	STATUS      : CRITICAL, URGENT, NUISANCE
	RESP        : who handles the problem
	REPORTED BY : Site, person, email, phone, fax
	DESCRIPTION : SHORT description

	Full description given by reporter

	Description by analyst

	Description of the cure by responsible

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
BUGNO:  A100D0001       CLOSED
DATE opened     : 05-JUN-1992
DATE closed     : 26-JUN-1992
STATUS          : URGENT
RESP            : Fabrizio Cane
REPORTED BY     : DESY, Georg Tysarczyk, F01TYS@DHHDESY3.BITNET
DESCRIPTION     : DPM lexicals cannot start with a digit

Description     :

Dear Dr. Robertson,
we are interested to install the SHIFT at the ZEUS
DEC-cluster for DPM usage.
I am writing to you, hoping that you could help us a little.
I am setting up the configuration files and I have noticed
that according to the manual (from August 8 1991) the
filesystem name can begin with any combination of letters and
digits.
In our case the disk have the names 06a etc.
In the shift.pools file a configuration like

SERVER  zow06 /shift/zow06
  /06a
  are not accepted it requires a letter first like
  /d06a

Could you tell me where I should change things in your code
so that such configuration is accepted ?
greetings + thank you
georg tysarczyk

Analysis        :

DPM lexical analysis does not allow lexicals to start with a digit
Already reported to FC in the context of sfmake and "8MM" tapes

Cure            :

Lexicals now can start with a digit and contain _,-

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
BUGNO:  A129T0001       CLOSED in 1.2.11
DATE opened : 13-JUL-92
DATE closed : 24-Jul-92
STATUS      : CRITICAL
RESP        : JPB
REPORTED BY : CERN, BALLINTIJN Maarten, NIKHEF-K/NA37 6524,
	      maartenb@cernvm.cern.ch
DESCRIPTION : SECURITY problems with TAPENV

The TAPENV variable allows a user to specify a file in an area where
he is not allowed to write (ex: /etc/passwd). This implies security
problems.

Analysis :

True. Not only TAPENV but also -P. Also for rls -a (remove the PATH).
At the time these are performed, the progs are already in suid to
talk to the tape daemon. Only suid's progs are allowed to talk to
tpdaemon (otherwise anybody could srew it up).

Cure:

Check write permissions.

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
BUGNO:  A129M0001       CLOSED
DATE opened : 15-JUL-92
DATE closed : 20-Jul-92
STATUS      : URGENT
RESP        : FH
REPORTED BY : CERN, ISNARD Christian. CERN/CN 2394, isnard@dxcoms.cern.ch
DESCRIPTION : MSGDAEMON generates zombies on SUN 4/490 SunOS 4.1.1 Rev B

Bonjour Frederic,

	Comme tu le sais nous avons demarre le msgdaemon sur notre machine
de development (cheops-dev). Nous avons constate qu'il produit des "defunct":

	csh> sps
	root      5406 /usr/local/bin/msgdaemon
	 |        5483 ** Exit **
	 |        5739 ** Exit **


Analysis :

The msgdaemon does not catch SIGCHLD and wait for child to finish. This
applies to all 4.2 fork() machines. We should trap SIGCHLD and perform
a wait3(). Only in msgdaemon.c and only for handling a (long) reply.
This bug appears on all SMI 4.1.x systems.

Cure:

Applied previous solution.

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
BUGNO:  A1212M0002       CLOSED in 1.2.14
DATE opened : 28-Aug-92
DATE closed : 02-Sep-92
STATUS      : URGENT
RESP        : FH
REPORTED BY : CERN, BAUD Jean-Philippe. CERN/CN 4981, baud@cernvm.cern.ch
DESCRIPTION : MSGDAEMON reports bad file number

Depuis le 22 juillet, aucun REPLY par les operateurs ne fonctionne.
le msgdaemon indique dans son log
netread(): Bad file number
As-tu une idee?
Qu'as-tu modifie?
				Jean-Philippe

Analysis :

The msgdaemon accept loop was not reaccepting on EINTR, causing further
communication to go on on a bad socket. All platforms.

Cure :

Correct the loop.

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
BUGNO:  C1214C0001       CLOSED in SunOS 4.1.2
DATE opened : 23-Sep-92
DATE closed :
STATUS      : CRITICAL
RESP        : FH
REPORTED BY : CERN, ISNARD Christian. CERN/CN 2384, isnard@dxcoms.cern.ch
DESCRIPTION : (Remote) tape read fails with EIO on SUN 4/490

Remote tape staging fails to reread a file written on a EXABYTE
with -B 32768 -L 32768

Analysis

The si device driver refuses to read on non-aligned buffers. However
SHIFT software is buffering 4 records at a time to increase network I/O
speed. This bug currently only shows up on 4/4xx and VME Sun3 SCSI adapter
on SUNOS 4.1.1. It does not seem to happen in 4.1.2 (at least on 4/6xx
series). The following message is displayed on the console :

si0:  illegal odd byte DMA address= 0x476

Cure

A compilation flag will be provided to get around buffering
