[Prev][Next][Index][Thread]
Re: chown-patch
On Fri, 28 Jan 1994, Michael Schneider wrote:
> hi managers,
>
> i had problems with "dxuam" under ultrix4.3a on ds5000/150.
> dxuam wasn't able to create the users home directory or something
> like this.
> the problem was solved with a patch to the /usr/bin/chown routine.
>
>
> i think we should mail each other the patches we get from dec
> and the corresponding problems, because dec don't publish a
> list of available patches/bugs and they don't make patches available
> on gatekeeper.
>
Yeah it's a problem.
Anyway, here's a README for ultrix 4.3a patches and another
for uws4.3a, that I got my hands on. I hope you dont send it
around though, this could hurt my personal connections to dec...
Hope this helps, regards, Harald
PS. I don't know if there's a newer one, this one is from august...
--- clip ---
Digital Internal Use Only
V4.3a RISC Patches
August 26, 1993
/usr/lib/cmplrs/as3.0/as13.0 (CFS_2490, CXO.10139, VBO-01010)
CHECKSUM: 38401 392
----------------------------
Problem description:
Programs that contain over-long branch instructions (branch
instructions that branch over more than 32K instructions) no
longer produce sporadic errors in the linker.
Solution:
Since the R3000 architecture does not support branch
instructions that branch over more than 32K instructions, the
assembler translates such long branches into sequences of
absolute jumps. Since the targets of these jump instructions
are absolute addresses, the assembler must set up relocation
entries for these addresses. This change fixes a bug with these
relocation entries.
The problem was with the 'extern' bit in the relocation entry.
This bit indicates whether the relocation is relative to an
external symbol or a local symbol. Since these relocation
entries are always for local labels inside the file, they should
always have the 'extern' bit set off. Unfortunately, the code
forgot to set the bit either way. Therefore, the value of the
bit was inherited from the previous relocation entry. If the
previous entry was for an external symbol, this relocation
entry was generated incorrectly. We fix the problem by always
setting the 'extern' bit to '0' for relocation entries
associated with jump instructions that were generated from
translating an over-long branch instruction.
==============================================================================
/sys/MIPS/BINARY/cdfs_gnodeops.o (CLD IPO-6222)
CHECKSUM: 29388 71
/sys/MIPS/BINARY/cdfs_namei.o
CHECKSUM: 34314 55
--------------------------------
The system is unable to read a specific Hi-Sierra format CDROM
with ISO 9660 CDFS file system on ULTRIX. Using correct CDFS mount syntax,
cdrom mounts correctly with no errors reported. However, when
traversing cdrom, no files are visable. Also, the directory structure
on the cdrom is visible, but when an ls -l command is issued, an error
is returned saying no files.
==============================================================================
/sys/MIPS/BINARY/clnt_kudp.o (CFS.3517)
CHECKSUM: 21724 73
/sys/MIPS/BINARY/if.o
CHECKSUM: 03648 73
/sys/MIPS/BINARY/kudp_fastsend.o
CHECKSUM: 37890 68
/sys/MIPS/BINARY/subr_kudp.o
CHECKSUM: 14739 39
/sys/MIPS/BINARY/svc.o
CHECKSUM: 12337 33
/sys/MIPS/BINARY/svc_kudp.o
CHECKSUM: 12081 58
/sys/MIPS/BINARY/sys_socket.o
CHECKSUM: 28593 61
/sys/MIPS/BINARY/tcp_input.o
CHECKSUM: 33508 84
/sys/MIPS/BINARY/tcp_output.o
CHECKSUM: 27588 47
/sys/MIPS/BINARY/tcp_subr.o
CHECKSUM: 28215 53
/sys/MIPS/BINARY/uipc_socket.o
CHECKSUM: 07921 114
/sys/MIPS/BINARY/uipc_socket2.o
CHECKSUM: 49176 62
/sys/MIPS/BINARY/uipc_syscalls.o
CHECKSUM: 43184 74
---------------------------------
This patch fixes the following kernel panic whose text was "panic tbl_miss".
The following is a sample output from the stack.
0 boot(0x0, 0x801a2b9c, 0x6c, 0xd801, 0xc1812080) [0x8011373c]
1 panic(0x8011f3c4, 0x39, 0x6c, 0xc17b8240, 0xc) [0x800af66c]
2 tlbmiss(0x8, 0xc1895000, 0x8, 0xfffffffc, 0xc1895002) [0x8011d798]
3 VEC_tlbmiss(0x0, 0x1522, 0x8008d0fc, 0xc1894fec, 0xc17c6224) [0x80111bd0]
The problem appears to be in the macro SO_LOCK in socketvar.h. This macro
replaced an smp_lock call with some code that is only executed if smp is
set. The lock is needed for non-smp code as well, and so there is a small
window of time where a page table entry could be modified. This can be
exacerbated by heavy socket I/O to slow devices (nfs mounting a cdrom, for
example).
For socketvar.h in 4.3a, a new macro defined that was used in uipc_socket.
This macro, SO_LOCK_KEY, replaced a SO_LOCK call in uipc_socket.o. All the
other binaries were unaffected by this change.
A kernel rebuild is required.
==============================================================================
/usr/sys/MIPS/BINARY/kn02.o (CLD CXO8988)
CHECKSUM: 16384 168
/usr/sys/MIPS/BINARY/kn02ba.o
CHECKSUM: 32817 127
/usr/sys/MIPS/BINARY/kn02ca.o
CHECKSUM: 24616 126
/usr/sys/MIPS/BINARY/kn03.o
CHECKSUM: 56502 153
/usr/sys/MIPS/BINARY/kn220.o
CHECKSUM: 06045 145
/usr/sys/MIPS/BINARY/kn5800.o
CHECKSUM: 54759 150
/usr/sys/MIPS/BINARY/xmiinit.o
CHECKSUM: 62623 119
---------------------------------
Kernel error logging hangs. Cause by kernel memory modules logging non-severe
errors as severe which causes error logging to stop logging to the syserr
log file. Other kernel modules were found to not panic on severely logged
errors potentially causing the same symptom.
Solution
Modified the modules that logged non-severe errors as severe to log them as
high instead. Modified the modules that were ignoring severe errors to panic
after those errors are logged as severe. In both cases, a severe error log
should only occur just before a panic and non-severe errors will be logged in
the syserr log file for customer to view using the uerf command.
A Kernel rebuild is required
Module and CPU cross reference table
MIPS
****
DS5000 kn02.o
DS5000/100 kn02ba.o
DS5000/25 kn02ca.o
DS5000/240 and DS5900 kn03.o
DS5500 kn220.o
DS5800 kn5800.o and xmiinit.o
==============================================================================
/usr/lib/cmplrs/cc2.1/lint1 (VBO_956)
CHECKSUM: 47436 286
/usr/lib/cmplrs/cc2.1/lint2
CHECKSUM: 34081 87
---------------------------
Problem description:
RISC lint aborts on an unaligned access when processing a file which works
on VAX lint. Lint used to crash when run on just the following two lines
of c-code.
x("x");
y("y",a+b);
Solution:
Lint1 was changed to insure that the null parameters passed to pftn routines
that formerly caused unaligned access problems were not processed after they
were known to be null.
==============================================================================
/usr/etc/lockd
CHECKSUM: 48154 148
---------------------
Problem 1: (CXO-10497)
**********
The problem is that a NFS client may test for locks and not find one
when there actually was one.
Problem 2: (CXO-9171)
**********
Querying locks from a Sun client and VAX NFS server hangs the client.
Problem 3: (UVO-1577)
**********
After locking two files with two programs, ctrl-c'ing out of the first will
not release the lock.
Problem 4: (MUH-2524)
**********
Multiple users and multiple locking confuses lockd and hangs the application
until lockd is killed.
After the patch has been install reboot the system.
==============================================================================
/usr/etc/named (AKO-2207, CXO-08577, CFS.3110)
CHECKSUM: 40816 427
res_init.o
CHECKSUM: 36799 5
res_send.o
CHECKSUM: 44021 5
------------------------
Named daemon has memory leak problem, the process size can graduately increase
to be very large. The solution was to call free() in a few routines to free
memory which was allocated but will never be used.
Installation:
1. Login as root.
2. Copy named to /usr/etc/named, save the original named
before copying.
3. Bring down the current running named process.
4. Start the new named process.
5. Archive the fix res_init.o and res_send.o into
/lib/libckrb.a, make a copy of original libckrb.a.
Use 'ar -r libckrb.a file', file is either res_init.o
or res_send.o
==============================================================================
/usr/sys/MIPS/BINARY/route.o (IPO5704, IPO6160, CXO9961, CXO9891)
CHECKSUM: 56595 54
----------------------------
This fixes a problem where machines with SLIP connections sometime panic
with "kpteseg miss" "segmentation fault", "bad rmfree", or memory corruption.
A kernel rebuild is required.
==============================================================================
/etc/snmpd (CXO-8902)
checksum: 07070 466
/usr/sys/MIPS/BINARY/ip_input.o
checksum: 39695 54
-------------------------------
MIB variables were missing from the ULTRIX SNMP agent. Snmpd was not
reporting the values for the following fields:
ifOutDiscards
ipInUnknownProtos
ipInDiscards
ipInDelivers
ipReasmOKs
tcpPassiveOpens
Some values in the "interfaces.IfTable.ifEntry" group only show up for the
first interface, but do not show up for the second interface. The solution
was to implement the missing MIB variables.
Solution:
Implemented the missing MIB variables.
Kernel rebuild required for /usr/sys/MIPS/BINARY/ip_input.o
==============================================================================
/usr/lib/sprintf.o-libc (CLD CXO-10229)
CHECKSUM: 43941 2
/usr/lib/sprintf.o-libcP
CHECKSUM: 09313 2
/usr/lib/sprintf.o-libcV
CHECKSUM: 41986 2
/usr/lib/sprintf.o-libc_G0
CHECKSUM: 18665 2
/usr/lib/sprintf.o-libc_mips3
CHECKSUM: 23162 2
/usr/lib/sprintf.o-libc_mips3_G0
CHECKSUM: 18749 2
/usr/lib/sprintf.o-libckrb
CHECKSUM: 43941 2
--------------------------------
This patch resolves a problem with string formatting stopping after 32767
bytes when using sprintf to format strings that are longer than 32k.
The names of each of modules are named to reflect which library the
modules are to be archived into. Example, /usr/lib/sprintf.o-libcP_p
is to be archived in /usr/lib/libcP_p.a and /usr/lib/sprintf.o-libckrb
will be archived into /usr/lib/libckrb.a. Before the archive is done
remove the library extension from the sprintf.o module. What should
be archived into each of the libraries is sprintf.o
Note: /usr/lib/libc_mips3.a and /usr/lib/libc_mips3_G0.a are the libraries
for R4000 machines only.
Installation instructions (examples):
Extract the module out of /usr/lib/libc.a by using:
# /bin/ar -x /usr/lib/libc.a sprintf.o
ave the old one:
# /bin/mv sprintf.o sprintf.o.orig-libc
Extract the module out of /usr/lib/libc_G0.a by using:
# /bin/ar -x /usr/lib/libc_G0.a sprintf.o
Save the old one :
# /bin/mv sprintf.o sprintf.o.orig-libc_G0
Then replace the libraries modules with the ones provided: You will need
to move each of the named patches to be sprintf.o. You will want
to do this 1 patch at a time.
If the patches are located in /tmp
ON RISC
# mv sprintf.o-libc sprintf.o
# /bin/ar -r /usr/lib/libc.a /tmp/sprintf.o
# mv sprintf.o-libc_G0 sprintf.o
# /bin/ar -r /usr/lib/libc_G0.a /tmp/sprintf.o
Making sure it is in place:
# /usr/bin/ranlib /usr/lib/libc.a
# /usr/bin/ranlib /usr/lib/libc_G0.a
==============================================================================
/usr/bin/MIPS/BINARY/uipc_usrreq.o (CLD ipo5728)
CHECKSUM: 27580 65
----------------------------------
System panics "smp_lock_long: lock owner"
$x
cpu 2 proc = 802937e8 (idle)
stack = 80001964
state = boot run panic
panicstr = secondary cpu requested
locks held: 0
cpu 4 proc = 802937e8 (running)
stack = 7ffffd68
state = panic stop
panicstr = smp_lock_long: lock owner
locks held: 81bfc474 801483b8 80287fe0 0
$c
_stop_secondary_cpu() from 80003136 at 7ffffd18
_panic(800cbe23) from 800561bd at 7ffffd68
_smp_lock_long(81bfc474,1,8006d5c8) from 8004e1f5 at 7ffffdc8
_unp_connect(81bfc400,81c1cf80) from 8006d5c8 at 7ffffe10
_uipc_usrreq(81bfc400,9,81c54a80,81c1cf80,0) from 8006c848 at 7ffffe54
_sosend(81bfc400,81c1cf80,7fffff00,0,0) from 80067373 at 7ffffeb0
_sendit(d,7fffff54,0) from 8006b61e at 7fffff18
_sendto() from 8006b48e at 7fffff6c
_syscall() from 80099024 at 7fffffa4
_Xsyscall() from 800028c0 at 7fffb8ec
800cbe23/s
800cbe23: smp_lock_long: lock owner
===============================================================================
/usr/sys/MIPS/BINARY/vm_smem.o (CXO-9281, CXO-9597, CXO-9608,
CHECKSUM: 20315 53 CXO-9615, MUH-2451, UVO-1762,
------------------------------ UVO-1826, CFS.3534, VBO-1178)
Problem #1
**********
The panic "sm_retrieve_sa: Could not find SMS in proc" is caused by a
race between a process with multiple shared memory segments exiting and the
pageout daemon attempting to page out a page from one of those shared
memory segments.
Problem #2
**********
This patch fixes a race condition when shared memory is in heavy use
and the pager/swapper is looking to free memory. This problem is
identified by a panic "flushpte: smindex == -1"). A kernel rebuild is
required.
==============================================================================
/usr/sys/MIPS/nfs_subr.o (QAR 12744)
CHECKSUM: 12275 98
/usr/sys/MIPS/nfs_vfsops.o
CHECKSUM: 54696 78
/usr/sys/MIPS/nfs_vnodeops.o
CHECKSUM: 07691 96
/usr/sys/MIPS/vnodeops_gfs.o
CHECKSUM: 39963 70
/usr/sys/fs/nfs/vfs.h
CHECKSUM: 08019 6
/usr/etc/automount
CHECKSUM: 60551 310
/usr/include/sys/mount.h
CHECKSUM: 20504 12
/usr/sys/conf/mips/param.c
CHECKSUM: 36974 11
This patch is for XPG3 NFS branding. This patch provides support for the
"-o grpid" mount option. By default, NFS create file and directory semantics
are the "normal" ULTRIX semantics. XPG3 mandated behavior is accomplished by
editing "param.c" and changing nfs_mount_default to 1.
/usr/sys/MIPS/vnodeops_gfs.o also fixes: (CLD CXO9768, CXO10030, QAR10546)
CHECKSUM: 39963 70
----------------------------
The chgrp command would not succeed on files owned by the user. When exporting
an ULTRIX filesytem through nfs and doing a chgrp command on a file
that the user owns you will get:
chgrp: <filename>: not owner
This patch resolves the problem. When installing this patch on your ULTRIX
server you will also need to do a:
chmod 0755 /usr/bin/chgrp
/usr/etc/automount also fixes: (CLD CXO-9643)
CHECKSUM: 60551 310
---------------------
When automounter is using the built-in "hosts" map, it correctly mounts
files from some NFS servers (including Suns), but does not mount files
from ULTRIX NFS servers which export files to hosts by name. Any access
to one of these files fails with "no such file or directory". If the
ULTRIX server exports its files to "everyone", then the clients automount
mounts the files just fine. Also note that manually mounting the file
when its exported to the client by name works correctly.
In addition, this automount fixes a problem when automount'ing from
UCX running on a VMS system. This contains a fix for "*" in the
VMS exports file.
A kernel rebuild is required.
==============================================================================
XvmsCalloc.o (CLD CFS.3496)
CHECKSUM: 32436 4
---------------------
This patch fixes a problem in the X system version of calloc that causes
segmentation fault. The calloc() routine in Xlib calls malloc() to get an
unitialized piece of memory and tries to clear it. However, if malloc()
was unable to allocate space, calloc tries to zero out NULL and crashes.
INSTALLATION INSTRUCTIONS:
1) Save the previous version (assuming the new version is in /patches).
cd /usr/lib
ar -xv libX11.a XvmsCalloc.o
mv XvmsCalloc.o XvmsCalloc.o.orig
2) Replace the object file.
cp /patches/XvmsCalloc.o /usr/lib
cd /usr/lib
ar -rv libX11.a XvmsCalloc.o
3) Reinitialize libX11.a with new XvmsCalloc.o patch
ranlib libX11.a
==============================================================================
/usr/lib/lpdfilters/ln03of
CHECKSUM: 08033 64
--------------------------
Problem 1: (no cld associated)
*********
When lprm is used to kill the active printjob, the filter will now send
a printer reset command.
Problem 2: (QAR 12495)
*********
Underscore backspace characters causes incorrect linecount.
Files using backspace underscore characters to underline text would cause the
linecount to be incorrectly updated. This would cause an early page eject.
==============================================================================
/usr/lib/lpdfilters/ln03rof (no cld associated)
CHECKSUM: 46150 52
---------------------------
NOTE: This filter requires the current lpd patch file.
Problem 1: (no cld associated)
*********
Exceeding serial postscript printer memory causes data loss.
This problem involves serial postscript printers using the ln03rof filter
as the 'if' and 'of' filter.
When the the queue contains many files it is possible to exceed the printer
memory capacity. When the memory is exceeded the rest of the current
print job be lost.
This fix requires an lpd patch file, and setting fs#02 in the printcap file.
Problem 2: (no cld associated)
*********
Specifying a width greater than 132 would be ignored.
If the command 'lpr -w200' was entered the filter would not change the width
value because the number passed was to large. The filter has been modified
to use a width of 132 for any value greater than 132.
==============================================================================
/usr/lib/lpd
CHECKSUM: 25465 204
-----------------------
NOTE: If using the ln03rof filter, apply the current ln03rof patch file.
Problem 1: (no cld associated)
*********
Exceeding serial postscript printer memory causes data loss.
This problem involves serial postscript printers using the ln03rof filter
as the 'if' and 'of' filter.
When the the queue contains many files it is possible to exceed the printer
memory capacity. When the memory is exceeded the rest of the current
print job be lost.
This fix requires an ln03rof patch file, and setting fs#02 in the printcap
file.
Problem 2: (cld UVO-1988)
*********
XON character sent when DTR flow control is used.
When ps=PS is set in the printcap file, the print daemon would send and
XON character. This was causing problems for printers using DTR flow
control. This patch now tests the fs printcap entry to see if bit 01 is
set before sending the XON character.
Problem 3: (QAR-12456)
*********
TCPIP printing to a terminal server hangs daemon.
If the print daemon fails to connect to the terminal server because the
listener is not available, the daemon would hang. The daemon will now retry
until the listener is available.
Problem 4: (no cld associated)
*********
Filenames starting with # would not print on printservers.
It is now possible to print files who's name starts with a # character.
Problem 5: CXO-09125
*********
The lpd has been modified to process remote print requests from systems
that send the control file first followed by the data.
Problem 6: (no cld associated)
*********
The print daemon will no longer core dump when printing a non-postscript
banner page.
If the printcap pw value was greater than 132 and the filename printed in
big type on the banner page exceeded 132 characters the lpd would core dump.
==============================================================================
rpc_server.o (CLD UVO-01439)
CHECKSUM: 43711 54
rpc_client.o
CHECKSUM: 22871 40
---------------------
There was a problem with DECRPC V1.0 and excessive retransmissions wherein a
fragment of a fragmented message was missed. The server and the client would
sometimes never get back in sync because the same network congestion (actually
a kernel MBUF shortage) which caused the original problem would continue to
result in lost fragments, and the message would never arrive.
PATCH INSTALLATION:
1) first extract the modules out of /usr/lib/libnck.a
by using:
ar x /usr/lib/libnck.a rpc_server.o rpc_client.o
2) save them to:
mv rpc_server.o rpc_server.org.o
mv rpc_client.o rpc_client.org.o
3) then replace the library's modules with the provided ones:
ar r /usr/lib/libnck.a rpc_server.o rpc_client.o
4) to check it is in place:
ar ts /usr/lib/libnck.a
5) then recompile the application with links to /usr/lib/libnck.a
==============================================================================
/sys/MIPS/BINARY/gfs_sysquota.o (CLD UVO-1856)
CHECKSUM: 43566 57
-------------------------------
The system panic with message "closedq:stray dquot". The reason for the
problem is that when setting quota on user root, the dquot structure is
created in setdlim() which is not suppose to, because Q_NDQ flag is set
in quota structure. As a result, when quota is turned off or file system
is unmounted, closedq() is called to clean up the dquot structure and
its reference count. The dquot reference count is invalid because quota
structure has reference to dquot structure but it thinks that there is no
dquot structure because Q_NDQ flag is set. The solution is not allow setdlim()
to allocated dquot structure for a user if Q_NDQ flag is set in its quota
structure.
In setdlim(), in gfs_sysquota.c, return EINVAL error if Q_NDQ flag is
set in quota structure for a user, e.g root. This means that dquot
structure is not to be allocated and disc quota is not set for the user.
Command 'edquota' which is used to set quota limit for a user will return
error if it is used to set quota limit for the root.
A kernel rebuild is required.
==============================================================================
/sys/MIPS/BINARY/sim_94.o (KAOB65389,cxo10489) (RISC only)
CHECKSUM: 59325 173
/usr/sys/MIPS/BINARY/sim_sched.o
CHECKSUM: 49422 58
/usr/sys/MIPS/BINARY/sim_sm.o
CHECKSUM: 38609 61
--------------------------------
The following 3 kernel object modules are for fixing system hanging problem
under SCSI/CAM due to SCSI BUS RESET.
Kernel rebuild required
==============================================================================
/sys/MIPS/BINARY/vm_kmalloc.o (CXO-10607) (RISC only)
CHECKSUM: 07325 70
---------------------------------
On a heavily loaded system equipped with CITCA devices, we sometimes get a
"null kup" panic. This happens because the CITCA device driver allocates
wired memory, but the km_free() routine frees it to the unwired memory pool.
Kernel rebuild required
==============================================================================
/usr/sys/MIPS/BINARY/mem.o (CXO-10613) (V4.3a only)
CHECKSUM: 28183 45
--------------------------
On a heavily loaded system equipped with CITCA devices, we sometimes get a
"tlbmiss" panic. The frequency of the panic increases the heavier the
CITCA load is and the more we run "vmstat -k". This is because vmstat reads
/dev/kmem, and CITCA drivers can free memory at interrupt time. In general,
a system can "tlbmiss" panic during a read of /dev/kmem. This can happen
when an interrupt in the critical section changes the PTE for memory about
to be read, or when a second processor on the machine changes the PTE.
Kernel rebuild required
==============================================================================
Digital Internal Use Only
--- clip ---
--- clip ---
Digital Internal Use Only
V4.3a UWS Patches
July 30, 1993
/usr/bin/Xws
CHECKSUM: 18355 2228
/usr/lib/DPS/masterDPSVM
CHECKSUM: 09232 123
-------------------------
DESCRIPTION OF FIXED PROBLEMS:
Line Attributes Problem [PX] (CFS.2391)
Lines drawn with a line width of 19 pixels or greater and a fill style
of FillStippled are not displayed correctly.
Problem With Backing Store [HX] (CFS.2416)
If a window is brought up overlapping another window, lines drawn in
the original window show up in the overlapping window.
X server crashes when running RR Donnelley app. (CXO_08214)
If you bring up certain JCPenney catalog pages, using RRD app,
the Xwserver crashes if you have an HX adaptor. Sometines the images
are corrupted as well.
This fix requires that the -noacc option be specified to Xws for
it to work.
XWS DS5000/240-MX backingstore problems [MX] (MUH_02324)
On the DS5000 with MX card, the client calls "XFillRectangle"
Xlib call routine, then it will draw correctly 120 different
patterns. However, if use "XFillPolygon" with fg been black,
bg been white, the fg/bg color are then inverted. Problem does
not exist if the fg is white and bg is black or fg/bg are the
same color.
Display PostScript Problem (UVO-1928)
In a DPS application, text is postioned incorrectly after a
composite character (i.e. a letter with an accent mark).
Display PostScript Crashes Server (CXO-08843)
Certain Display PostScript sequences will cause the server to abort.
INSTALLATION INSTRUCTIONS:
To install the new image on a V4.3 system (assuming that the new
image is in the directory /patches):
$ su
# cd /usr/bin
# mv Xws Xws.orig <save original image>
# cp /patches/Xws .
# chmod 755 Xws
# cd /usr/lib/DPS
# mv masterDPSVM masterDPSVM.orig <save original file>
# cp /patches/masterDPSVM .
# chmod 644 masterDPSVM
To start the new server (Warning: this will kill your session and
bring you back to the login screen):
# ps -ax | grep Xws <find pid for Xws>
# kill -9 <Xws pid>
==============================================================================
/usr/bin/Xtx
CHECKSUM: 64270 2128
/usr/lib/DPS/masterDPSVM
CHECKSUM: 09232 123
-------------------------
DESCRIPTION OF FIXED PROBLEMS:
Display PostScript Problem (UVO-1928)
In a DPS application, text is postioned incorrectly after a
composite character (i.e. a letter with an accent mark).
INSTALLATION INSTRUCTIONS:
To install the new image on a V4.3 system (assuming that the new
image is in the directory /patches):
$ su
# cd /usr/bin
# mv Xtx Xtx.orig <save original image>
# cp /patches/Xtx .
# chmod 755 Xtx
# cd /usr/lib/DPS
# mv masterDPSVM masterDPSVM.orig <save original file>
# cp /patches/masterDPSVM .
# chmod 644 masterDPSVM
To start the new server (Warning: this will kill your session and
bring you back to the login screen):
# ps -ax | grep Xtx <find pid for Xtx>
# kill -9 <Xtx pid>
===============================================================================
/usr/bin/Xwst3d
CHECKSUM: 40426 2536
/usr/lib/DPS/masterDPSVM
CHECKSUM: 09232 123
------------------------------
DESCRIPTION OF FIXED PROBLEMS:
Line Attributes Problem [PXG] (CFS.2391)
Lines drawn with a line width of 19 pixels or greater and a fill style
of FillStippled are not displayed correctly.
Display PostScript Problem (UVO-1928)
In a DPS application, text is postioned incorrectly after a
composite character (i.e. a letter with an accent mark).
INSTALLATION INSTRUCTIONS:
To install the new image on a V4.3a system (assuming that the new
image is in the directory /patches):
$ su
# cd /usr/bin
# mv Xwst3d Xwst3d.orig <save original image>
# cp /patches/Xwst3d .
# chmod 755 Xwst3d
# cd /usr/lib/DPS
# mv masterDPSVM masterDPSVM.orig <save original file>
# cp /patches/masterDPSVM .
# chmod 644 masterDPSVM
To start the new server (Warning: this will kill your session and
bring you back to the login screen):
# ps -ax | grep Xwst3d <find pid for Xwst3d>
# kill -9 <Xwst3d pid>
==============================================================================
/usr/bin/dxsession_motif
CHECKSUM: 64487 4068
/usr/lib/X11/uid/DXsession.uid_motif
CHECKSUM: 57592 128
/usr/bin/dxsession_xui
CHECKSUM: 31733 2848
/usr/lib/X11/uid/dxsession.uid_xui
CHECKSUM: 60874 132
-------------------------------------
Both the Motif and XUI version of this patch fixes the following problem:
Problem 1: ~/.Xdefaults not read in correct sequence at login time
********* (CLD CXO_09214)
~/.Xdefaults is not being read in the correct sequence under ULTRIX v4.3a.
The .Xdefaults in the user's home directory is now read first prior to
changing to a new directory when logging in.
INSTALLATION INSTRUCTIONS:
PLEASE NOTE: In v4.3a there are two version of the dxsession, the
OSF/Motif version for Motif 1.1.3 and the XUI version. For these
patches, when installing either the Motif version, called dxsession_motif
and DXsession.uid_motif, or the XUI, called dxsession_xui and
dxsession.uid_xui, please rename them to their original names, i.e
dxsession, DXsession.uid, and dxsession.uid respectfully. The user will
only need to install either the Motif or XUI version of these patches.
To install the new files on a V4.3 system (assuming that you have
copied the new files to the directory /patches on your system):
If OSF/MOTIF v1.1.3 is installed on the system, PLEASE install patches
/usr/bin/dxsession_motif and /usr/lib/X11/uid/DXsession.uid_motif:
$ su
# cd /usr/lib/X11/uid
# mv DXsession.uid DXsession.uid.orig <save original file>
# cp /patches/DXsession.uid_motif DXsession.uid
# cd /usr/bin
# mv dxsession dxsession.orig <save original file>
# cp /patches/dxsession_motif dxsession
# chmod 4755 dxsession
Lastly, quit the session and log back in to start up the new
session manager.
If OSF/MOTIF is NOT installed on system, PLEASE install patches
/usr/bin/dxsession_xui and /usr/lib/X11/uid/dxsession.uid_xui:
$ su
# cd /usr/lib/X11/uid
# mv dxsession.uid dxsession.uid.orig <save original file>
# cp /patches/dxsession.uid_xui dxsession.uid
# cd /usr/bin
# mv dxsession dxsession.orig <save original file>
# cp /patches/dxsession_xui dxsession
# chmod 4755 dxsession
Lastly, quit the session and log back in to start up the new
session manager.
================================================================================
Digital Internal Use Only
~
--- clip ---
> micha
> --
> ==================================================================
> Michael Schneider
> Institut fuer Experimentelle Kernphysik der Universitaet Karlsruhe
> Engesserstr. 7(Physikhochhaus), 76128 Karlsruhe 1
> micha@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx | BD14@xxxxxxxxxxxxxx
> ==================================================================
>
Harald Lundberg <hl@xxxxxxxx>;Tekla Oy,Koronakatu 1,FIN-02210,ESPOO,FINLAND
tel +358-0-8879449;fax +358-0-8039489;res +358-0-8026752;2nd +358-11-728013
References:
- chown-patch
- From: micha@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx (Michael Schneider)