has requested a copy or has subscribed to the document named "New_AIXV4_Fixes".
If you would like to be removed from this mailing list, send e-mail to
aixservaustin.ibm.com with a subject of "unsubscribe New_AIXV4_Fixes", or
send a note to owner-aixservaustin.ibm.com with your request.

APAR: IY30971 COMPID: 5765E6100 REL: 510
ABSTRACT: ALTERNATE ROOTVG MIGRATION INSTALL OF BOS.CLVM.ENH FAILS

PROBLEM DESCRIPTION:
Customer gets error during bos.clvm.enh install process:
  Error: Enhanced Concurrent volume group(s) still active.
  Varyoff all Enhanced Concurrent volume groups
when attempting alternate rootvg migration install.

PROBLEM SUMMARY:
alternate rootvg migration install of bos.clvm.enh fails
with error:
Error: Enhanced Concurrent volume group(s) still active.
Varyoff all Enhanced Concurrent volume groups
This failure causes bos.clvm.enh not to be installed.

PROBLEM CONCLUSION:
Install scripts for fileset now appropriately check for
EC vgs only when doing a rootvg (not alternate rootvg)
install.

TEMPORARY FIX:
Varyoff concurrent volume groups, do standard migration
install rather than alternate rootvg.

------

APAR: IY31781 COMPID: 5765E6100 REL: 510
ABSTRACT: AUTOFS: PERMISSION DENIED WHEN A SUBDIRECTORY IS NOT MOUNTABLE

PROBLEM DESCRIPTION:
If the customer is using automount with the -hosts map on a
client, and the client does not have mount access to all
directories on a server, the customer will not be able to
access any of the directories from the server.

LOCAL FIX:
Enable COMPAT_AUTOMOUNT in the environement and restart
the automount daemon.

PROBLEM CONCLUSION:
When the client builds the list of exported directories to be
mounted from the server, check to see if there are any access
groups on the export, and if there are, don't add the directory
to the list unless the client is included in those groups.

------

APAR: IY33081 COMPID: 5765E0200 REL: 710
ABSTRACT: TOTALVIEW PROBLEM WITH MODULES & QSMP=NOOPT

PROBLEM DESCRIPTION:
Totalview's Tools|Fortran Modules menu item greys out when the
source is compiled with -qsmp=noopt.

------

APAR: IY33421 COMPID: 5765E6100 REL: 510
ABSTRACT: THREADINIT DSI W/ PSALLOC=EARLY

PROBLEM DESCRIPTION:
On 64bit kernel, setting PSALLOC=early, then malloc'ing
enough memory to reach psdanger (ie low paging space
condition) and then calling sync from within the
PSALLOC=early process will cause the system to
crash with the following stack trace...
    pvthread+009A80 STACK:
     002FA13C bcopy+00011C ()
     000B50B4 threadinit+0000C8 ()
     000D5DF8 procinit+000214 ()
     000240C8 initp+0000A8 ()
     05DA8DE8 npm_start+0000E4 ()
     05DA481C nfssync+000058 ()
     05DF7B48 nfs_sync+000014 ()
     05DF6F4C aix2onc_sync+000048 ()
     0047CD64 vfs_sync+000014 ()
     004635A4 sync+000074 ()
     00003780 sc_msr_2_point+000028 ()

PROBLEM SUMMARY:
The early allocation policy should not apply to stack
segments, but the stackseg flag for the segment control
block was not being checked in that case.

PROBLEM CONCLUSION:
The early allocation policy should not apply to stack
segments, but the stackseg flag for the segment control
block was not being checked in that case.

------

APAR: IY33800 COMPID: 5765E6100 REL: 510
ABSTRACT: VMSTAT TAKES MORE TIME

PROBLEM DESCRIPTION:
vmstat takes more time to report the memory related
data. Above is because of the logic of extracting the
data from the kernel.

------

APAR: IY34604 COMPID: 5765E6100 REL: 510
ABSTRACT: UNEXPECTED PAGEOUTS OF WORKING PAGES

PROBLEM DESCRIPTION:
A system on which the number of client pages are lower
than the maximum (numclient < maxclient), but free pages
are running low, might steal computational pages instead
of file pages even though numperm >= maxperm.

PROBLEM CONCLUSION:
Modify the way "file_only" page steals are decided upon when
the LRU daemon is started to steal client pages. Right now it
does not always take numperm/maxperm into consideration.

------

APAR: IY34622 COMPID: 5765F7000 REL: 810
ABSTRACT: QUERY REGARDING USE OF SEVER ERROR MESSAGE 1516-078

PROBLEM DESCRIPTION:
Query regarding severe error message 1516-078

------

APAR: IY34876 COMPID: 5765E6100 REL: 510
ABSTRACT: KSH DISPLAYS UNWANTED OUTPUT WITH BACKSPACE IN JA_JP

PROBLEM DESCRIPTION:
In Ja_JP locale, entering backspace in ksh command line vi mode
results in unexpected characters.

PROBLEM SUMMARY:
ksh displays unexpected output with backspace in Ja_JP
locale. Although the input is not corrupted, the display
is.

PROBLEM CONCLUSION:
The shell (ksh) displays unexpected output with backspace
in Ja_JP. The display is corrupt because the shell prints
some output before the prompt in cases when there is a 'c'
character in an echo statement.

------

APAR: IY34915 COMPID: 5765E6100 REL: 510
ABSTRACT: USB MOUSE DOES NOT WORK AFTER SYSTEM REBOOT

PROBLEM DESCRIPTION:
USB mouse does not work after system reboot. Mouse becomes
functional if it is disconnected and then reconnected. This
problem typically occurs when USB mouse is plugged into USB
keyboard. Generally hardware error will be logged with
the following detailed data:
FAILED
usb_intr
ADDITIONAL INFORMATION
4 5 ??????? F9 ???????

LOCAL FIX:
unplug and replug the mouse.

PROBLEM SUMMARY:
USB Driver will retry failed connect at IPL.

PROBLEM CONCLUSION:
Install PTFs for this APAR when available.

TEMPORARY FIX:
Unplug and re-plug USB Mouse.

------

APAR: IY34946 COMPID: 5765E6100 REL: 510
ABSTRACT: DEVICE NOT FOUND MESSAGE FROM CHLV AND RMLV

PROBLEM DESCRIPTION:
Device not found error message is sometimes generated by
chlv and rmlv although there is no problem.

PROBLEM CONCLUSION:
Corrected chlv and rmlv not to needlessly call tcbck.

------

APAR: IY34999 COMPID: 5765E6100 REL: 510
ABSTRACT: ERRPT -C MUST CLEANUP ODM WHEN KILLED WITH -1,-2,-3,-15

PROBLEM DESCRIPTION:
Any application can be killed by error daemon provided that it
has the same pid as the pid in errnotify entry left behind by
'errpt -c' after it is killed with either of the following
signals: 1, 2, 3, 15

PROBLEM SUMMARY:
Signals 1,2,3,15 were not handled, so odm was not
cleaned up properly (errnotify) after errpt -c ran.

PROBLEM CONCLUSION:
Handling signals 1,2,3,15 ensures that the errnotify
entry (en_name=concurrent) is deleted after errpt -c
is killed using the signals described above. This
eliminates the possibility of the error daemon
killing a process that has grabbed a pid previously
used by errpt (the error daemon runs the method
pointed to in the errnotify entry.
i.e. en_method = "/usr/bin/kill -30 41184")

TEMPORARY FIX:
Delete the odmentry by running:
odmdelete -q"en_name=concurrent" errnotify

------

APAR: IY35051 COMPID: 5765E6100 REL: 510
ABSTRACT: ETHERCHANNEL CANNOT SWITCH GB ADAPTERS W/BACKUP MODE

PROBLEM DESCRIPTION:
Etherchannel running over the IBM Gigabit
ethernet adapters is not able to recover
from a link down event when one of the
adapters is in standby mode.

PROBLEM SUMMARY:
Adapter device driver needs to implement link
callback function for link down events so
that etherchannel driver is aware that the
link is down.

PROBLEM CONCLUSION:
Add link callback function to link down
routine.

------

APAR: IY35056 COMPID: 5765E6100 REL: 510
ABSTRACT: CRASH CREATING LV WITH PP >32768

PROBLEM DESCRIPTION:
System crash when creating LV with a physical partition
greater than 32678 (likely on very large disks).

PROBLEM CONCLUSION:
Corrected code to handle greater physical partition numbers.

------

APAR: IY35088 COMPID: 5765E6100 REL: 510
ABSTRACT: GLOB() ROUTINE FAILS WITH GLOB_ABORTED ON FILES

PROBLEM DESCRIPTION:
If GLOB_ERR flag is set glob() routine should return with
GLOB_ABORTED only upon encountering a directory which cant
be opened/found. But glob() is failing with GLOB_ABORTED
for any file not found.

PROBLEM SUMMARY:
glob() returns with GLOB_ABORTED even incase of files
which are not found/readable

PROBLEM CONCLUSION:
glob() modified to see that it returns with GLOB_ABORTED
when GLOB_ERR is specifed only incase it is not able to
open a directory.

------

APAR: IY35111 COMPID: 5765E6100 REL: 510
ABSTRACT: MKTIME() CONVERSION FAILS WHEN MONTH SPECIFIED IS A MULTIPLE OF

PROBLEM DESCRIPTION:
mktime() routine is not able to convert properly when the
input month specified is negative and a multiple of 12.

PROBLEM SUMMARY:
mktime() routine outputs a wrong value, incase input month
is negative and a multiple of 12.

PROBLEM CONCLUSION:
There is a normalisation of month and year values done inside
mktime() routine. The logic has been modified to handle
negative values properly.

------

APAR: IY35121 COMPID: 5765E6100 REL: 510
ABSTRACT: DBX CAN CRASH UPON RECEIVING FAULTY LDR INFO

PROBLEM DESCRIPTION:
DBX can erroneously accept invalid file information from
ptrace about loaded modules. This results in unpredictable
core dumps when a user attempts to step or continue through
a process.

PROBLEM SUMMARY:
DBX can crash upon receiving faulty ldr info

------

APAR: IY35284 COMPID: 5765E6100 REL: 510
ABSTRACT: AT COMMAND ZULU OPTION PROBLEMS

PROBLEM DESCRIPTION:
at command ZULU option is not scheduling jobs at expected time

PROBLEM SUMMARY:
ZULU option doesnt work as expected particularly incase
of timezones with daylight savings on.

PROBLEM CONCLUSION:
at command modified to see that ZULU and GMT work in a
similar fashion. Also logic to determine the day of scheduling
has been enhanced.

------

APAR: IY35302 COMPID: 5765E6100 REL: 510
ABSTRACT: CSH BUILTIN COMMAND JOBS GIVES UNEXPECTED OUTPUT WITH PIPES

PROBLEM DESCRIPTION:
The jobs builtin command in csh gives incorrect output
when piped to another process.

PROBLEM SUMMARY:
csh jobs builtin does not give any output when piped
to another process

PROBLEM CONCLUSION:
The jobs builtin function in csh does not give any output
when piped to another process

------

APAR: IY35426 COMPID: 5765F7000 REL: 810
ABSTRACT: ADJUSTL PRODUCES INCORRECT OUTPUT

PROBLEM DESCRIPTION:
ADJUSTL produces incorrect output.

------

APAR: IY35582 COMPID: 5765E6100 REL: 510
ABSTRACT: MULTIPLE BINDS TO A PORT ERRONEOUSLY SUCCESSFUL

PROBLEM DESCRIPTION:
multiple binds to a port erroneously successful

PROBLEM SUMMARY:
multiple binds to a single port may be erroneoulsy
successful

PROBLEM CONCLUSION:
Added locks for the port to be bound.Each thread needing to
bind to a port must acquire a port specific lock before
running the bind code.This prevents multiple threads to be
able to bind to a single port.

------

APAR: IY35680 COMPID: 5765E6100 REL: 510
ABSTRACT: MOUNT AND/OR DF PRODUCES GARBAGE OUTPUT

PROBLEM DESCRIPTION:
Some garbage output may be displayed when running the 'mount'
or 'df' commands

PROBLEM SUMMARY:
df and mount can produce garbage data

PROBLEM CONCLUSION:
check validity of fields prior to printing them

------

APAR: IY35691 COMPID: 5765E0200 REL: 710
ABSTRACT: PERFORMANCE DEGRADATION FROM XLF 8.1 OVER XLF 7.1

PROBLEM DESCRIPTION:
Performance degradation from XLF 8.1 over XLF 7.1

PROBLEM SUMMARY:
Same as submitter's text

PROBLEM CONCLUSION:
Development found the problem

------

APAR: IY35839 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR IN XLFCODE

PROBLEM DESCRIPTION:
The following testcase gives an Internal Compiler Error message
in xlfcode:
   MODULE TEST
! Compilation command: xlf95 -c -qfree=f90 testcase.f
   CONTAINS
    Function ALLOC_CA (p, u, l, iv)
    character(*), pointer :: p(:)
    character(len(p)), pointer :: alloc_ca(:)
    character(*), optional, intent(in) :: iv
    integer, optional, intent(in) :: l, u
    alloc_ca=>allochar(p,u,l,iv)
    return
    end function ALLOC_CA
    Function ALLOCHAR (p, u, l, iv)
    character(*), pointer :: p(:)
    character(len(p)), pointer :: allochar(:)
    character(*), optional, intent(in) :: iv
    integer, optional, intent(in) :: l, u
    integer :: il, iu, lb, ub, rc
    allocate(allochar(il:iu),stat=rc)
    return
    end function ALLOCHAR
    END MODULE TEST
The error is:
/8.1.0.0/bin/xlf95: 1501-230 Internal compiler error; please
contact your Service Representative
1501-511 Compilation failed for file testcase.f.

PROBLEM SUMMARY:
Same as submitter's text

PROBLEM CONCLUSION:
Development fixed the defect. It will be part of the XL Fortran
for AIX 8.1.0.03 PTF.

------

APAR: IY35861 COMPID: 5765E6100 REL: 510
ABSTRACT: AUTOMOUNTD IGNORES MULTIPLE HOSTS MOUNT POINT IN MAP FILE

PROBLEM DESCRIPTION:
 utomountd takes into consideration only the first host listed
for the mount point in the automount map file and ignores the
rests of the hosts.
For example: If the map file contains an entry
dir ${HOST}:/export/data/& foo:/export/data/&
Then hosts foo and bar will be ignored.

PROBLEM SUMMARY:
If automount is able to mount from the first host then the
problem may pass unnoticed. But if it fails then automount
will not try the next host and when a user trys to cd to
the automount directory he will get a permission denied
error.
For example:
# ls /data/foo
ls: 0653-345 /data/foo: Permission denied.

PROBLEM CONCLUSION:
The loop to try all the mount points in the function mount_nfs
is missing. Fix is to restore the do while loop.

------

APAR: IY35862 COMPID: 5765F7000 REL: 810
ABSTRACT: INVALID SUBSCRIPT OUT OF BOUNDS TRAP WHEN -C IS USED

PROBLEM DESCRIPTION:
The customer's testcase runs as expected when compiled without
-C. However, adding -C -qsigtrap causes the application to trap
as runtime with an invalid error:
 Signal received: SIGTRAP - Trace trap
   Fortran language trap: subscript out of bounds
However, the code is not accessing memory out of the bounds of
the array.

LOCAL FIX:
do not use -C

------

APAR: IY35954 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR ONLY WITH XLF 8.1 USING -O3 -QSMP=OMP

PROBLEM DESCRIPTION:
With a testcase, compiling with -O3 gives an Internal Compiler
Error message. Removing -O3, or compiling with -O or -O4 does
not give an Internal Compiler Error. In addition to -O3, adding
the option -qsmp=omp gives the error message:
1586-346 (S) An error occurred during code generation. The code
generat(the code generation return code for error 1586-346 was
40).

LOCAL FIX:
do not compile with -O3.

PROBLEM CONCLUSION:
Development fixed this defect in XLF 8.1.0.02 PTF.

------

APAR: IY35995 COMPID: 5765E6100 REL: 510
ABSTRACT: PROGRAM HANG/FAILURE WITH FORK AND SIGWAIT

PROBLEM DESCRIPTION:
Child process from a fork may hang or fail if
parent was doing sigwait at the time of fork.

LOCAL FIX:
Cancel sigwaiting threads before forking;
or do exec from child process.

PROBLEM SUMMARY:
Child process from a fork may hang or fail if parent was doing
sigwait at the time of fork. The faulting thread will be in
either _pthread_deactivatex or _wait_list_add.

PROBLEM CONCLUSION:
Properly initialize the pthread library's internal list
of sigwaiting threads in its post-fork handler.

------

APAR: IY36069 COMPID: 5765F7000 REL: 810
ABSTRACT: ICE:TRAP ENCOUNTERED IN ASTI

PROBLEM DESCRIPTION:
ICE:Trap encountered in ASTI

------

APAR: IY36187 COMPID: 5765F7000 REL: 810
ABSTRACT: INCORRECT OUTPUT WHEN USING -QINTSIZE=8

PROBLEM DESCRIPTION:
Incorrect output when using -qintsize=8

------

APAR: IY36203 COMPID: 5765E6100 REL: 510
ABSTRACT: TOPAS DISPLAYS INCORRECT DISK AND NETWORK VALUES

PROBLEM DESCRIPTION:
Topas will display incorrect KBPS data for Disk and
Network values.

PROBLEM CONCLUSION:
Correct the KBPS and MBPS calculations in topas.

------

APAR: IY36217 COMPID: 5765E6100 REL: 510
ABSTRACT: MKSECLDAP FAILS FOR LDAP 4.1

PROBLEM DESCRIPTION:
mksecldap command fails to configure LDAP servers for
IBM Directory 4.1.

PROBLEM CONCLUSION:
Update the mksecldap script to work with the new IBM
Directory 4.1 version

------

APAR: IY36254 COMPID: 5765F7000 REL: 810
ABSTRACT: QUERY REGARDING DISCREPANCY BETWEEN XLF AND SUN OUTPUT

PROBLEM DESCRIPTION:
Query regarding discrepancy between XLF and Sun output

------

APAR: IY36267 COMPID: 5765F7000 REL: 810
ABSTRACT: CODE GENERATION ERROR IN XLF 8.1.0.01 PTF ONLY USING -O4 OR -O5

PROBLEM DESCRIPTION:
The customer's testcase gives the following error using -O4 or
-O5:
1586-346 (S) An error occurred during code generation. The code
generation return code is 217.
1501-511 Compilation failed for file bug.f.
At opt levels 3 and below, the testcase compiles.
This defect only exists in XLF 8.1. XLF 7.1.1 does not have this
defect.

LOCAL FIX:
Do not compile with opt levels 4 and 5.

------

APAR: IY36338 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR IN XLFENTRY

PROBLEM DESCRIPTION:
The customer's testcase gives an Internal Compiler Error in
xlfentry. The error occurs with all levels of XLF 7.1.1 and XLF
8.1.

PROBLEM CONCLUSION:
Development fixed this apar. It will be available in the XLF
8.1.0.03 PTF.

------

APAR: IY36445 COMPID: 5765E6100 REL: 510
ABSTRACT: NIS: BAD FREE IN SV_YPALL_CALLBACK

PROBLEM DESCRIPTION:
Calling the nis version of getservbyname more than
once will coredump if debug malloc postfree checking
is turned on. Even if debug malloc postfree checking
is turned off we are likely corrupting the heap the
second time getservbyname is called.

PROBLEM SUMMARY:
The sv_ypall_callback function will call strdup
to allocate new memory for the result of the
getservbyname call. This memory will be free'ed
by subsequent calls to getservbyname.

PROBLEM CONCLUSION:
The sv_ypall_callback function will call strdup
to allocate new memory for the result of the
getservbyname call. This memory will be free'ed
by subsequent calls to getservbyname.

------

APAR: IY36560 COMPID: 5765E6100 REL: 510
ABSTRACT: CHANGE PROCFS TO NOT IGNORE TRACED SIGNALS

PROBLEM DESCRIPTION:
signals that are traced through procfs and ignored
are actually held and may be delivered if disposition changed
to default

PROBLEM CONCLUSION:
do not allow procfs traced signals to be ignored, similar
to ptrace

------

APAR: IY36564 COMPID: 5765F7000 REL: 810
ABSTRACT: EXCEPTION TERMINATES PROGRAM

PROBLEM DESCRIPTION:
Exception terminates program

LOCAL FIX:
Use -qnohot compiler option

------

APAR: IY36566 COMPID: 5765F7000 REL: 810
ABSTRACT: ICE IN ASTI

PROBLEM DESCRIPTION:
ICE in ASTI

LOCAL FIX:
Compile using -qnohot

------

APAR: IY36746 COMPID: 5765E6100 REL: 510
ABSTRACT: FIND -FOLLOW DOES NOT FOLLOW LINKS PROPERLY

PROBLEM DESCRIPTION:
The find command when given with -follow option does
not follow links properly and gives error messages.

PROBLEM SUMMARY:
The find command does not follow the links properly when
given with -follow option.

PROBLEM CONCLUSION:
Code changes are done to 'find' command to follow
links properly with -follow option.

------

APAR: IY36785 COMPID: 5765F7000 REL: 810
ABSTRACT: XLF V8 AND TOTALVIEW V6 INTEROPERABILITY

PROBLEM DESCRIPTION:
XLF V8 and Totalview V6 Interoperability

------

APAR: IY36792 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR WITH 8.1.0.02 PTF (NO OPTIMIZATION)

PROBLEM DESCRIPTION:
Using the 8.1.0.02 PTF, the customer is getting an Internal
Compiler Error. Optmization is not being used.

PROBLEM CONCLUSION:
The defect appears as an Internal Compiler Error when compiling
without optimization and the user code is calling matmul.

------

APAR: IY36955 COMPID: 5765E6100 REL: 510
ABSTRACT: MMAP SYSTEM CALL CAN CRASH THE SYSTEM

PROBLEM DESCRIPTION:
System can crash while executing the mmap
system call, dump will show below given
stack trace.
vm_map_enter+000488
mmap_common+00046C
_mmap64+000040
mmap+0000E8

PROBLEM SUMMARY:
Applications which does mamp can crash
the system and stack trace will be
as given below:
vm_map_enter+448
mmap_common+00046C
_mmap64+000040
mmap+0000E8

PROBLEM CONCLUSION:
mmap code modified to resolve the problem.

------

APAR: IY36980 COMPID: 5765E6100 REL: 510
ABSTRACT: 64BIT RPC SERVER APPL. DON'T WORK

PROBLEM DESCRIPTION:
rpc server application when compiled in 64 bit (using -q64
compiler option) do not work. A way to detect this is by
testing it with rpcinfo.
rpcinfo -T tcp host prognum versnum
would fail with the error:
rpcinfo: RPC: Timed out program <prognum> version <vernum> is
not available

PROBLEM SUMMARY:
Customer will find their 64 bit rpc server though up and
running but not responding to rpc requests (as it is not
able to decode/encode the messages properly).
Also during the investigation of this bug, it was found
64-bit rpc applications may core dump and fail to start
due to a bug in libpthreads.a. This has also been fixed
but under a separate APAR (IY37240).

PROBLEM CONCLUSION:
The function xdr_callmsg which decodes and encodes rpc
messages in libnsl.a is missing. In the past it was getting
resolved by linking with libc.a (which has xdr_callmsg). For
32-bit the structure rpc_msg in libc.a and libnsl.a match
hence it worked fine. For 64bit the structure don't match
hence the messages which are decoded properly on libc side
are not interpreted properly on libnsl side. The fix is to
introduce the xdr_callmsg in libnsl.a.
The core dump (see above) occurs because libnsl's entry point
calls pthread routine pthread_self (before libpthreads's
initialization routine is called). It is being fixed by
a check in pthread_self. The check calls pthread's init
function if not already invoked. The APAR number for this
is IY37240.
Both the APARs are needed for proper functioning.

------

APAR: IY37008 COMPID: 5765E6100 REL: 510
ABSTRACT: RPC.LOCKD HUNG IN STOPPING STATE

PROBLEM DESCRIPTION:
If there are nfs clients blocked waiting for file locks when
stopsrc -s rpc.lockd is attempted on the server the
rpc.lockd will enter stopping state but never reach
inactive state. This means the client will not be able
to unmount the filesystem where the lock resides and
the server will not be able to restart rpc.lockd.
Rebooting the server would be required to clear the
condition.

PROBLEM SUMMARY:
The transport specific close routine for rpc.lockd
was not being properly called.

------

APAR: IY37039 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR IN XLFHOT WITH THE 8.1.0.02 PTF

PROBLEM DESCRIPTION:
With the 8.1.0.02 PTF, the testcase supplied by the customer
fails with an Internal Compiler Error in xlfhot. The same
testcase, when compiled with -O2, passes. The testcase also
passes with the 8.1.0.0 and 8.1.0.01 level.

LOCAL FIX:
compile with -O2

PROBLEM CONCLUSION:
Development fixed apar. It will be available in the XLF 8.1.0.03
PTF.

------

APAR: IY37058 COMPID: 5765E6100 REL: 510
ABSTRACT: KSH DUMPS CORE IN PATH_PARENTNAME ROUTINE

PROBLEM DESCRIPTION:
When the PWD env variable is unset, executing a script
more than once may cause ksh to core dump.

PROBLEM SUMMARY:
When the PWD environment variable is unset in
the parent shell, ksh will core dump in the
child.

PROBLEM CONCLUSION:
The problem will be fixed through code changes
to ksh.

------

APAR: IY37063 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR IN XLFCODE WITH

PROBLEM DESCRIPTION:
With the customer's testcase, there is an Internal Compiler
Error in xlfcode in all levels of the XLF 8.1 product including
the latest 8.1.0.02 PTF. The following is the compile line:
xlf -qarch=pwr3 -c -O3 -qstrict -qfloat=rsqrt
If I remove any of the above compile options, then the compiler
passes.

LOCAL FIX:
remove any of options:
-qarch=pwr3 -c -O3 -qstrict -qfloat=rsqrt

PROBLEM CONCLUSION:
Development fixed the apar and it is part of the 8.1.0.04 PTF

------

APAR: IY37083 COMPID: 5765E6100 REL: 510
ABSTRACT: RACE CONDITION IN LIO_LISTIO WITH LIO_WAIT FLAG

PROBLEM DESCRIPTION:
customer has testcase with two similar programs that call
lio_listio with LIO_WAIT flag set. One program uses alarm with
a signal handler that causes lio_listio to return with EINTR.
This program works but causes race condition that causes other
program not using signal handler to hang.

PROBLEM CONCLUSION:
We use an extra knot in the knot_struct that is used in
keeping track of requests completed to also keep track of
et_post done by kproc to insure neithor calling code or kproc
delete knot structure until the other is done with it.

------

APAR: IY37137 COMPID: 5765E6100 REL: 510
ABSTRACT: ALT_DISK_MIGRATION ERROR RUNNING BOSBOOT

PROBLEM DESCRIPTION:
When running nimadm an error creating the
client's boot image is reported:
0505-180 alt_disk_mig: Error creating client
boot image.
The errors reported by bosboot can be seen in the
alt_disk_migration log. The message is different
depending the version:
FOR AIX 5.1
bosboot: The memory or boot logical volume size
is too small.The boot logical volume size may be
increased by executing the kdb_resize command
(Example: kdb_resize hd5).
FOR AIX 5.2:
0301-179 bosboot: The boot logical volume size
must be at least 12 MB.

LOCAL FIX:
An e-fix can be made available upon request.
There are no known ways to work around the problem
at this time.

PROBLEM SUMMARY:
nimadm fails with the following error messages:
0505-180 alt_disk_mig: Error creating client boot image.
The message seen by nimadm from bosboot
(recorded in the log), is:
FOR AIX 5.1
bosboot: The memory or boot logical volume size is too small.
The boot logical volume size may be increased by executing the
kdb_resize command (Example: kdb_resize hd5).
FOR AIX 5.2:
0301-179 bosboot: The boot logical volume size must be at
least 12 MB.

PROBLEM CONCLUSION:
Fixed nimadm bosboot setup to validate hd5 by major/minor
access numbers.

------

APAR: IY37225 COMPID: 5765E6100 REL: 510
ABSTRACT: UNSUCCESSFUL_LOGIN_COUNT ATTRIBUTE MAY NOT BE INCREMENTED

PROBLEM DESCRIPTION:
The attribute unsuccessful_login_count may
not be incremented in domains like LDAP
for a ldap user when the user authentication
fails. This may also be true for other
domains like AFS, DCE, NIS, NISPLUS. There
is no problem with the local updates in the
/etc/security/lastlog.

PROBLEM CONCLUSION:
Set the registry value of the user correctly.

------

APAR: IY37239 COMPID: 5765F7000 REL: 810
ABSTRACT: ICE IN ASTI WITH -Q64 AND -C

PROBLEM DESCRIPTION:
!***************************************************************
!*
!* DEFECT ABSTRACT : NCAR: ICE in asti
!* CUSTOMER : NCAR:James Edwards
!* PMR NUMBER : 67780,033,000
!* COMPILER INVOCATION : xlf90_r
!* COMPILER OPTION(S) : -c -C -q64
!* OS OPTION(S) : N/A
!* OS LEVEL : AIX 5.1
!* COMPILER LEVEL : XLF 8.1.0
!* MACHINE TYPE/MODEL :
!*
!* DESCRIPTION : ICE in asti (see below)
!*
!*
================================================================
!
! TEST CASE:
!
 subroutine surf_inti( plon, plonl, platl, pplon )
      implicit none
      real, allocatable, dimension(:,:,:) :: desert
      integer, intent(in) :: plon, plonl, platl, pplon
      real :: wrk2d(plon,platl)
      allocate( desert(plonl,platl,pplon))
      desert = reshape( wrk2d, (/plonl,platl,pplon/), order =
(/1,3,2/) )
end subroutine surf_inti
!
!---------------------------------------------------------------
!
! OUTPUT RECEIVED USING 8.1.0.02:
!
! $ xlf90_r -v -qdebug=except -c -C -q64 test.f
!exec:
/usr/lpp/xlf/bin/xlfentry(/usr/lpp/xlf/bin/xlfentry,test.f,/tmp/
F8DFeC7b,/tmp/
F8DFeC7bF.lst,xlfsmsg.cat,xlfmsg.cat,test.f,64(largetype),XLF90(
noautodealloc,
nosignedzero),FREE(f90),THREADED,DEBUG(except),CHECK,64,WSTREAMS
(/tmp/F8DDeC7ah1,
/tmp/F8DDeC7ab1,/tmp/F8DDeC7as1),DEFMSG(/usr/lpp/xlf/bin/default
_msg),DEBUG(NTRANS,ASTI),
-I/usr/lpp/xlf/include_64ldt,NULL)
!No licenses available. Contact your program supplier to add
!additional users. Compilation will proceed shortly.
!** surf_inti === End of Compilation 1 ===
!exec:
/usr/lpp/xlf/bin/xlfhot(/usr/lpp/xlf/bin/xlfhot,/tmp/F8DDeC7ah1,
/tmp/F8DDeC7ah2,
!/tmp/F8DDeC7ab1,/tmp/F8DDeC7ab2,/tmp/F8DDeC7as1,/tmp/F8DDeC7as2
,/tmp/F8DFeC7b,/tmp/F8DFeC7bA.lst,
!-qdebug=except,NULL)
!
!Internal Compiler Error: chaos: compilation abandoned -
operands have differing types, code 628
  Trap encountered in ASTI:
  SIGABRT - Abort
  Traceback:
  Offset 0x0000011c in Error
  Offset 0x00000034 in chaos__FPci
  Offset 0x0000005c in op_cmp__Fi
  Offset 0x000000ac in op_logical__Fi
  Offset 0x000000ac in op_logical__Fi
  Offset 0x000000ac in op_logical__Fi
  Offset 0x0000005c in wc_expr__Fi
  Offset 0x000001d4 in flow_stmt__Fi
  Offset 0x00000178 in gen_stmts__Fi
  Offset 0x00000bc4 in htow__FP22TOBEY_W_Code_Options_t
  Offset 0x00000038 in hil_to_wcode
  Offset 0x00000804 in main
  --- End of call chain ---
1501-511 Compilation failed for file test.f.
!
!
!
! EXPECTED OUTPUT:
!
! $ xlf90_r -v -c -C -q64 -F/orbit/xlf.v7r11/deliver/xlf.cfg51
test.f
exec:
/orbit/xlf.v7r11/deliver/bin/xlfentry(/orbit/xlf.v7r11/deliver/b
in/xlfentry,test.f,
/tmp/F8lleAEb,/tmp/F8lleAEbF.lst,xlfsmsg.cat,xlfmsg.cat,test.f,6
4(largetype),XLF90
(noautodealloc,nosignedzero),
FREE(f90),THREADED,NOLM,CHECK,64,WSTREAMS(/tmp/F8lleAEah1,/tmp/F
8lleAEab1,/tmp/
F8lleAEas1),DEFMSG(/orbit/xlf.v7r11/deliver/bin/default_msg),-I/
orbit/xlf.v7r11/deliver/
include_64ldt,NULL)
** surf_inti === End of Compilation 1 ===
1501-510 Compilation successful for file test.f.
! ------------------------ EOF

LOCAL FIX:
Compile without either -q64 or -C.

PROBLEM CONCLUSION:
Development fixed the defect in xlfhot.
Missing converts on some range-checking generated code

------

APAR: IY37246 COMPID: 5765E6100 REL: 510
ABSTRACT: PLOCK WITH UNLIMITED STACK CAUSES HANG

PROBLEM DESCRIPTION:
The sbrk()/brk() system calls can hang in an infinite loop
if the stack limit is set to unlimited and the plock() system
call is used to pin the data segment.

LOCAL FIX:
Do not set stack limit to unlimited.

PROBLEM CONCLUSION:
Fix the brk/sbrk system calls.

------

APAR: IY37250 COMPID: 5765E6100 REL: 510
ABSTRACT: AUTHENTICATION ONLY LOADMODULES SUPPORT BROKEN

PROBLEM DESCRIPTION:
The user defined authentication-only loadmodule which doesn't
have identification methods like method_getgrgid will
cause login to coredump when the module is used for
authentication. This is in AIX 4.3.
The core dump is also seen when method_getentry is not defined
in the loadmodule for AIX 5.1 and above where compound
loadmodules are defined, in addition to the above problem
seen in AIX 4.3.

PROBLEM CONCLUSION:
The if condition now checks if the methods are defined in
the module or not.

------

APAR: IY37320 COMPID: 5765E0200 REL: 710
ABSTRACT: ICE IN ASTI WITH -QFIXED

PROBLEM DESCRIPTION:
!*************************************************************
!*
!* DEFECT ABSTRACT : ICE in ASTI with -qfixed
!* PMR NUMBER : 66511,033,866
!* COMPILER INVOCATION : xlf90_r
!* COMPILER OPTION(S) : -qfixed -c
!* OS OPTION(S) : N/A
!* OS LEVEL : AIX 5.1
!* COMPILER LEVEL : XLF 7.1.1 and 8.1.0
!* MACHINE TYPE/MODEL :
!*
!* DESCRIPTION : xlf90_r crashes at ASTI when compiling test
!* case.
!*
!* ============================================================
!**************************************************************
! TEST CASE: (/:/projects/vacsup/testcase/aix/aix/66511/man5.f)
!
           CALL TWOPHASE(
. BTDB,BEE,DBEE,JAC,JAC1,SAMP,ORT ,
. COORD,COORDF,GPCOR,KP,DTKD,C,DERF,DERIVF,KDERIV ,
. GRAV,VOLF,KAY,DERIVT,DERIV,CNEW,THM,KM,G,NFVAL ,
. DIRCS,WEI,NCOL,NDEG,NLV,JACT,DF,DQ,DDQ,FLOW ,
. PTIM,WT,CPP,SGPR,SCRATCH,OLDBDY,GVAL ,
. CSS,CPS,CSP,HSP,QP,CC,HPP,HPS,HSS,QPT,QS,CPPF ,
. CSSF,CPSF,CSPF,VOLF2,C2,KAYT,KAYW,KAYN,KAYW0 ,
. KAYN0,DTKD2,KDERVN,KDERVW,TXDPC,TKT,TKS,TKN,TKW ,
. TPOR2,TKAYW1,TKAYN1,TKAYW2,TKAYN2,TKAYW3,TKAYN3 ,
. TCAYW1,TCAYN1,TCAYW2,TCAYN2,TCAYW3,TCAYN3,CAYW ,
. CAYN,TWETV,TOILV,TIKAP,TVKRW,TVKRN,TVKRW0,TVKRN0,
. EDGBD2,EDGBD3,FUNFD,FUNFD1,FUNFD2,VELCT,VELCW ,
. VELCN,PELD,SELD,DPELD,DSELD,DTKD41,DTKD42 ,
. TOILV2,TWETV2,VOLBN,VOLBW,PREFN,PREFW,PREFR ,
. BODYS,BODYN,BODYW,DEPTH,RTHICK,PREINI,SATINI ,
. STORKB,RPERM,KTUP,KWPARM,PORUP,VAL3,VAL3S,NUM3 ,
. NODNUM3,NUM3T,NUM2T,NUM4T,VAL4,LOADSCC,RSCRTCH ,
. NUM4,NODEW,WELINJ,WELPRO,WELPRW,DISINI,LOADSC ,
. VIPTYP,VWPARM,VAL5,VAL5D,NUM5,NUM5D,VAL6,NUM6 ,
. KDIAGT,LOADT,LOAD0T,XV,YV,ZV,NELHIT,ZICOD,ETACOD,
. ZETACOD,LAYMAX,WMAX,NUM7,VAL7,NF2,LOADSN,G2 ,
. CRLDST,SHEATR,SHEATW,SHEATN,ROCKE,TEMRES,NFT ,
. INITW,NELTYP,CWELL,RWELL,SNODE,SNODN,VELNEL ,
. IPE,STRNI,LOADL,NELNAD,STORBKT,VALCG,BK,AZ,BZ ,
. PIVOT,RESTRN,IRESTRN,IRESTR,VONEM,IVONEM ,
. ITAP,PBL,PBR,WORK,RR,PP,ZZ,BK2,BK3,BKET,BKDT ,
. PBLT,PBRT,WORKT,BK2T,BK3T,ELASTT,NSTBLKT ,
. VARINT,DVTTAB,DVTVAL,CRPTOT,TVTIM,CRPBDY ,
. CLOADS,CRLOAD,CLOAD,TEVPTOT,VISTIM,VISTIMJ ,
. TOTKE1,TOTKE2,TOTKE3,TOTKE4,BK2ELE,XWELLP,YWELLP,
. NLR,IVISC,CARD,IPOROS,PTOL,PTOLT,IFEMVI,LOGSTA ,
. LOGSCR,IELAST,IFREQN,IPERM,IFLUX,IPHASE,IKEEPF ,
. IDATEC,INDFL,IFLOW,ILOADSC,INTR,RELINP,PRNMUL ,
. SIZUP1,SIZUP2,SIZUP3,EASTX,MAXJOI,IFORS ,
. CONVRU,MAXUNI,ITOL,IWET,INON,IWETS,INONS,IKDIAGT,
. IBKET,IBKDT,ILOADT,ICRLDST,IHEATR,IHEATW,IHEATN ,
. IROCKE,HILOEL,ICON_G ,
. IRBK2T,IRBK3T,IELASTT,ISTBLKT,IPIVOT,IDEFT,ITEMR,
. INFT,CORNER,ILOADL,ISTORT,IV1,IV2,IV3,W1S,W2S ,
. W3S,ISCALE,NPR,NFLW,WELLNAME,NPERF,NPERFA,DOOR ,
. FLUID ,
. RADRW,FRADRW,RADR0,BDEPTH,FLWI,FLWP,QUART,IFLWI ,
. IFLWP,IPERF,IPERFA,NODLAY,NODLAYA ,
. IPRODA,WELLARRAY,IWALL,NUMPER,NPER,BHP ,
. IBHPI,BHPI,IBHPP,BHPP,EQUBHP ,
. PDW,ISHED,NWELL,NTOTP,BHPVAL,NBHPVAL,FLOVAL ,
. NFLOVAL,ICNNCT,NCNNCT,IFILEP,IVISKEY,NODTES ,
. INODTE,TFAIL,N29,TITLET,SIGNIF,NPHASE,ITKET ,
. IDLTU,IDLTP,TIMEST,TIMESU,ISTBLKS,NSTBLKS,NITMAX,
. ISETD0,WELLN,IWELLN,WELINT,IWELINT,WELINTA,JWCNT,
. NPERFT,NPERFTC,NMAT2,IRBKE1 ,
. IRBKE2,ICORELE,NDOFF,ITELE1,ITELE2,IKE1,IKE2 ,
. IXV1,IXV2,IYV1,IYV2,IZV1,IZV2,NSAMPR,SAMPR,ILAYV,
. NP8,INEL1,INEL2,ICRIT,IRFOS,IRYFA,IHARD,IPORE ,
. ISCTH,IDTSIZ,IPPSIZ,NNFVAL,NINIT,IDVTT1,IDVTV1 ,
IDVTV2,NPOINTS,IWMAX,INGI,ITOTP,IDEFCR,NN2,NDISP2,
ILOADSN,DVTTYPE,NDVTTB,NHIGH,IHIGH,ICALCDT,SLDDAT,
. PRDDAT,IDISPE,CMETHOD,PNGTOL,IMPLOT,IBOXN,IHELPN,
. RJTYPE,VMATRIX,ILOADSCC,NVJ1,NVJ2,NVJ3 ,
. NVJ4,NROT ,
. YLDTOL,NQUICK_GP,NQUICK,VDIR,PERFAC,NUMFIND ,
. NUMTIME,FINDNAME,FINDWELL,FINDTIME ,
. FINDTYPE,WELLNUM,NEWELL ,
. RSMREAD,RESDATA,ICALCR,RSIZEX,RSIZEY,RSIZEZ ,
. INODLAY,IJWCNT,NFMP,MPC1,MPC2,NUMP,NUMP1,NUMP2 ,
. IMPCON,NMPCON,RNUMP,NUMPR1,NUMPR2,REGION,IREG ,
. SATDATA,ISAT1,ISAT2,ISAT3,ISAT4,ISATREG ,
. PVTDATA,IPVT1,IPVT2,IPVT3,IPVT4,IPVTREG,NVAL7 ,
. SATTYPE,PVTTYPE,NFCO,NFCO1,NFCO2,NMPCONF,NMPCOND,
. findbname,findblock,numfindb,TVITIME ,
. NGI_METHOD,ICALTH,FILES_ARR,NPOINT ,
SIZ_IA,SIZ_JA,IA,JA,TRANS,G_MIN,G_MAX,NOD_POINT,doi10)
END
!
! -------------------------------------------------------------
!
! OUTPUT RECEIVED USING 7.1.1.04 and 8.1.0.03:
! ============================================
!
!$ xlf90_r -c -qfixed -v -qdebug=except man5.f
!exec: /usr/lpp/xlf/bin/xlfentry(/usr/lpp/xlf/bin/xlfentry,
!man5.f,/tmp/F8Wl0oib,/tmp/F8Wl0oibF.lst,xlfsmsg.cat,xlfmsg.cat,
!man5.f,32,XLF90(noautodealloc,nosignedzero),FREE(f90),THREADED,
!DEBUG(except),FIXED,WSTREAMS(/tmp/F8Wl0oiah1,/tmp/F8Wl0oiab1,
!/tmp/F8Wl0oias1),DEFMSG(/usr/lpp/xlf/bin/default_msg),
!DEBUG(NTRANS,ASTI),-I/usr/lpp/xlf/include_32,NULL)
!No licenses available. Contact your program supplier to add
!additional users. Compilation will proceed shortly.
!** _main === End of Compilation 1 ===
!exec: /usr/lpp/xlf/bin/xlfhot(/usr/lpp/xlf/bin/xlfhot,
!/tmp/F8Wl0oiah1,/tmp/F8Wl0oiah2,/tmp/F8Wl0oiab1,/tmp/F8Wl0oiab2
,
!/tmp/F8Wl0oias1,/tmp/F8Wl0oias2,/tmp/F8Wl0oib,
!/tmp/F8Wl0oibA.lst,-qdebug=except,NULL)
!
!Internal Compiler Error: wstack: stack overflow
! Trap encountered in ASTI:
! SIGABRT - Abort
!
! Traceback:
!
! Offset 0x0000011c in Error
! Offset 0x0000030c in enter__5wilcuFP2bu
! Offset 0x0000003c in readcu__5wilcuFl
! Offset 0x00000254 in WHTran__Fv
! Offset 0x0000078c in main
!
! --- End of call chain ---
!1501-511 Compilation failed for file man5.f.
!
!
! -------------------------------------------------------------
!
! EXPECTED OUTPUT:
! ================
! Successful compile.
! ---------------------------- EOF ----------------------------

PROBLEM CONCLUSION:
The w-code stack used in wh is of
fixed size, currently limited to
500 w-code instructions. For each
argument of this call we need two
w-code instructions, so we overflow
the queue.

------

APAR: IY37425 COMPID: 5765E6100 REL: 510
ABSTRACT: INCORRECT SMIT FIELDS FOR GIGIBIT PCI-X ADAPTER

PROBLEM DESCRIPTION:
"10/100/1000 Base-TX PCI-X Adapter (14106902)" and
"Gigabit Ethernet-SX PCI-X Adapter (14106802)" SMIT
identifiers do not match the possible input values.

PROBLEM CONCLUSION:
Modified driver to have the correct identifiers for the input
fields.

------

APAR: IY37440 COMPID: 5765E6100 REL: 510
ABSTRACT: DBX: TRACEI BROKEN FOR 64-BIT APPS ON P6XX SYSTEMS

PROBLEM DESCRIPTION:
breakpoints and tracepoints don't work when using DBX on
64-bit applications.

PROBLEM CONCLUSION:
Don't attach application text segment when ptracing self.

------

APAR: IY37455 COMPID: 5765E6100 REL: 510
ABSTRACT: NFSD DIES AND FAILS TO RESPAWN

PROBLEM DESCRIPTION:
nfsd keeps dying (error id E18E984F) and gets respawned with
errcode -9035 ( subsystem ended will be restarted). After a
couple of restarts nfsd stops respawning with errorcode -9020
(retry limit exceeded).
It is normally noticed when there are large number of PC
clients or when port monitoring software like nmap is run
continuously.

PROBLEM CONCLUSION:
The nfs kernel extension does not handle the error code
ECONNABORTED/ECONNRESET returned by soaccept gracefully. It
just restarts nfsd if the error is anything other than
EACCESS.
The fix is to handle the error conditions.

------

APAR: IY37456 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH IN VM_WRITEP USING NOAC NFS MOUNT OPTION

PROBLEM DESCRIPTION:
System crash in vm_writep when using the "noac" mount
option on an NFS filesystem. Stack trace would be
similar to:
 0061B690 vm_writep+000080
 0754C414 nfs.ext:writerp+0001F8
 07553274 nfs.ext:nfs_write+000750
 075406BC nfs.ext:aix2onc_rdwr+000188
 0753FAA4 nfs.ext:aix2onc_rdwr_attr+0000A0
 006110EC vnop_rdwr+0000A0
 006558D4 vno_rw+000074
 005E6DC0 rwuio+0000F0
 005E6FDC rdwr+000130
 005E697C kwrite+000040
 0000379C sc_msr_2_point+000028
 D01DE7CC write+00011C
 100FB09C 100FB09C+000000
 100CB414 as+0006F4
 1002BA90 1002BA90+000000
 1002AD60 1002AD60
 10000218 10000218+000000

PROBLEM CONCLUSION:
Pass sid to vm_handle instead of handle.

------

APAR: IY37463 COMPID: 5765E6100 REL: 510
ABSTRACT: SEGMENTATION FAULT WITH ZH_TW/ZH_TW IM

PROBLEM DESCRIPTION:
The Zh_TW and ZH_TW imput methods can core dump with a special
input string. The stack trace would be similar to:
_aixAuxDraw(??, ??, ??, ??) at 0xd06eca34
IMAuxDraw_2(??) at 0xd2f51614
aux_page_down(??) at 0xd2f4eb40
tim_filter(??, ??, ??) at 0xd2f4f7c0
TIMFilter(??, ??, ??, ??, ??) at 0xd2f50150
IMFilter(??, ??, ??, ??, ??) at 0xd015c4b0
_aixXmbFilter(??, ??, ??, ??) at 0xd06e3430
_XaixKeyPressFilter(??, ??, ??, ??) at 0xd06e29f4
XFilterEvent(??, ??) at 0xd06c3dd8
main(0x1, 0x2ff22268) at 0x10001148

PROBLEM CONCLUSION:
Check input buffer and empty output buffer every time.

------

APAR: IY37468 COMPID: 5765E6100 REL: 510
ABSTRACT: ALT_DISK_INSTALL DOES NOT HANDLE PARALLEL/ROUND ROBIN POLICY

PROBLEM DESCRIPTION:
Any alt_disk_install operations on a rootvg
with logical volumes with SCHED_POLICY set
to parallel/round robin with fail to be
recreated with the same SCHED_POLICY.

LOCAL FIX:
No way to work around it without
a code change.

PROBLEM SUMMARY:
alt_disk_install does not preserve LV attributes
parallel/round robin and parallel/sequential.

PROBLEM CONCLUSION:
alt_disk_install behavior changed to look for and preserve
parallel/round robin and parallel/sequential LV attributes.

------

APAR: IY37480 COMPID: 5765E6100 REL: 510
ABSTRACT: DBX CAUSES SYSTEM CRASH.

PROBLEM DESCRIPTION:
Attempts to use dbx can cause a system crash with a stack trace
like:
   00681560 ptrace_map_text64+000070
   006802A8 ld_ptrace_load64+0003C8
   006821A0 ld_ptracecall+00007C 006E173C ld_proc_load+0007C8
   006DF26C loadx+00008C 0000379C sc_msr_2_point+000028

PROBLEM CONCLUSION:
Fix the system call processor in ptrace to not process system
calls in loader entries that don't have any.

------

APAR: IY37529 COMPID: 5765E6100 REL: 510
ABSTRACT: NET_MALLOC PROBLEMS MANAGING CROSS-MEMORY DESCRIPTORS

PROBLEM DESCRIPTION:
Backporting the changes of IY35553(5.2 apar) to 5.1.
"Network memory does not take advantage of memory
pretranslation on the 32-bit kernel because cross-memory
descriptors are not used correctly. This results in
non-optimal network performance".

PROBLEM SUMMARY:
Backporting the changes of IY35553(5.2 apar) to 5.1.
"Network memory does not take advantage of memory
pretranslation on the 32-bit kernel because cross-memory
descriptors are not used correctly. This results in
non-optimal network performance".

PROBLEM CONCLUSION:
Use cross-memory descriptors correctly so that memory
pretranslation can be done for network memory. This
improves network performance.

------

APAR: IY37544 COMPID: 5765F7000 REL: 810
ABSTRACT: INCORRECT OUTPUT FROM XLF V8 IF STATEMENT

PROBLEM DESCRIPTION:
Incorrect output from XLF V8 IF statement

------

APAR: IY37563 COMPID: 5765E6100 REL: 510
ABSTRACT: YPPASSWD: ROOT CANNOT CHANGE USERS PASSWORD

PROBLEM DESCRIPTION:
On NIS master server, root cannot change a user's
password while the user himself/herself can.
This was noticed while using alternate passwd file.
The error message generated is:
3004-604 Your entry does not match the old password.
yppasswd: 1831-155 could not change passwd

PROBLEM CONCLUSION:
The root password is not being passed properly to the
password change function. It is being saved in one buffer and
a different buffer is being passed to the change function.

------

APAR: IY37566 COMPID: 5765E6100 REL: 510
ABSTRACT: STALE BITS NOT CLEARED BY SYNC OF CONCURRENT VG WITH FACTOR

PROBLEM DESCRIPTION:
Stale bits may not be cleared on passive node when doing sync
of classic (not enhanced) concurrent VG created with factor
greater than 1.

PROBLEM CONCLUSION:
Corrected code to clear proper stale bits on passive node.

------

APAR: IY37571 COMPID: 5765E6100 REL: 510
ABSTRACT: STRIPE WRITE MAY BE MISSED UNDER HEAVY IO

PROBLEM DESCRIPTION:
Write to a stripe may be missed under heavy IO, potentially
causing data inconsistency.

PROBLEM CONCLUSION:
Corrected code to detect conflict during stripe
serialization.

------

APAR: IY37588 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH IN HD_AUDIT WHEN DOING MKVG ON 64-BIT KERNEL

PROBLEM DESCRIPTION:
System crash in hd_audit when doing mkvg on 64-bit kernel
with audit enabled. Stack trace simiar to:
hd_audit+000314
hd_cfg+000100
sc_msr_2_point+000028
lvm_config+000058
lvm_defvg+00015C
lvm_verify+000154
main+000780
__start+000088

PROBLEM SUMMARY:
System crash in hd_audit when doing mkvg on 64bit kernel
with audit enabled.

PROBLEM CONCLUSION:
Corrected code to copy audit structure correctly.

------

APAR: IY37618 COMPID: 5765E6100 REL: 510
ABSTRACT: UDP PROCESS FROM REMOTE M/C. DIES AFTER THE INTERFACE TAKEOVER

PROBLEM DESCRIPTION:
UDP process from the remote m/c. dies after the interface
takeover. In this case it's oracle process (uses UDP) which
was opened from the remote m/c. dies after the interface
takeovear in HACMP environment.

PROBLEM SUMMARY:
UDP process from the remote m/c. dies after the interface
takeover. In this case it's oracle process (uses UDP) which
was opened from the remote m/c. dies after the interface
takeovear in HACMP environment. This is because when the
first interface is down, the UDP packets will go out through
default interface becase the route corresponding to the
interface will be deleted. When the backup is up, these
packets will continue to go out through default interface
because the default route is up and there is a cached route
in protocol control blocks (PCB), which will result in
deleting the connection.

PROBLEM CONCLUSION:
dding a new function in_pcblocalipnotify() to notify all the
UDP PCBs whose local IP matches with the new interface address

TEMPORARY FIX:
Restart the process.

------

APAR: IY37626 COMPID: 5765E6100 REL: 510
ABSTRACT: XMBRESETIC() IS NOT WORKING WELL ON CHINESE LOCALES

PROBLEM DESCRIPTION:
XmbResetIC() returns unconverted preedit string

PROBLEM CONCLUSION:
modify getstring funtion to reset every thing

------

APAR: IY37636 COMPID: 5765E6100 REL: 510
ABSTRACT: MOTIF APPLICATIONS CORE DUMP IN JA_JP/ZH_TW LOCALES

PROBLEM DESCRIPTION:
Motif 2.1 applications may crash in multibyte encoded locales.
Stack trace similar to:
aix_mbstocs
_XomConvert
_XomGenericTextPerCharExtents
_XwcGenericTextPerCharExtents
XwcTextPerCharExtents
CTLRenditionSegPerCharExtents
_XmRenditionTextPerCharExtents
_XmRenditionPosCharExtents
_XmTextPosSegment
TextOut.CursorAtBoundary
TextOut.PaintCursor
DrawInsertionPoint
ProcessReturn
HandleActions
HandleSimpleState
_XtTranslateEvent
XtDispatchEventToWidget
_XtDefaultDispatcher
XtDispatchEvent
XtAppMainLoop
main

PROBLEM CONCLUSION:
 Adjustment to Motif 2.1 logic which branches between
multibyte and widechar codepaths.
Fix made in libXm.a in fileset X11.motif.lib.

------

APAR: IY37666 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH ON JFS2 FTRUNCATE

PROBLEM DESCRIPTION:
JFS2, on a truncate upwards, crashes with a stack trace in
pageOut then a static brkpt.
Stack trace was:
 0007B3C8 .brkpoint+000000 (00000000 ?? )
 00151A00 pageOut+000584 (??, ??, ??, ??)
 00152B30 j2PagerStrategy+0001A0 (??)
 000907CC v_pdtsio_common+000860 (??, ??, ??, ??, ??)
 00090C84 v_pfsiox+000100 (??)
 0008E790 begbt_603_patch_2+000018 (??, ??, ??)
 0008E64C .backt+000000 (0001D4FD, 00000000, 00000020 ?? )
 004921B0 vm_writep+0001A8 (??, ??, ??)
 0014F4D4 pagerReadWriteBehind+000180 (??, ??, ??, ??, ??, ??)
 000E1B1C vm_uiomove+000568 (??, ??, ??, ??)
 004AF91C writeI+00018C (??, ??, ??, ??, ??)
 004AF0FC j2_rdwr+00057C (??, ??, ??, ??, ??, ??, ??, ??)
 00486658 vnop_rdwr+000094 (??, ??, ??, ??, ??, ??, ??, ??)
 0045F030 rwuio+0000D0 (??, ??, ??, ??, ??)
 0045F248 rdwr+00013C (??, ??, ??, ??, ??)
 0045E974 kwrite+0000EC (00000016, 200E9BDC, 00000800)
 00003A94 .sys_call+000000 ()

PROBLEM CONCLUSION:
Fix truncate logic to deal with finalizing RBNA regions

------

APAR: IY37674 COMPID: 5765E6100 REL: 510
ABSTRACT: ICONV BUFFERING PROBLEMS WITH UTF-8 CONVERSIONS

PROBLEM DESCRIPTION:
When iconv() converts data from UTF-8 to UCS-2 and
doesn't have enough room in the output buffer, the
resulting position of the input buffer pointer is
incorrect ( i.e. advanced one character too far ).
As a result, other conversions that depend on this
conversion as an intermediate step, such as
iconv from UTF-8 to just about anything via
Universal_UCS_Conv converter, may produce incorrect
results with certain buffer sizes.

PROBLEM CONCLUSION:
Modify UTF-8 to UCS-2 iconv() converter so that it doesn't
advance the input buffer pointer when the output buffer
is not large enough to contain the result.

------

APAR: IY37696 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR IN XLFHOT WITH -C

PROBLEM DESCRIPTION:
The customer's testcase ICEs in xlfhot when compiled
with -C. The testcase compiles successfully
without -C option.
The ICE is only in XLF 8. XLF 7 is okay.
Here is the testcase:
...
xlf90 -c -C test.f
Program testcase
   character(len=100) fname
   integer iens
   character(len=3) rungen
   open(10,file='infile.ens')
      read(10,*)fname ; print *,'fname= ',trim(fname),'.uf'
      rungen=trim(fname)(1:3)
   close(10)
end program testcase
...
Output is:
1585-100 Internal compiler error -- please contact your service
representative.
1501-511 Compilation failed for file test.f.

LOCAL FIX:
remove -C

PROBLEM CONCLUSION:
The problem has to do with trim(fname).
Development fixed the code and there is no longer an Internal
Compiler Error.

------

APAR: IY37697 COMPID: 5765F7000 REL: 810
ABSTRACT: POINTER ARITHMETIC ERROR IN XLF 8 PRODUCES DIFFERENT RESULTS

PROBLEM DESCRIPTION:
DESCRIPTION : At runtime, the output between the XLF 7.1
and XLF 8.1 compilers is different wrt the customers
testcase and pointer arithmetic. XLF 7.1 has the correct
output, whereas XLF 8.1 does not.
! xlf90 bug.f
program bug
  implicit none
  type opts
     integer method
  end type opts
  type msr
     type(opts) option(0:5)
     type(opts), pointer :: opt
  end type msr
  type(msr), target :: system(32)
  type(msr), pointer :: sys
  sys => system (1)
  sys%opt => sys%option (1)
  system(1)%option(1)%method = 4
  write (*,*) 'sys%opt%method = ' , sys%opt%method
  write (*,*) 'system(1)%option(1)%method = ',
system(1)%option(1)%method
end program bug
The failing output is:
sys%opt%method = 0
system(1)%option(1)%method = 4
The expected output is:
sys%opt%method = 4
system(1)%option(1)%method = 4

PROBLEM CONCLUSION:
Development fixed the APAR and it will be part of the 8.1.0.04
PTF.

------

APAR: IY37715 COMPID: 5765E6100 REL: 510
ABSTRACT: CHECKPOINT OF RESTARTED PROCESSES ERRNO=ESRCH SY ERROR=0X000D

PROBLEM DESCRIPTION:
Checkpoint of restarted processes fails with errno=ESRCH
and Sy Error=0x000D(KCHKPT_NON_CHKPNTABLE) even though all
processes being checkpointed have the CHECKPOINT=yes
environment variable set.

PROBLEM CONCLUSION:
The zombie ghost process leader created during restart must
be created with the checkpointable flag.

------

APAR: IY37747 COMPID: 5765E6100 REL: 510
ABSTRACT: DRAW_SEPARATOR SEGV IN XMIM.C

PROBLEM DESCRIPTION:
If an applet in the Java appletviewer is quickly and
repeatedly reloaded or restarted, the appletviewer may
cause a segmentation violation.

PROBLEM CONCLUSION:
Motif code fix.

------

APAR: IY37777 COMPID: 5765F7000 REL: 810
ABSTRACT: ICE IN ASTI WHEN USING -C

PROBLEM DESCRIPTION:
ICE in ASTI when using -C

------

APAR: IY37780 COMPID: 5765E6100 REL: 510
ABSTRACT: BOS.SYSMGT.NIM.MASTER GUI STILL AROUND AFTER ALT_DISK_MIGRATION

PROBLEM DESCRIPTION:
When using alt_disk_migration (nimadm) to
migrate a system from AIX V4.3 to AIX V5.1, the
fileset bos.sysmgt.nim.master_gui is left on
the system when it should be removed.

LOCAL FIX:
Remove the fileset from the system using
"smitty remove"

PROBLEM SUMMARY:
After using nimadm to migrate from 5.1 or 5.2, some obsolete
filesets are still installed on the system.

PROBLEM CONCLUSION:
Updated nimadm to remove the obsolete filesets during the
migration.

------

APAR: IY37797 COMPID: 5765E6100 REL: 510
ABSTRACT: UNSUPPORTED RESOLUTION SETTING DURING INSTALL

PROBLEM DESCRIPTION:
Unsupported monitor resolution is set
during install of AIX when a 7316 monitor
is used with avocent KVM switch. This
results in a black screen on monitor
during install.

LOCAL FIX:
Bypass switch, directly connect 7316 to system

PROBLEM CONCLUSION:
Recognize the EDID monitor string used by avocent switch and
set default to 1024x768.

------

APAR: IY37799 COMPID: 5765F7000 REL: 810
ABSTRACT: ICE:EXCEPTION TERMINATES PROGRAM

PROBLEM DESCRIPTION:
ICE:Exception terminates program

LOCAL FIX:
Use -O2 instead of -O3.

------

APAR: IY37812 COMPID: 5765E6100 REL: 510
ABSTRACT: INCREMENTAL BACKUP FAILS ON JFS2

PROBLEM DESCRIPTION:
backbyinode on jfs2 filesystems fails with incremantal
backups reporting the following internal errors.
1) backup: 0511-202 internal error recieved a
        segmentation violation
2) backup: 0507-558 Inode 12326 has stale b-tree page
        at block 141026
        An error occurred when reading data from the device.
        backup: 0511-203 Internal error; exit status 1 is
        not recognized.
The command being executed is as given below :
"backup -0 -u -f /dev/rmt0.1 /jfs2fs"

PROBLEM CONCLUSION:
The problem has been fixed by making necessary
code changes.

------

APAR: IY37840 COMPID: 5765B9501 REL: 330
ABSTRACT: HSM HANGING IN DSMRECALLD

PROBLEM DESCRIPTION:
Fix case where some events are on the queue, but a dmGetEvents
call get no events returned. This fix should post the
EV_INACTIVE events that are sitting on queue.

PROBLEM SUMMARY:
HSM recall hanging in dsmrecalld

PROBLEM CONCLUSION:
A summary bit that indicated where there were events on the
queue was not being kept up to date correctly. Change to
code to use an existing counter of the number of events that
have not been given to the DMAPI events handler. Wake up any
event waiters whenever the counter is positive instead of
when there is just 1 event on the queue.

------

APAR: IY37852 COMPID: 5765F7000 REL: 810
ABSTRACT: COREDUMP WITH RESHAPE AND -C

PROBLEM DESCRIPTION:
Coredump with reshape and -C

LOCAL FIX:
Compile without using -C.

------

APAR: IY37853 COMPID: 5765F7000 REL: 810
ABSTRACT: INCORRECT OUTPUT WHEN COMPILING WITH 8.1

PROBLEM DESCRIPTION:
Incorrect output when compiling with 8.1

------

APAR: IY37869 COMPID: 5765E6100 REL: 510
ABSTRACT: POSSIBLE SYSTEM HANG WHEN ATM FIBER RECONNECTED TO THE SWITCH

PROBLEM DESCRIPTION:
After disconnecting and reconnecting the ATM fiber or
rebooting the ATM switch the system may hang.

------

APAR: IY37870 COMPID: 5765E6100 REL: 510
ABSTRACT: ATMSVCD NEEDS TO IGNORE ILMI ENTERPRISE TRAPS FROM SWITCHES.

PROBLEM DESCRIPTION:
Atmsvcd restarts its attachment to the ATM network everytime
it receives an ILMI Enterprise Trap from the network.

PROBLEM CONCLUSION:
The only trap atmsvcd needs to take action on is a Cold Start.
Changing atmsvcd to ignore all ILMI traps except Cold Starts.

------

APAR: IY37871 COMPID: 5765E6100 REL: 510
ABSTRACT: ATMSVCD NEEDS TO HANDLE INVALID "NO SUCH NAME" RESPONSES.

PROBLEM DESCRIPTION:
atmsvcd's ILMI will never startup when it receives an
ILMI response of "No Such Name" for the SysUpTime MIB.
This is an invalid response which needs to be fixed on the
switch but atmsvcd needs to handle this invalid response
as well to protect itself.

PROBLEM CONCLUSION:
Changed atmsvcd to accept a response "No Such Name" for the
SysUpTime MIB as verification of the LINK but store that
the response was invalid.

------

APAR: IY37872 COMPID: 5765E6100 REL: 510
ABSTRACT: POSSIBLE SYSTEM HANG IF ATMCM AND ATMSVCD LOOP REGISTERING.

PROBLEM DESCRIPTION:
When a registration packet gets lost on the network ATMCM
and atmsvcd can get into a loop where ATMCM registers and
atmsvcd returns an error that it is registering. This
can lead ATMCM and atmsvcd to get into a loop which can
cause a system hang.

PROBLEM CONCLUSION:
Change atmsvcd to NOT notify ATMCM that the address is
already being registered. ATMCM only wants to know
if the address registration was successful or not.

------

APAR: IY37873 COMPID: 5765E6100 REL: 510
ABSTRACT: CDIAGATT GROWS LARGE: DIAGD CONSUMES CPU

PROBLEM DESCRIPTION:
CDiagAtt and CDiagAtt.vc continue to grow quite large. The
"/" filesystem becomes full. diagd consumes an abnormally
large percentage of the processor time.
odmget -q"attribute=diag_reminder" CDiagAtt displays
thousands of duplicate stanzas like the following:
CDiagAtt:
        name = ""
        attribute = "diag_reminder"
        value = "9999"
        type = "T"
        rep = "s"

LOCAL FIX:
Kill diagd.
odmdelete -q "attribute=diag_reminder" -o CDiagAtt
cp /usr/lpp/diagnostics/obj/CDiagAtt /etc/objrepos
cp /usr/lpp/diagnostics/obj/CDiagAtt.vc /etc/objrepos
set the permissions of the object classes to 644.
Restart diagd.

PROBLEM SUMMARY:
CDiagAtt and CDiagAtt.vc continue to grow quite large.
The "/" filesystem becomes full. diagd consumes an
abnormally large percentage of the processor time,
because ODM queries to CDiagAtt take much longer.

PROBLEM CONCLUSION:
Correct /usr/lpp/diagnostics/bin/diagd in bos.diag.rte to
remove the duplicate stanzas if they already exist on the
system and prevent the duplicate stanzas from being created.

------

APAR: IY37877 COMPID: 5765E6100 REL: 510
ABSTRACT: ALT_DISK_INSTALL ERRORS 0505-102 AND 0516-1208 SEEN

PROBLEM DESCRIPTION:
During an alt_disk_install operation with the
LANG variable set to languages other than en_US
or C, the following errors could be seen:
0505-102 alt_disk_install: mkvg has returned an error.
0516-1208 mkvg: Warning, The Physical Partition Size
of <PP's> requires the creation of <PP's>
partitions for hdisk#. The system limitation is
#### physical partitions per disk at a factor
value of #. Specify a larger Physical Partition
Size or a larger factor value in order to
create a volume group on this disk.
0516-862 mkvg: Unable to create volume group.

LOCAL FIX:
Use LANG=C or en_US to avoid the problem.

PROBLEM SUMMARY:
alt_disk_install may fail in some locales with the following
errors: "Invalid range within expression." This message
will appear in the designated language.
The language dependant code has to be modified.

PROBLEM CONCLUSION:
Changed alt_disk_install to use LANG=C when evaluating
regular expression ranges.

TEMPORARY FIX:
None without code change.

------

APAR: IY37882 COMPID: 5765E6100 REL: 510
ABSTRACT: LSSRC & DADMIN SHOW INCORRECT LEASE AFTER REFRESH OF DHCPSD

PROBLEM DESCRIPTION:
When the customer changes the leasetimedefault or option 51 in
the /etc/dhcpsd.cnf and refreshes the dhcpsd, the database
built by the server doesnot have the correct lease time
duration for all the addresses which have already been leased
by the server (ie the client info is from /etc/db_file.cr).

LOCAL FIX:
As the server is working as per design but the lssrc
and dadmin tools give out the wrong lease duration
, so customer can ignore the lease duration in
this scenario.

PROBLEM SUMMARY:
This apar was opened to fix the problem
with lssrc & dadmin showing wrong
lease time for an IP address when the
IP address has already been leased to
a client and then the system admin goes
and changes the lease time in /etc/dhcpsd.cnf
either by using option 51 or leasetimedefault.
In this case the dhcpsd server should retain
the old lease time as there is no mechanism
for server to inform client about the change
in lease time. Our AIX server is performing
as designed but when we look at the leased
addresses using lssrc or dadmin the lease time
used to be latest as per the /etc/dhcpsd.cnf.

PROBLEM CONCLUSION:
Both lssrc and dadmin should show the correct
lease time.

TEMPORARY FIX:
As this doesnot affect the server, this should
not cause any customer issues. But if the
system admin want more info he can look at
/etc/db_file.cr and look for the start time
and expire time. If wants to know the lease
time then (lease time = expiretime - starttime).

------

APAR: IY37886 COMPID: 5765E6100 REL: 510
ABSTRACT: SECURITY: BUFFER OVERFLOW IN XFS

PROBLEM DESCRIPTION:
There are a number of places in xfs where invalid requests
sent over the network could cause the font server to crash
because of buffer overflows. There is no known exploit, but
in theory buffer overflows could be used to execute arbitrary
code included by a hacker.

PROBLEM CONCLUSION:
Add a number of checks to detect invalid font server requests
and return errors.

------

APAR: IY37937 COMPID: 5765E6100 REL: 510
ABSTRACT: OPENGL APPLICATION HANGS WHEN USING MULTIPLE TEXTURES

PROBLEM DESCRIPTION:
An OpenGL based application called VE0TEXTURE hangs on
GXT6500P graphics adapter when mutiple textures are
being used simultaneously.

PROBLEM CONCLUSION:
Handle a case where not enough memory can be made available,
by kicking out an in use texture.

------

APAR: IY37944 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH IN HD_FREE_LVOLS

PROBLEM DESCRIPTION:
System crash in hd_free_lvols().

PROBLEM SUMMARY:
System crash in hd_free_lvols() with lv in deleting state.

PROBLEM CONCLUSION:
Corrected code to allow lv in deleting state.

------

APAR: IY37989 COMPID: 5765E6100 REL: 510
ABSTRACT: SMITTY NIM_CHNET PRODUCES ERROR ON JA_JP LANG

PROBLEM DESCRIPTION:
An error (1800-083) will be reported by smitty
when using the "smitty nim_chnet" fastpath and the
LANG is set to Ja_JP. The error seen is the following:
    1800 083 Data error: could not find matching ring
    field value for corresponding Command_to_Discover
    value. sm_cmd_opt 2 .disc_field_name:
    "ent_type".
    The current actual value will be used as the
    default value for the following dialog. Use
    local problem reporting procedures.
One of the SMIT ODM stanzas has the wrong catalog name
and causes one of the values, specifically "ent_type" to
be loaded with the incorrect data.

LOCAL FIX:
Executing the same function "smitty nim_chnet" with
the LANG set to "C" or "en_US".

PROBLEM SUMMARY:
When running "smitty nim_chnet" on a NIM
ethernet network in a language other than
English (i.e. Ja_JP), an error message will
occur:
1800 083 Data error: could not find matching ring
    field value for corresponding Command_to_Discover
    value. sm_cmd_opt 2 .disc_field_name:
    "ent_type".
    The current actual value will be used as the
    default value for the following dialog. Use
    local problem reporting procedures.
The problem was caused by an ODM stanza pointing to
the wrong catalog.

PROBLEM CONCLUSION:
The code will be changed to modify the
nim_chnet_opt_ieee stanza in sm_nim_master.add
to point to the cmdnim.cat catalog instead of the
smit.cat catalog.

TEMPORARY FIX:
Change the LANG to "C" or "en_US" or use
the command line to change the network object.

------

APAR: IY37991 COMPID: 5765E6100 REL: 510
ABSTRACT: 10/100 ETHERNET PCI ADAPTER II DRIVER NO MP-SAFE

PROBLEM DESCRIPTION:
The 10/100 Ethernet PCI Adapter II device driver does not
call tstop() in a while loop, making it not multi-processor
safe.

PROBLEM CONCLUSION:
Change the way the device driver calls tstop() and tstart() to
make it MP-safe.

------

APAR: IY38007 COMPID: 5765E6100 REL: 510
ABSTRACT: VARONVG -CN FAILS WITH DUPLICATE LV NAME

PROBLEM DESCRIPTION:
varyonvg -cn fails when a lv name is deleted and readded with
a different lvid.

PROBLEM CONCLUSION:
Corrected synclvodm to handle this case.

------

APAR: IY38015 COMPID: 5765E6100 REL: 510
ABSTRACT: LIST WIDGET MAY RESIZE UNEXPECTEDLY

PROBLEM DESCRIPTION:
The list widget within a file selection dialog may resize
itself and become larger than expected.

PROBLEM CONCLUSION:
Motify changes to the list widget introduced in Motif 2.1.30.

------

APAR: IY38022 COMPID: 5765E6100 REL: 510
ABSTRACT: MKLVCOPY RUNS OUT OF ENVIRONMENT SPACE WITH LARGE LV

PROBLEM DESCRIPTION:
mklvcopy runs out of environment space working on large LV.

PROBLEM CONCLUSION:
Write to file instead of using environment.

------

APAR: IY38046 COMPID: 5765F7000 REL: 810
ABSTRACT: INCORRECT OUTPUT FROM SIZE INTRINSIC

PROBLEM DESCRIPTION:
Incorrect output from size intrinsic

------

APAR: IY38065 COMPID: 5765E6100 REL: 510
ABSTRACT: CANNOT ADD ROUTE IF THE GATEWAY IS NOT ON THE SAME NETWORK

PROBLEM DESCRIPTION:
Unable to add route, if the gateway is not on the same network.

PROBLEM CONCLUSION:
Removed the check for the gateway.

------

APAR: IY38117 COMPID: 5765E6100 REL: 510
ABSTRACT: GETROOTFS BROKEN WITH LARGE CAPACITY DISKS

PROBLEM DESCRIPTION:
When booting from a non-disk device and selecting the option
for accessing a root volume group, systems with large
capacity disks (greater than 100 GB) will see an error from
lqueryvg, and the process of accessing a rootvg will fail.

PROBLEM CONCLUSION:
add white space to separate the different fields of disk info

------

APAR: IY38161 COMPID: 5765B9501 REL: 330
ABSTRACT: DMCREATESESSION DEADLOCK

PROBLEM DESCRIPTION:
DM create session hangs when DM Set Disposition called at same
time.

PROBLEM SUMMARY:
DMCREATESESSION deadlock

PROBLEM CONCLUSION:
DM create session hangs when DM Set Disposition called at
same time. releseAccess must be called before
dmSessMgrMutex.acquire so that ccMgrMutex/dmSessMgrMutex
locking order is not violated.

------

APAR: IY38166 COMPID: 5765E6100 REL: 510
ABSTRACT: SCSI HOT SWAP MANAGER DOES NOT DISPLAY ALL SCSI ENCLOSURES

PROBLEM DESCRIPTION:
SCSI Hot Swap Manager:
  1. Displays *Catalog Error* as one of its devices.
  2. Some Enclosure and its devices are missing from
     the device selection menu; therefore, the user
     will not able to perform SCSI Hot Swap Operations.

PROBLEM CONCLUSION:
Correct /usr/lpp/diagnostics/bin/uscsicfg in
bos.diag.util.

------

APAR: IY38201 COMPID: 5765E6100 REL: 510
ABSTRACT: MEMEORY LEAK IN CSH

PROBLEM DESCRIPTION:
csh dumps core when DEBUGMALLOC is set, and the 'eval' command
is executed.

PROBLEM SUMMARY:
DEBUG MALLOC causes csh to dump core. The SIGSEGV is caused
because a memory is being freed twice.

PROBLEM CONCLUSION:
Memory leak observed in csh, since setting DEBUG MALLOC
causes csh to dump core.

------

APAR: IY38208 COMPID: 5765B9501 REL: 330
ABSTRACT: HSM DEADLOCK IN GNTYPE

PROBLEM DESCRIPTION:
Hang in gnType waiting for dmShared lock. gnType should be
treated like a stat call which does not get a DM lock.

PROBLEM SUMMARY:
HSM deadlock in gnType

PROBLEM CONCLUSION:
Hang in gnType waiting for dmShared lock holding up a
mmfs_dm_rdwr_invis call waiting for the gnode to get out of
transition state. gnType should be treated like a stat call
which does not get a DM lock.

------

APAR: IY38215 COMPID: 5765E6100 REL: 510
ABSTRACT: SEGFAULT IN DESKTOP.DESTROY()

PROBLEM DESCRIPTION:
A segfault may occur when destroying an XmDesktopObject.

PROBLEM CONCLUSION:
Motif code fix.

------

APAR: IY38217 COMPID: 5765E6100 REL: 510
ABSTRACT: INSTALLATIONS FROM FLOPPY THROUGH SMITTY COULD FAIL

PROBLEM DESCRIPTION:
When installing from a diskette through
smitty and the F4 listing is used to
select the filesets, the installation
could fail reporting that no filesets were
found on the media.

LOCAL FIX:
To work around this problem use the keyword "all"
in the "SOFTWARE to install" field. If this is not
possible then use geninstall or installp from the
command line.

PROBLEM SUMMARY:
When using a floppy or tape device to install
using geninstall (or SMIT), it will fail if the
install list or bundle contains prefixed entries.
The error message will state that the filesets are not
found on the media, or that the media is invalid.
Having prefixes in front of the filesets to
install is handled properly for CD installs but
tape and floppy installs did not handle it.

PROBLEM CONCLUSION:
The code has to change to remove prefixes from
install lists when the media is present on
tape of floppy.

TEMPORARY FIX:
none without modifying the code.

------

APAR: IY38223 COMPID: 5765E6100 REL: 510
ABSTRACT: DIRECT I/O BEING DEMOTED

PROBLEM DESCRIPTION:
After an unaligned read, direct i/o is no
longer attempted for the file.

------

APAR: IY38226 COMPID: 5765E6100 REL: 510
ABSTRACT: FAILURE TO UPDATE AIO_ERRNO AND AIO_RETURN FIELD PROPERLY

PROBLEM DESCRIPTION:
Sybase may fail with sddone 823 errors indicating a
size mismatch. Generally this will cause Sybase to
crash. This problem only occurs on p690's (and related
systems).

PROBLEM SUMMARY:
Sybase may fail on p690 with sddone 823 (size mismatch)
error.

PROBLEM CONCLUSION:
Changed rw_request and arl_iodone code to run LWSYNC
after updating aio_return field, and before updating
aio_errno field.

------

APAR: IY38248 COMPID: 5765E6100 REL: 510
ABSTRACT: TX TIMEOUT ERRORS WHEN RUNNING WITH LARGE-SEND ENABLE

PROBLEM DESCRIPTION:
Depends on the workload, TX timeout errors can occur when
unning with Large-send enabled on the 10/100 Ethernet Adapter
II. The device driver is able to recover from the error.

PROBLEM CONCLUSION:
Fix the driver code so that it gets and dma-map additional
buffer properly in the case that the driver larg-send buffer
pool is used up.

------

APAR: IY38264 COMPID: 5765B9501 REL: 330
ABSTRACT: FCNTL LOCK REVOKE WORKER HUNG

PROBLEM DESCRIPTION:
fcntl lock revoke worker hung.
One node is holding a byte range lock that it already has.
Stopping GPFS on this node will cleanup the waiters.
This information is from PMR 83743,999,866 (R340). However
there is no information how the determine which node has
the problem. Refer to IY35958 and defect 386912.

PROBLEM SUMMARY:
fcntl lock revoke worker hung

PROBLEM CONCLUSION:
Add missing ctTransitionWake after ctState marked CL_STABLE.
Dump tm sequence number in RecLock pending revoke string.

------

APAR: IY38288 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH IN V_PFREFREEDQ

PROBLEM DESCRIPTION:
Customer system can crash while executing the "v_prefreedq"
function. The stack for above crash will be as shown below
 00095318 .panic_trap+000000 ()
 00095314 .panic+000068 ()
 00138948 v_prefreedq+000080 (??, ??, ??)
r28 : F100009E15A43838 r27 : 0000000000000001
 00327A94 .backt+00007C ()

PROBLEM CONCLUSION:
"v_prefreedq" function was trying to unlock "pdtx" without
taking lock on it, kernel code modified to fix the problem.

------

APAR: IY38293 COMPID: 5765E6100 REL: 510
ABSTRACT: "CHLV CAUSING LVM DRIVER TO STOP RETURNING NEW DEV SUBTYPE."

PROBLEM DESCRIPTION:
After creating a logical volume with the "mklv -T O"
option the new device subtype will not returned after
a subsequent chlv command is issued.
A varyoffvg/varyonvg sequence will cause the driver
to reset the logical volume to return the new device subtype.

PROBLEM CONCLUSION:
Prevent chlv from overwriting the LVM driver indicator
that specifies that the new subtype is to be returned.

------

APAR: IY38294 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH IN DISPATCH

PROBLEM DESCRIPTION:
After starting HTX, system crashes within a few
hours. All the dumps observed after this crash
fail with a -3 error code.

PROBLEM CONCLUSION:
The argument passed to trace hook dereferences a freed
thread pointer.

------

APAR: IY38328 COMPID: 5765B9501 REL: 330
ABSTRACT: SOME SIGNALS ARE MASKED AT PROECESSES WHICH ARE STARTED FROM

PROBLEM DESCRIPTION:
The following signals are masked at processes which are started
from gpfsready script.
SIGHUP SIGINT SIGQUIT SIGPIPE SIGALRM SIGCHLD SIGIO SIGUSR2
SIGDANGER.
And some important signals are ignored also.(ex. SIGINT,SIGQUIT)

PROBLEM SUMMARY:
Signals masked when calling external scripts

PROBLEM CONCLUSION:
In the daemon, certain signals are masked off. When shell
scripts are executed by the daemon, the signal handlers get
reset, but the signal mask persists. This may have
undesirable consequences, as in this defect:
mmfs_varyonvg.scr tries to verify that mmsdrfs is current,
and in the course of processing has to run a command on a
different node, which is done through mmdsh, which issues
ping to verify that the node is alive. If the node is dead,
ping with hang forever because SIGALRM is masked off. Reset
the signal mask before exec'ing shell scripts from the
daemon.

------

APAR: IY38329 COMPID: 5765B9501 REL: 330
ABSTRACT: SOME SIGNALS ARE MASKED AT PROECESSES WHICH ARE STARTED FROM

PROBLEM DESCRIPTION:
The following signals are masked at processes which are started
from gpfsready script.
SIGHUP SIGINT SIGQUIT SIGPIPE SIGALRM SIGCHLD SIGIO SIGUSR2
SIGDANGER.
And some important signals are ignored also.(ex. SIGINT,SIGQUIT)

PROBLEM SUMMARY:
Signals masked when calling external scripts

PROBLEM CONCLUSION:
In the daemon, certain signals are masked off. When shell
scripts are executed by the daemon, the signal handlers get
reset, but the signal mask persists. This may have
undesirable consequences, as in this defect:
mmfs_varyonvg.scr tries to verify that mmsdrfs is current,
and in the course of processing has to run a command on a
different node, which is done through mmdsh, which issues
ping to verify that the node is alive. If the node is dead,
ping with hang forever because SIGALRM is masked off. Reset
the signal mask before exec'ing shell scripts from the
daemon.

------

APAR: IY38371 COMPID: 5765E6100 REL: 510
ABSTRACT: CHECKPOINT/RESTART: LL CMD CORE'G UPON RESTART

PROBLEM DESCRIPTION:
On the 64-bit kernel, a process that is forked by a restarted
process and then exec's a multi-threaded will coredump inside
pthread_init().

PROBLEM CONCLUSION:
Replace the call to getpid() with the call to getrpid() in
pthread_init().

------

APAR: IY38420 COMPID: 5765F7000 REL: 810
ABSTRACT: ICE IN XLFCODE (TOBEY)

PROBLEM DESCRIPTION:
!*************************************************************
!*
!* DEFECT ABSTRACT : ICE in xlfcode (tobey)
!* CUSTOMER : UCAR- Julie Rew
!* PMR NUMBER : 71346,033,000
!* COMPILER INVOCATION : xlf90
!* COMPILER OPTION(S) : none
!* OS OPTION(S) : none
!* OS LEVEL : AIX 5L
!* COMPILER LEVEL : XLF 8.1.0.2 (latest
!* MACHINE TYPE/MODEL :
!*
!* DESCRIPTION :
!* Customer hits an ICE when compiling test case.
!* They first noticed the problem with an eFix we had
!* sent to them. But upon further testing, I found that
!* it occurs on the latest dev build as well. Please consider
!* the following:
!*
!*
!* $ /orbit/xlf.v8r1/deliver/bin/xlf90 -F/orbit/xlf.v8r1/ >
!* deliver/xlf.cfg51 upper>
!* ** main === End of Compilation 1 ===
!* /orbit/xlf.v8r1/deliver/bin/xlf90: 1501-230 Internal
!* compiler error; please contact your Service Representative
!* 1501-511 Compilation failed for file uppercase.f.
!* $ /orbit/xlf.v8r1/deliver/bin/xlf90 -v -qdebug=except
!* -F/orbit/xlf.v8r1/deliv>
!* exec: /orbit/xlf.v8r1/deliver/bin/xlfentry(/orbit/
!* xlf.v8r1/deliver/bin/xlfentry,uppercase.f,/tmp/F8FH29ab,
!* /tmp/F8FH29abF.lst,xlfsmsg.cat,xlfmsg.cat,uppercase.f,32,
!* XLF90(noautodealloc,nosignedzero),FREE(f90),NOLM,
!* DEBUG(except),WSTREAMS(/tmp/F8FH29aah1,/tmp/F8FH29aab1,
!* /tmp/F8FH29aas1),DEFMSG(/orbit/xlf.v8r1/deliver/bin/
!* default_msg),DEBUG(NTRANS,ASTI),-I/orbit/xlf.v8r1/deliver/
!* include,NULL)
!* ** main === End of Compilation 1 ===
!* exec: /orbit/xlf.v8r1/deliver/bin/xlfhot(/orbit/xlf.v8r1/
!* deliver/bin/xlfhot,/tmp/F8FH29aah1,/tmp/F8FH29aah2,
!* /tmp/F8FH29aab1,/tmp/F8FH29aab2,/tmp/F8FH29aas1,
!* /tmp/F8FH29aas2,/tmp/F8FH29ab,/tmp/F8FH29abA.lst,
!* -qdebug=except,NULL)
!* exec: /orbit/xlf.v8r1/deliver/bin/xlfcode
!* (/orbit/xlf.v8r1/deliver/bin/xlfcode,-qdebug=except,
!* /tmp/F8FH29aah2,/tmp/F8FH29aab2,uppercase.o,
!* /tmp/F8FH29abB.lst,/tmp/F8FH29aas2,NULL)
!* Invalid scope difference -1 for symbol 20 (#1)
!* Symbol owner is 14 (upper_case)
!* Current function is 1 (main)
!* EXCEPTION terminates program: no handler for exception.
!* Exception of type: TRAP at 10117F88. Regs:
!* 10118648 2FF21C70 3029CB94 FFFFFFFF F0044490 00000008
0000F0B2 00000000
!* 10006FBD 10006FBD 00000000 F012B364 10118640 DEADBEEF
00000007 2FF21F40
!* 2FF21F60 A00031C6 00010B0C 00000008 00000002 00000004
300181F0 30018FE4
!* 30015FB0 00000005 00000000 00100000 47050670 00000000
00000000 576B97D8
!* resume: cr: lr: ctr: xer: fpscr: msr:
mq:
!* 10117F8C 62422444 10118648 D0062F00 20000000 00002000
0000D0B2 00000000
!* Trap condition: -1 ? 0 arithmetic compare
!* Trap instruction: TI 20,R03,0
!* Traceback:
!* Line ? Disp 00000004
Dsa: 2FF21C70
!* Line ? Disp 00000160
Dsa: 2FF21CF0
!* Line ? Disp 00000508
Dsa: 2FF21D60
!* Line ? Disp 00000128
Dsa: 2FF21DE0
!* Line ? Disp 00000120
Dsa: 2FF21E70
!* --- End of call chain ---
!* 1501-511 Compilation failed for file uppercase.f.
!* unlink: uppercase.o
!* $
!*
!*
!*************************************************************
!
! TEST CASE: uppercase.f
program main
  implicit none
  select case(upper_case("everest"))
  end select
  contains
  function upper_case(str) result(ustr)
    implicit none
  character(len=*), intent(in) :: str
  character(len=len(str)) :: ustr
    integer i
    integer,parameter :: il2u=ichar('A')-ichar('a')
    ustr=str
    do i=1,len_trim(str)
      if(str(i:i).ge.'a'.and.str(i:i).le.'z') &
        ustr(i:i)=char(ichar(str(i:i))+il2u)
    end do
  end function upper_case
end program main
!
! -----------------------------------------------------------
!
! OUTPUT RECEIVED USING 8.1.0.03:
!
! $ /orbit/xlf.v8r1/deliver/bin/xlf90 -F/orbit/xlf.v8r1/ \
! deliver/xlf.cfg51 uppercase.f
! ** main === End of Compilation 1 ===
! /orbit/xlf.v8r1/deliver/bin/xlf90: 1501-230 Internal
! compiler error; please contact your Service Representative
! 1501-511 Compilation failed for file uppercase.f.
!
!
!
! -----------------------------------------------------------
!
! EXPECTED OUTPUT:
! $ /orbit/xlf.v8r1/deliver/bin/xlf90 -F/orbit/xlf.v8r1/ \
! deliver/xlf.cfg51 uppercase.f
! ** main === End of Compilation 1 ===
! 1501-510 Compilation successful for file uppercase.f.
!
! ---------------------------- EOF --------------------------

------

APAR: IY38423 COMPID: 5765E6100 REL: 510
ABSTRACT: CRASH IN XTINSERT USING JFS2 FILESYSTEMS

PROBLEM DESCRIPTION:
Crash using JFS2 filesystems with stack trace of:
 00042DFC .brkpoint+000000 (0000000000000000 ?? )
 00157388 xtInsert+000108 (??, ??, ??, ??, ??, ??, ??)
 0014D12C pagerAllocateExtent+000138 (??, ??, ??, ??, ??)
 0014FB90 pageProtect+000114 (??, ??, ??, ??)
 00150E6C j2PagerService+000148 (??)
 0014E148 j2PagerThread+000080 (??)

PROBLEM SUMMARY:
crash in xtinsert

PROBLEM CONCLUSION:
pageProtect should check for RBNA when determining if a page
is backed.
Fix made in the kernel.

------

APAR: IY38430 COMPID: 5765E6100 REL: 510
ABSTRACT: WRITE CMD FAILS WITH E_BUSY STATUS, ON STK TAPE DRIVES

PROBLEM DESCRIPTION:
mksysb backup command fails with BUSY status, in SAN
configuration, where device is shared between multiple
hosts.

PROBLEM CONCLUSION:
Retry command on SCSI_BUSY status.

------

APAR: IY38434 COMPID: 5765E6100 REL: 510
ABSTRACT: SECURITY: INTEGER OVERFLOW IN XDRMEM_GETBYTES

PROBLEM DESCRIPTION:
Malicious data could cause programs depending on libnsl to
crash

PROBLEM CONCLUSION:
Change libnsl to correctly use unsigned integers for sizes

------

APAR: IY38440 COMPID: 5765E6100 REL: 510
ABSTRACT: LIBNCK.A FAILS ON AIX 5.1 AND 5.2

PROBLEM DESCRIPTION:
An executable that links to the library
libnck.a will fail to load in AIX 5.1
with the following error:
The program which calls the following library fails.
Library : /usr/lib/libnck.a
<<Error message>>
0509-036 Cannot load program
0509-130 Symbol resolution failed
0509-136 Symbol connect (***) is not
exported from dependent.
0509-192 Examine .loader section
symbols with the"dump -T"

LOCAL FIX:
none without code change.

PROBLEM SUMMARY:
A change in a kernel symbol from
select to _select caused the library
/usr/lib/libnck.a to not function.
This causes problems for all the code
that references this library.

PROBLEM CONCLUSION:
To fix this problem the library was modified
to reference the right symbol.

------

APAR: IY38454 COMPID: 5765E6100 REL: 510
ABSTRACT: KERNEL KDB CRASHES SYSTEM

PROBLEM DESCRIPTION:
User enters the debugger by typing control-. Upon return
from the debugger, the system will assert with the following
stack:
 0000932C .unlock_enable_mem+000020 ()
 04594268 vcon_offl+0004C0 (??)
 0005E9A4 i_offlevel+0000D4 (??, ??)
 0005E5DC i_softmod+0002DC ()
 000A1CD4 .finish_interrupt+0000BC (??, ??, ??, ??, ??, ??)

PROBLEM CONCLUSION:
A lock was released before calling the debugger which should
not have been released.

------

APAR: IY38469 COMPID: 5765E6100 REL: 510
ABSTRACT: PCI 10/100 ETHERNET II REQUIRES TIMES TO CHANGE ITS CU STATE.

PROBLEM DESCRIPTION:
Customer does the TSM backup through the PCI 10/100
Ethernet II adapter that causes a lot adapter errors
logged in the AIX error log.
The TSM backsup software keeps asking the ethernet
adapter about the statistics every 150ms while the adapter
transmit packets. The adapter has to constantly switching mode
between transmitting packets and gathering statistics and
it could cause the state machine not changing fast enough.

PROBLEM CONCLUSION:
Give the adapter more times to update it state machine
before call out the error.

------

APAR: IY38471 COMPID: 5765E6100 REL: 510
ABSTRACT: CHANGE LIBC TCP RPC TO USE NON-BLOCKING I/O

PROBLEM DESCRIPTION:
Any programs that use the libc rpc implementation with tcp
are vulnerable to being starved by a program that sends data
very slowly

PROBLEM CONCLUSION:
Change the libc rpc implementation to use non-blocking I/O
for tcp

------

APAR: IY38487 COMPID: 5765E6100 REL: 510
ABSTRACT: NIS+ GETSERVBYNAME W/NULL PROTO

PROBLEM DESCRIPTION:
getservbyname with NIS+ fails if proto is NULL

PROBLEM CONCLUSION:
Added NIS+ call to deal with NULL proto case appropriately.

------

APAR: IY38489 COMPID: 5765E6100 REL: 510
ABSTRACT: RPC.RSTATD NOT REPORTING SYSTEM TIME

PROBLEM DESCRIPTION:
System time is shown as 0.

PROBLEM CONCLUSION:
Set system time in structure.

------

APAR: IY38506 COMPID: 5765E6100 REL: 510
ABSTRACT: SU DOESN'T LOG AN AUDIT RECORD FOR A NON-EXISTANT USER

PROBLEM DESCRIPTION:
The SU command doesn't produce an audit
record for a non existant user. For an
existant user, it records the event.

PROBLEM CONCLUSION:
Call the audit() when the su fails for the user.

------

APAR: IY38548 COMPID: 5765E6100 REL: 510
ABSTRACT: COMMAND TIMEOUTS - FIBRE CHANNEL ADAPTER

PROBLEM DESCRIPTION:
Small window exists when small spurts of disk I/O are followed
by no disk I/O activity. Could miss an interrupt resulting in
a cmd timeout. I/O will succeed on retry.

PROBLEM CONCLUSION:
Insert primitives to ensure coherency between memory and
interrupt state.

------

APAR: IY38553 COMPID: 5765E6100 REL: 510
ABSTRACT: PAX PROBLEM WITH -RW OPTION

PROBLEM DESCRIPTION:
The 'pax' command when given as 'pax -rw . <outdir>'
does not copy the data of the first file it writes
to the outdir.

PROBLEM SUMMARY:
The pax command when given as 'pax -rw . <outdir>'
does not copy the data of the first file it copies
into the output directory.

PROBLEM CONCLUSION:
Code changes are made to 'pax' code to make sure
that the data of all the files are copied properly
when given as 'pax -rw . <outdir>'.

------

APAR: IY38558 COMPID: 5765E6100 REL: 510
ABSTRACT: IN 32-BIT KERNEL SYSTEM CALLS, 64BIT MAPPING DOES NOT TAKE

PROBLEM DESCRIPTION:
In 32-bit kernel system calls, 64 bit address mapping does not
take place when the lower order bits of the 64bit parameter
are all 0s.

PROBLEM CONCLUSION:
Currently, in all the system calls in syscalls.c, a check is
made to see if the parameter (the lower 32 bits of the actual
64-bit parameter) passed to the call is 0. In such a case,
it's being assumed that the parameter passed is NULL and
the calls to get64bitparm/as_remap64 are bypassed irrespective
of whether the higher order 32 bits of the 64 bit parameter is
non-zero.
Changes should be made in all the system calls to make
compulsory call to get64bitparm() in order to ensure that the
entire 64 bit parameter is non-null.

------

APAR: IY38559 COMPID: 5765E6100 REL: 510
ABSTRACT: SECLDAPCLNTD SOMETIMES HANGS WHILE READING THE DATA FROM SLAPD

PROBLEM DESCRIPTION:
The secldapclntd daemon some times hangs while
reading the data from ldap server. The user
login will not occur at that time. The daemon
needs to be killed and restarted again to
resolve the problem.

LOCAL FIX:
Kill the daemon and restart it again.

PROBLEM CONCLUSION:
The read from the socket is modified to read the complete
message from the server instead of reading BUFSIZ.
The daemon will now try to connect to the replica servers
ever 5 sec once it detects the ldap server is down.

------

APAR: IY38560 COMPID: 5765E6100 REL: 510
ABSTRACT: D_MAP_SLAVE_ISA ORDERING PROBLEM

PROBLEM DESCRIPTION:
Potential data error using ISA slave DMA on POWER4 systems.
The only device using this service is the floppy controller.
The impact is most likely to be that the desired data will
(very rarely) be read/written from the wrong page offset,
possibly also with an incorrect transfer count.

PROBLEM CONCLUSION:
Add the appropriate memory barrier instruction to the code
that sets up the ISA DMA controller.

------

APAR: IY38565 COMPID: 5765E6100 REL: 510
ABSTRACT: CRASH ATTEMPTING LVM BAD BLOCK RELOCATION

PROBLEM DESCRIPTION:
System crash in xmemdma64_list during bad block relocation
due to bad block directory not being reliably pinned.

PROBLEM CONCLUSION:
Always keep bad block directory pinned.

------

APAR: IY38573 COMPID: 5765E6100 REL: 510
ABSTRACT: AUTOFS PROTO FAILBACK

PROBLEM DESCRIPTION:
If the proto specified for a mount on the client is not
available the mount will fail.

PROBLEM CONCLUSION:
When proto is set, connection attempt is made and if this fails
the alternate proto is attempted.

------

APAR: IY38607 COMPID: 5765E6100 REL: 510
ABSTRACT: PROBLEM WITH GREP -W OPTION

PROBLEM DESCRIPTION:
The grep -w command should print lines if the pattern is
a valid word in the line. Instead it also print lines
if the pattern lies in the end of any word.

PROBLEM SUMMARY:
The grep -w command should print lines only if the pattern
matches a complete word in the line. Instead it also prints
lines if the pattern lies in the end of any word.

PROBLEM CONCLUSION:
Code changes are done to grep code so that the -w
options prints line only if the pattern is a
complete word in the line.

------

APAR: IY38618 COMPID: 5765E6100 REL: 510
ABSTRACT: UNABLE TO START LOADLEVELER AFTER TERMINATING LL_STARTD

PROBLEM DESCRIPTION:
Unable to start LoadLeveler after terminating LL_Startd.
Problem occurs due to thread looping in the kernel. The
stack trace for the thread is similar to:
 00471BD0 crref+000028 ()
 00472DF8 suser+000014 (??)
 00123FC0 soconnect+00010C (??, ??)
 00123938 _connect+000288 (??, ??, ??)
 0012E7CC connext+000014 (??, ??, ??)
 0000379C sc_msr_2_point+000028 ()
 D026CB34 connect+000140 (??, ??, ??)
 D107C648 D107C648 ()
 D10FE7C4 D10FE7C4 ()
 D10F52A0 D10F52A0 ()
 D10F49EC D10F49EC ()
 D10F7930 D10F7930 ()
 D10F3320 D10F3320 ()
 D1072A40 D1072A40 ()
 D004D3FC _pthread_body+0000E8 (??)

PROBLEM CONCLUSION:
Add "volatile" to the declaration of U_cred in the user.h
system header file.

------

APAR: IY38623 COMPID: 5765E6100 REL: 510
ABSTRACT: ALT_DISK_INSTALL 31 :TEST: ARGUMENT EXPECTED

PROBLEM DESCRIPTION:
The use of alt_disk_install to operate
on an invalid disk or device could
cause the command to fail device validity
checks which in turn makes it continue
execution.
This leads to unpredictable results and
perhaps misleading error messages.
An example of such behavior is documented below:
    # alt_disk_install -W testcase
    /usr/sbin/alt_disk_install 31 : test: argument expected
    Waking up volume group ...
    Device open failure: /dev/testcase
    /usr/sbin/alt_disk_install 6 : v1 > v2 : bad number

LOCAL FIX:
Avoid the use of invalid disks or devices when
performing alt_disk_install operations.

PROBLEM SUMMARY:
If the user specifies an invalid disk parameter, they will
get the following confusing errors:
/usr/sbin/alt_disk_install 31 : test: argument expected
    Waking up volume group ...
    Device open failure: /dev/testcase
    /usr/sbin/alt_disk_install 6 : v1 > v2 : bad number

PROBLEM CONCLUSION:
Changed alt_disk_install behavior to display the following
error message:
0505-125 alt_disk_install: target disk <input> does not exist.

------

APAR: IY38625 COMPID: 5765E6100 REL: 510
ABSTRACT: JFS2 INLINE LOGS, CHFS MAY CRASH

PROBLEM DESCRIPTION:
chfs with inlinelog may cause system crash.
stack trace typically looks like:
 00160D40 lmLogShutdown+000058 ()
 0016227C lmLogQuiesce+00000C ()
 004AF48C j2_extendfs+0005CC ()
 004AED34 j2ExtendFS+00000C ()
 004AEE4C j2_cntl+00009C ()
 004B1140 vfs_cntl+00001C ()
 004BC76C fscntl+0000F4 ()
 0000379C sc_msr_2_point+000028 ()

PROBLEM CONCLUSION:
Ensured that the quiesce is complete
before the shutdown phase is entered.

------

APAR: IY38635 COMPID: 5765E6100 REL: 510
ABSTRACT: SAR -V REPORTS 0 VALUES

PROBLEM DESCRIPTION:
sar -v reports 0 values for all columns in the output.

PROBLEM CONCLUSION:
The incorrect values are reported because the arguments
passed by sar to sadc are incorrect.

------

APAR: IY38667 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH 800-102-700-0C0 USB MOUSE

PROBLEM DESCRIPTION:
System crash with led 800-102-700-0c0 using mouse attached
to system via USB adapter.
From the dump:
01FAF828 usbmse:kmusb_get_event+00001C (34E8A400, 00000003,
01FB223C usbmse:mseOfflvl+000338 (??)
0001E00C i_offlevel+000090 (??, ??)
...

PROBLEM SUMMARY:
AIX 5.1 ML02 crash with led 800-102-700-0c0 using mouse
attached to system via USB adapter.
From the dump:
01FAF828 usbmse:kmusb_get_event+00001C (34E8A400, 00000003,
01FB36CC ?? ) 01FB223C usbmse:mseOfflvl+000338 (??)
0001E00C i_offlevel+000090 (??, ??)
...

PROBLEM CONCLUSION:
Add additional locking to properly serialize close and
offlevel processing in loaded system.

------

APAR: IY38699 COMPID: 5765E6100 REL: 510
ABSTRACT: GXT4000P ANTI-ALIASED LINE RENDERING ANAMOLY OPENGL

PROBLEM DESCRIPTION:
Env: AIX5.1/OpenGL 7044-170/GXT4000P Prob: antialiasing
partiality doesn't work on diagonal line with GXT4000P.
Rendering anamoly on anti-aliased diagonal linesegments where
line has small horizontal segments as well as diagonal
segments. Some segments are rendered "bold".

------

APAR: IY38736 COMPID: 5765E6100 REL: 510
ABSTRACT: AUTHENTICATE COREDUMPS IN 64BIT APPLICATION

PROBLEM DESCRIPTION:
The authenticate subroutine call coredumps
in a 64-bit application. The 32-bit
application works fine.

PROBLEM CONCLUSION:
The headerfile stdlib.h is added which has the malloc
prototype. Since the prototype is not defined, the malloc
pointer gets typecase to int and coredump occurs.

------

APAR: IY38775 COMPID: 5765E6100 REL: 510
ABSTRACT: CRASH IN PRSYSCALLEXIT

PROBLEM DESCRIPTION:
System hits a DSI in the kernel when tracing using procfs.
The stack trace is:
 00009514 .simple_lock+000014 ()
 00045DD0 prWakeup+00009C (??, ??)
 00024190 stop_thread+000110 (??)
 00044F50 prSyscallExit+000130 ()
 00003E2C .update_errno+0001F0 ()

PROBLEM CONCLUSION:
Ensure that the same lock is used for sleep and wakeup for the
event list lock protecting a debugger process sleeping for an
event in the traced process.

------

APAR: IY38850 COMPID: 5765C3403 REL: 430
ABSTRACT: AIX 4330-11 RECOMMENDED MAINTENANCE PACKAGE

PROBLEM DESCRIPTION:
This APAR delivers the 4330-11 Recommended Maintenance package
for AIX 4.3.3.
To install selected updates from this package, use the command:
  smit update_by_fix
To install all updates from this package that apply to installed
filesets on your system, use the command:
  smit update_all
It is highly recommended that you install all updates from this
package.
This APAR and the PTFs associated with this Maintenance
Level are not installed when systems are preloaded with
AIX 4.3.3.

PROBLEM CONCLUSION:
Packaging only.

------

APAR: IY38856 COMPID: 5765E6100 REL: 510
ABSTRACT: NEGATIVE ACCESS CACHE

PROBLEM DESCRIPTION:
when authentication failed, nfs failed the access
because of the access cache. The customer has to re-mount
the mount for the access cache to be clear in addition to
authenticate itself on the network. nfs needs to
add in negative access caching.

PROBLEM CONCLUSION:
add in negative access caching in nfs3write and nfs3_access

------

APAR: IY38860 COMPID: 5765E6100 REL: 510
ABSTRACT: 64-BIT APP, 32-BIT KERNEL, TRACE TIMES CAN BE MESSED UP

PROBLEM DESCRIPTION:
If a 64-bit application traces using the 32-bit
trace, the reported elapsed and delta times can
be wrong.

PROBLEM CONCLUSION:
Fix a sign extention problem in 4.3 and 5.1.
Fix a data retrieval problem in 5.2 and beyond.

------

APAR: IY38861 COMPID: 5765E6100 REL: 510
ABSTRACT: TRACE_UTIL HOOK REPORT CORRUPTED ENTRIES

PROBLEM DESCRIPTION:
A long program name can cause trace to
incorrectly report processes and threads.

PROBLEM CONCLUSION:
Make sure the line buffer is long enough to
adequately handle the ps command output.

------

APAR: IY38863 COMPID: 5765E6100 REL: 510
ABSTRACT: DSI IN IODONE ROUTINE ON MP SYSTEMS RUNNING NUMEROUS IOCTLS

PROBLEM DESCRIPTION:
The problem manifests itself as a DSI in the adapter's iodone
routine under heavy ioctl stress. This is not predictably
recreatable.

PROBLEM CONCLUSION:
To fix this problem, we need to OR in the B_MPSAFE_INITIAL flag
while building the buf structure for ioctls. This ensures that
the only function setting the B_DONE flag for the buf is the
adapter driver's iodone routine.

------

APAR: IY38864 COMPID: 5765E6100 REL: 510
ABSTRACT: ADD SUPPORT FOR PCI-X CMD REG TO DD, DPR AND PRESERVE

PROBLEM DESCRIPTION:
Customer will see lesser performance after an Enhanced I/O
Error Handling (EEH) event.

PROBLEM CONCLUSION:
Driver saves off PCI-X registers to restore performace values
after an EEH event.

------

APAR: IY38866 COMPID: 5765E6100 REL: 510
ABSTRACT: GIGABIT ETHERNET PCI-X ADAPTERS: TCP RETRANSMITS

PROBLEM DESCRIPTION:
Poor performance may be observed by the customer when running
applications which use TCP.

PROBLEM CONCLUSION:
Perform TCPIP context initialization prior to first
TCP segmentation context initialization during tx
initialization sequence.

------

APAR: IY38867 COMPID: 5765E6100 REL: 510
ABSTRACT: TAR -U NOT BEHAVING AS DOCUMENTED

PROBLEM DESCRIPTION:
tar -u will always archive files even if the files were not
modified since the last time they were archived.

PROBLEM SUMMARY:
tar -u will not archive the files correctly. Ideally it
should archive only files whose modification times have
changed after the last archive.

PROBLEM CONCLUSION:
The part of the tar code that searches and matches
the modification time was rectified.

------

APAR: IY38870 COMPID: 5765E6100 REL: 510
ABSTRACT: DHCPREMOVE DID NOT REMOVE PTR RECORDS WHEN A RECORDS WERE

PROBLEM DESCRIPTION:
dhcpremove did not remove PTR records when A records were
missing

PROBLEM SUMMARY:
dhcpremove will no longer require the existance
of A records when deleting PTR records.

PROBLEM CONCLUSION:
dhcpremove will no longer check
for an existing A records when
deleting a PTR record.

------

APAR: IY38913 COMPID: 5765E6100 REL: 510
ABSTRACT: CANNOT EXECUTE BINARIES SMALLER THAN 1032 BYTES

PROBLEM DESCRIPTION:
Binaries smaller than 1032 bytes fail to execute.
They fail with a 'cannot execute' message.

PROBLEM CONCLUSION:
For binaries when reading in the COFF header do not assume
a size greater then the header size.

------

APAR: IY38914 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM HANG OR CRASH RUNNING CHFS ON JFS2 FILE SYSTEM

PROBLEM DESCRIPTION:
Running extendfs on a JFS2 filesystem with an inlinelog
could result in a system hang or crash.

PROBLEM CONCLUSION:
Support the use of chfs to change inlinelog size
and also lv size should count the increase of
inlinelog

------

APAR: IY38926 COMPID: 5765E6100 REL: 510
ABSTRACT: LOGREDO DOESN'T WORK WITH RAW DEVICES

PROBLEM DESCRIPTION:
When running fsck, customer may notice that logredo fails,
calling j2_logredo on jfs filesystems.

PROBLEM CONCLUSION:
Add a loop in logredo which will detect and properly handle
raw devices

------

APAR: IY38928 COMPID: 5765E6100 REL: 510
ABSTRACT: TRACE BASED PERFORMANCE TOOLS BROKEN

PROBLEM DESCRIPTION:
tprof, netpmon & filemon reports lot of Unknown process
and also tprof reports are empty / inaccurate.

PROBLEM CONCLUSION:
Fixed trace processing code to get correct process information
from the trace utility hook.

------

APAR: IY38930 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH (DSI) IN XTI_OSRQ_WAKEUP()

PROBLEM DESCRIPTION:
When gil(thread 13) is trying to respond a T_ORDREL_REQ from
application, Another thread is doing closing to pop up a
moudule above xtiso. Unfortunately, they are running on
same device, which cause privative data structure inconsistant,
and DSI happend.
 00009068 .disable_lock+000064 ()
 0254137C xtiso:xti_osrq_wakeup+000028 (??)
 02548824 xtiso:xti_ordrel_req+0001B4 (??, ??)
 02541D1C xtiso:xtiwsrv+0001F0 (??)
 02508104 pse:sq_wrapper+0000AC (??)
 024F81A8 pse:csq_run+0002B0 (??)
 024F7194 pse:csq_lateral+0000C0 (??, ??, ??)
 02508A78 pse:scheduled_run+0000D0 ()
 001086D0 Netintr+000140 ()
 00108550 netisr_thread+000014 ()
 000C3340 threadentry+000014 (??, ??, ??, ??)
and
 0002DD1C e_block_thread+0002EC ()
 024F780C pse:_csq_acquire+0000A0 (??, ??)
 02510D78 pse:osr_pop_subr+000100 (??, ??)
 024FE9DC pse:osr_close_subr+00039C (??)
 024FF0DC pse:pse_close+0004EC (??, ??, ??, ??)
 00469500 devcclose+0001F4 (??, ??)
 00515F00 spec_close+00007C (??, ??, ??, ??)
 0047DD58 vnop_close+000018 (??, ??, ??, ??)
 004B6364 vno_close+000028 (??)
 004AE038 closef+000078 (??)
 004530A0 closefd+0000A0 (??, ??)
 00453334 close+000230 (??)
 0000379C sc_msr_2_point+000028 ()
 D04926F0 t_close+000040 (??)
 D04555C4 _svc_vc_destroy_private+000068 (??, ??)
 D04318C4 _svc_destroy_private+000048 (??)
 D045626C _svc_done_private+0000C0 (??)
 D04567C8 _svc_run_mt+000420 ()
 D0457404 svc_run+000060 ()
 10000FA4 main+0005AC (00000001, 2FF22CD0)
 100001D8 __start+000088 ()

PROBLEM CONCLUSION:
Don't call osrq_wakeup() from xtiso module, but pass the
requirement to stream header, and call osrq_wakeup() there.

------

APAR: IY39007 COMPID: 5765C3407 REL: 210
ABSTRACT: V2R1 PTF-14

PROBLEM SUMMARY:
Added filesets for AIX v2r1 NSM PTF 14

PROBLEM CONCLUSION:
Filesets ready for ship

------

APAR: IY39009 COMPID: 5765C3407 REL: 210
ABSTRACT: V2R1 PTF-14 DUMMY

PROBLEM SUMMARY:
Added filesets for AIX v2r1 NSM PTF 14

PROBLEM CONCLUSION:
Filesets ready for ship

------

APAR: IY39091 COMPID: 5765B9501 REL: 330
ABSTRACT: SIGNAL 11 AT LOCATION 0X1000F36AC IN PROCESS 745696

PROBLEM DESCRIPTION:
signal 11 at location 0x1000f36ac in process 745696

PROBLEM SUMMARY:
mmfsd failing with signal 11 at location 0x1000F36AC

PROBLEM CONCLUSION:
signal 11 at location 0x1000F36AC in process 745696, link
reg 0 countTokens needed to handle compacted objects

------

APAR: IY39161 COMPID: 5765E6100 REL: 510
ABSTRACT: RPC.YPPASSWDD MAY COREDUMP W/ ALTERNATE PASSWD FILE

PROBLEM DESCRIPTION:
/usr/lib/netsvc/yp/rpc.yppasswdd may coredump if an alternate
password file is being used for NIS authentication and the
file /var/yp/passwd.ptmp already exists. The corefile would
have the following stacktrace:
(dbx) where
fclose_unlocked()
fclose()
changepasswd()
svc_getreqset()
svc_run()
main()

PROBLEM SUMMARY:
Fix cleanup code to ensure fclose is no longer called on
an invalid file pointer.

PROBLEM CONCLUSION:
Fix cleanup code to ensure fclose is no longer called on
an invalid file pointer.

------

APAR: IY39176 COMPID: 5765E6100 REL: 510
ABSTRACT: GSCLVMD DIES DUE TO FAILED EXECVE() CALL AND FAILS TO RESTART

PROBLEM DESCRIPTION:
When the enhanced concurrent LVM daemon dies, Volume Groups
varied on in enhanced concurrent mode remain varied on.

PROBLEM CONCLUSION:
Make the LVM enhanced concurrent daemon restartable.

------

APAR: IY39209 COMPID: 5765E6100 REL: 510
ABSTRACT: DBX CORE DUMPES WHEN TRYING TO LOAD AN APPLICATION

PROBLEM DESCRIPTION:
dbx will core when trying to load an executable.

PROBLEM SUMMARY:
dbx core dumps when trying to load an application

PROBLEM CONCLUSION:
dbx was statically linked to libdemangle.a. This has been
the cause of the problem. The solution is to link libdbx
to libC.a.

------

APAR: IY39239 COMPID: 5765E6100 REL: 510
ABSTRACT: CHANGE DIAG METHOD FOR IDENTIFYING TYPE OF PCI-X SCSI ADAPTER

PROBLEM DESCRIPTION:
Methodology for determining the type of PCI-X SCSI adapter
for the microcode download diagnostic service aid may not
accurately determine adapter type.

PROBLEM CONCLUSION:
Correct methodology for determining type of PCI-X SCSI
adapter. Affected fileset is devices.pci.14106602.diag

------

APAR: IY39249 COMPID: 5765E6100 REL: 510
ABSTRACT: ERRPT COMMAND CORE DUMPS IN STRCMP

PROBLEM DESCRIPTION:
errpt -c can core dump with a stack trace similar to:
strcmp D0
cat_alrea 34
catopen 1B0
__strerro 28
perror 84
logread B0
logget B4
udb_logge 28
rpt_c 90
main 180
__start 8C

PROBLEM CONCLUSION:
Check the error record length read from the file.

------

APAR: IY39262 COMPID: 5765E6100 REL: 510
ABSTRACT: 128-PORT ADAPTER CANNOT RECOVER FROM EEH EVENT.

PROBLEM DESCRIPTION:
The customer will see an adapter failure error (CXMA_ADP_FAIL)
in the error log after an Extended Error Handling (EEH) event,
and the adapter will stop working until it has been
unconfigured and reconfigured. If there is an adapter fail
data (CXMA_ADP_FAIL_DATA) message, it will contain all 0xFF's.

PROBLEM CONCLUSION:
Because the EEH event happened between the time that the
driver did an I/O memory attach and the time the driver
checked for an adapter failure, the EEH event is erroneously
logged as an adapter failure. The driver will now check for
and EEH event before it logs the adapter failure error, and
handle it accordingly.

------

APAR: IY39263 COMPID: 5765E6100 REL: 510
ABSTRACT: AUTOFS: CRASH IN AIX2ONC_RELE

PROBLEM DESCRIPTION:
System crash with the following stack trace...
> t -mk
IAR: .low+0 (00000000): ?
LR: . autofs.ext:aix2onc_rele +7c ()
2ff3ae40: . autofs.ext:aix2onc_mkdir +84 ()
2ff3aef0: .vnop_mkdir+30 ()
2ff3af50: .mkdir+188 ()
2ff3b3c0: .sys_call_ret+0 ()
05556000: kfork+0 ()
0000e000: .nodev+0 ()
Branched to the NULL pointer during an autofs
mkdir operation.

PROBLEM SUMMARY:
Modified aix2onc_mkdir so that it will no longer
release the wrong vnode.

PROBLEM CONCLUSION:
Modified aix2onc_mkdir so that it will no longer
release the wrong vnode.

------

APAR: IY39403 COMPID: 5765E6100 REL: 510
ABSTRACT: MODE_SENSE CMD CDB REPORTS ILLEGAL REQUEST ERROR

PROBLEM DESCRIPTION:
Unable to configure the tape drive, with "invalid request"
error msg.

PROBLEM CONCLUSION:
Correct MODE_SENSE cdb request.

------

APAR: IY39425 COMPID: 5765E6100 REL: 510
ABSTRACT: NODE HANG AFTER RESTART FROM CHECKPOINT

PROBLEM DESCRIPTION:
64bit kernel may crash for certain
instances where it should generate a
core dump.

PROBLEM CONCLUSION:
Fix system call handler in 64bit kernel
to correctly handle bad system call.

------

APAR: IY39429 COMPID: 5765E6100 REL: 510
ABSTRACT: PROGRAMS USING SMALL STACK THREADS CORE DUMP

PROBLEM DESCRIPTION:
After installation of bos.rte.libc 4.3.3.88 or 5.1.0.37,
programs calling gethostbyname2() using small stack threads may
core dump.

PROBLEM CONCLUSION:
Modifying the MAXPACKET size from 64K back to 1K in
ho_byname2()

------

APAR: IY39779 COMPID: 5765E6100 REL: 510
ABSTRACT: SYSTEM CRASH RUNNING DIAGNOSTICS ON 23100020 ETHERNET ADAPTER

PROBLEM DESCRIPTION:
The customer may experience a machine crash in certain slots
running a system in LPAR mode.

PROBLEM CONCLUSION:
We have fixed the cleanup routines for NULL addresses.

------

APAR: IY39833 COMPID: 5765F7100 REL: 810
ABSTRACT: INTERNAL APAR FOR XLF RUNTIME PTF 8.1.0.3

PROBLEM DESCRIPTION:
Internal APAR for xlf runtime PTF 8.1.0.3

------

APAR: IY39834 COMPID: 5765F7000 REL: 810
ABSTRACT: INCORRECT OUTPUT WITH SQRT AND -O AND AUTODBL=DBLPAD

PROBLEM DESCRIPTION:
Incorrect output with sqrt and -O and autodbl=dblpad

------

APAR: IY39835 COMPID: 5765F7000 REL: 810
ABSTRACT: XLC CALLS FE WITH -Q32 NO MATTER IF -Q64 IS INCLUDED IN .CFG

PROBLEM DESCRIPTION:
If we compile a simple testcase like :
#include<stdio.h>
main()
{
  printf("Hello World\n");
}
with the commnd :
xlc test.c -Fv.cfg
where the options -q64 and -V are added in the options
line of the xlc stanza,
we get the following output:
/.../torolab.ibm.com/fs/projects/vabld/run/matrix/600/
aix/daily/020813/exe/xlcentry -D_AIX -D_AIX32 -D_AIX41
-D_AIX43 -D_AIX50 -D_AIX51 -D_IBMR2 -D_POWER -q32 -
qansialias -otest.o test.c /tmp/xlcW0GDFria /tmp/
xlcW1GDFrib /dev/null test.lst /dev/null /tmp/xlcW2GDFric
1506-507 (W) No licenses available. Contact your program
supplier to add additio nal users. Compilation will
proceed shortly.
/.../torolab.ibm.com/fs/projects/vabld/run/matrix/600/
aix/daily/020813/exe/xlCcode -qansialias /tmp/xlcW0GDFria
/tmp/xlcW1GDFrib test.o test.lst /tmp/xlcW2GDFric
/bin/ld -b64 /lib/crt0_64.o -bpT:0x10000000
-bpD:0x20000000 test.o -L/.../torolab.ibm.com/fs/projects/
vabld/run/matrix/600/aix/daily/020813/lib/aix51 -L/usr/lpp
/xlopt -lxlopt -lc
ld: 0711-738 ERROR: Input file test.o:
        XCOFF32 object files are not allowed in 64-bit mode.
rm /tmp/xlcW0GDFria
rm /tmp/xlcW1GDFrib
rm /tmp/xlcW2GDFric
The -q32 is invoked and xlcentry compiles in 32-bit mode,
but when it links in 64-bit mode, error occurs.
We didn't invoke -q32 option anywhere, it is automatically
invoked by the driver.
This happens whether we invoked -q64 or not, as if -q32 is
a default setting for the xlc drvier.

LOCAL FIX:
A workaroun would be by adding -q64 in the command line, or by
setting the OBJECT_MODE 64 .

------

APAR: IY39836 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR WITH COMBINATIONS OF O3 FDPR TUNE AND

PROBLEM DESCRIPTION:
The following testcases will cause internal compiler errors:
Raised in a single APAR as ll have to do with switch statement
optimization.
Testcase #1:
============
int DummyFunc(void){}
int f(void)
{
   int status;
   int collection = 0;
   int objCollection = 0;
   if (status){
      collection = 1;
      DummyFunc();
   }
   if (collection){
      DummyFunc();
      objCollection = 1;
   }
   while (status == 0){
      switch(status){
      case 41:
  if (objCollection)
           break;
      case 42:
  if (status == 0)
            break;
      case 30:
      case 31:
      case 33:
  if (status == 0)
            status = 1;
      default:
      break;
      }
      if (!collection) status = 0;
   }
}
xlc -O3 -q64 -qfdpr -qpdf1 -qtune=rs64b -qcompact -c t1.c
1586-346 (S) An error occurred during code generation. The code
generation return code was 40
Testcase #2:
============
void f(void){
   short op;
   switch (op){
      case 136:
      case 134:
      case 4:
      case 2:
      case 135:
      case 133:
      case 3:
      case 1:
         switch (op){
            case 136:
            case 134:
            case 4:
            case 2:
            op = 1;
            break;
            case 135:
            case 133:
            case 3:
            case 1:
            op = 2;
            break;
         } /* nested switch */
      default:
      break;
   } /* switch */
} /* f */
xlc -O3 -qcompact -qfdpr -qtune=rs64b -c t2.c
1501-230 Internal compiler error; please contact your Service
Representative
Testcase #3:
============
dummy_func(void)
{
   int opt;
   switch (opt) {
      case 'U': case 'u': break;
      case 'K': case 'k': break;
      case 'M': case 'm': break;
      case 'G': case 'g': break;
      case 'T': case 't': break;
      case 'P': case 'p': opt = 0; break;
      case 'E': case 'e': opt = 1; break;
   }
}
dummyfunc1(void){
   dummy_func();
}
xlc -O3 -qfdpr -qtune=rs64b -c t3.c
1501-230 Internal compiler error; please contact your Service
Representative

LOCAL FIX:
Do not use optimization.

PROBLEM CONCLUSION:
Fixed in 6.0.0.2.

------

APAR: IY39837 COMPID: 5765F7000 REL: 810
ABSTRACT: -QFLOAT=EMULATE -O => "COMPILATION ENDED DUE TO LACK OF SPACE"

PROBLEM DESCRIPTION:
The customer sent a testcase fails with -qfloat=emulate -O and
gives "compilation ended due to lack of space".
I produced the preprocessed file and was able to reduce it from
10,000 lines of code to:
-3,600 lines failing on V5
-1,275 lines failing on V6
both tested on latest builds.

------

APAR: IY39838 COMPID: 5765F7000 REL: 810
ABSTRACT: BAD OBJECT CODE CAUSING LINKER FAILURE

PROBLEM DESCRIPTION:
Testcase:
===========
void dummy(void){
   short fakevar, len;
   if(&fakevar)
      fakevar = len;
   if(len && !fakevar)
      switch(len){
         case 3: if(len); break;
         case 2: len = 1; break;
         case 4: len = 1; break;
         case 6: {
            int var = 1;
            len = 1;
            break;
         }
         default: len =1; break;
      }
   switch(len) {
      case 3: len = 1; break;
      case 2: len = 1; break;
      case 4: len = 1; break;
      case 6: len = 1; break;
      default: len = 1; break;
   }
}
int main(){
   dummy();
}
xlc t.c -O3 -qfdpr -qtune=rs64b -q64 -qcompact
ld: 0711-596 SEVERE ERROR: Object t.o
   An RLD for section 2 (.data) refers to symbol 0,

LOCAL FIX:
Do not use the combination of -O3 -qfdpr -qtune=rs64b -q64
-qcompact.

PROBLEM CONCLUSION:
Fixed in V6 PTF #2.

------

APAR: IY39839 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR WHEN USING __CHECK_LOCK_MP() WITH O3 AND

PROBLEM DESCRIPTION:
Source code making the use of __check_lock_mp() causes the VAC++
V6 compiler to crash with ICE when compiled as follows:
xlc -c -DUSE_CHECK -O3 -qarch=pwr4 ... test.c

LOCAL FIX:
Undefine USE_CHECK symbolic constant.

PROBLEM SUMMARY:
The problem has been fixed in VAC++ Nov 12, 2002 build.

PROBLEM CONCLUSION:
The problem originated at the compiler backend component.

------

APAR: IY39840 COMPID: 5765F7000 REL: 810
ABSTRACT: DIVIDE BUILTIN ABENDS WITH ZERODIVIDE.

PROBLEM DESCRIPTION:
The ZERODIVIDE condition is rasied for the DIVIDE builtin
function with a precision of 31,5.
IBM0301S ONCODE=320 The ZERODIVIDE condition was raised.

PROBLEM SUMMARY:
****************************************************************
* USERS AFFECTED: All. *
****************************************************************
* PROBLEM DESCRIPTION: DIVIDE BUILTIN ABENDs with ZERODIVIDE. *
****************************************************************
* RECOMMENDATION: *
****************************************************************

------

APAR: IY39841 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR

PROBLEM DESCRIPTION:
The testcase has been produced by preprocessing a compiler
generated file from .tempinc directory, used for automatic
template instantiation. A lot of OpenClass headers have been
pulled in with a number of template instantiations. The build
takes between 1 and 5 hours depending on the system load, and
fails with the ICE and dumps the message below.
xlC -qnodigraph -c -O -qrtti=all -qdebug=EXCEPT -v iikssavl.i
EXCEPTION terminates program: no handler for exception.
Exception of type: TRAP at 100C0174. Regs:
000193DF 2FF1F400 301B4154 000101BC 00000000 3001F3CC 85C2D0EC
565526A8
00000000 0001D6A1 575BB208 00000000 00000000 57331908 00000004
0000000C
57331908 00000000 00000004 000155C5 80000000 58154BE8 0000038E
5734F008
565526A8 000000FF 2E5C7898 57F31658 000101BC 000101BC 00000000
FFC00000
resume: cr: lr: ctr: xer: fpscr: msr:
mq:
100C0178 42200840 100BFCDC 00000007 20000004 00002000 0000D032
00000000
Trap condition: -4194304 l>= 103391 logical
compare
Trap instruction: T 5,R31,R00
Traceback:
Line ? Disp 00000090 scan_list Dsa:
2FF1F400
Line ? Disp 00000118 build_the_alias_array Dsa:
2FF1F400
Line ? Disp 00000090 build_alias_array Dsa:
internal
Line ? Disp 00000148 gen_kill_vectors Dsa:
2FF1F4F0
Line ? Disp 0000007C late_recan Dsa:
internal
Line ? Disp 000001E4 canonize Dsa:
2FF1F750

LOCAL FIX:
Use option -qnotempinc instead of -qtempinc (default).

PROBLEM SUMMARY:
Fixed in V5. will be available with the next PTF (no target
date yet).

PROBLEM CONCLUSION:
Limitation fixed in V5.

------

APAR: IY39842 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR WHEN COMPILING AT OPT(2)

PROBLEM DESCRIPTION:
If the program is too complex and contains many (hundreds) of
data variables, the optimizer may run out of storage and the
compile may terminate with an "internal compiler" error.

LOCAL FIX:
turn optimization off

PROBLEM SUMMARY:
****************************************************************
* USERS AFFECTED: Users of OS/390 V2R10 C/C++ compiler *
* who experienced the CBC5104 error *
* message or the CBC5002 error message *
* when compiling a vary large source file *
* with the OPTIMIZE(1) or *
* OPTIMIZE(2) option. *
****************************************************************
* PROBLEM DESCRIPTION: The compiler itself operates within the *
* limitations set by the operating *
* environment. Thus when the source file *
* is large and complex the amount of *
* virtual storage required by the compile *
* can not be allocated. This is particula *
* so during the optimization phase where *
* the demand for virtual storage would be *
* much higher than NOOPTIMIZE. For the *
* particular case reported by the APAR, *
* the compiler failed to detect the *
* overflow condition when the storage siz *
* for a data structure was being computed *
* Thus a CBC5104 message was issued *
* when the compiler got into trouble late *
* The compiler is fixed to issue a CBC510 *
* message to inform about the program *
* being too complex to compile and return *
* RC=16. *
****************************************************************
* RECOMMENDATION: *
****************************************************************
For some large and complex C/C++ source files received by the
compiler the virtual storage required by the compiler may have
exceeded the limitation set by the operating environment
without being detected by the compiler.

PROBLEM CONCLUSION:
Apply provided service.

------

APAR: IY39843 COMPID: 5765F7000 REL: 810
ABSTRACT: DEBUGGER STOPS AT A COMMENTED LINE WHEN A TESTCASE IS INLINED

PROBLEM DESCRIPTION:
The customer sent a testcase that misleads the debugger when
compiled with -Q When debugging the testcase, the debugger stops
at a blank line, and a commented line !
dbx behaves as above, but idebug stops at blank lines in the tes
This is the second testcase for the same customer of defect 2294
and she would like to know, applying on this specific testcase,
why the debugger behaves like that.
testcase is located on DFS /:/projects/vacsup/testcase/aix/aix/<
the command line with the output:
$ xlC_r -Q -g e13.C n13.C
e13.C:
n13.C:
$ dbx a.out
Type 'help' for help.
reading symbolic information ...
(dbx) stop in main
1. e13.main
2. n13.main(int,char**)
Select one or more of [1 - 2]: 1
[1] stop in main
(dbx) s
[1] stopped in main at line 10 in file "e13.C" ($t1)
   10 {
(dbx) n
stopped in main at line 11 in file "e13.C" ($t1)
   11 Testd *lm = new Testd;
(dbx) n
stopped in main at line 12 in file "e13.C" ($t1)
   12 lm->main(argc, argv);
(dbx) s
stopped in unnamed block $b2 at line 24 in file "N13.C" ($t1)
   24 NEW_LOCK_CONFIG_READ;
(dbx) n
stopped in unnamed block $b2 at line 9 in file "N13.C" ($t1)
    9
(dbx) n
stopped in unnamed block $b2 at line 65528 in file "N13.C" ($t1)
"N13.C" has only 28 lines
(dbx) n
stopped in unnamed block $b2 at line 10 in file "N13.C" ($t1)
   10 //
(dbx) n
stopped in newNetProcess::main(int,char**) at line 27 in file "N
   27 return(0);
(dbx) n
stopped in newNetProcess::main(int,char**) at line 28 in file "N
   28 }
(dbx) q
$
Please advise, Thanks

PROBLEM CONCLUSION:
This was caused by a line-number adjustment problem. This
problem does not affect the v6 compiler, only v5. This is
because v6 doesn't allow -g and -Q to be used together. This
problem has been fixed. A fix for this problem should be
available in the next v5 PTF.

------

APAR: IY39844 COMPID: 5765F7000 REL: 810
ABSTRACT: PROGRAMME DUMPS CORE WHEN COMPILED WITH -Q64, -O3, AND -QCOMPACT

PROBLEM DESCRIPTION:
int donothing(){ }
int main(){
   unsigned int var=0xfffffffa;
   testfunc(var);
}
int testfunc(unsigned int par){
   unsigned int var = par;
   switch (var){
      case (((unsigned int) (4294967295U)) - 1):
      donothing();
      break;
      case (((unsigned int) (4294967295U)) - 5):
      donothing();
      break;
      case (((unsigned int) (4294967295U)) - 6):
      break;
      case 0x00000009:
      case 0x00000008:
      break;
   }
}
$ xlc t.c -q64 -O3 -qcompact
$ a.out
Segmentation fault
Remarks:
=========
- The testcase appears to be raising a SIGPIPE(?) exception as
soon as it hits the switch statement in testfunc().
- It happens only when compiled with the combination of options:
-q64, -O3, and -qcompact.
- It runs clean when compiled with V5.

LOCAL FIX:
Drop -qcompact or -O3.

PROBLEM SUMMARY:
Compiler optimization problem. Fixed in VAC++
V6 PTF 2.

PROBLEM CONCLUSION:
Fixed in VAC++ V6 PTF 2.

------

APAR: IY39845 COMPID: 5765F7000 REL: 810
ABSTRACT: -QDATALOCAL WITH A GLOBAL OBJECT OF TYPE CLASS MISLEADS LINKER

PROBLEM DESCRIPTION:
The testcase:
class A {
public:
       A() { };
       ~A() { };
} a ;
int main() { }
The command and output:
$ xlC test.cpp -qdatalocal
ld: 0711-596 SEVERE ERROR: Object test.o
        An RLD for section 2 (.data) refers to symbol 0,
        but the storage class of the symbol is not C_EXT or C_HI
ld: 0711-596 SEVERE ERROR: Object test.o
        An RLD for section 2 (.data) refers to symbol 0,
        but the storage class of the symbol is not C_EXT or C_HI
ld: 0711-596 SEVERE ERROR: Object test.o
        An RLD for section 2 (.data) refers to symbol 0,
        but the storage class of the symbol is not C_EXT or C_HI
ld: 0711-596 SEVERE ERROR: Object test.o
        An RLD for section 2 (.data) refers to symbol 0,
        but the storage class of the symbol is not C_EXT or C_HI
Error was reproducable on all levels of V5 and V6

------

APAR: IY39846 COMPID: 5765F7000 REL: 810
ABSTRACT: CMPL ERROR IN CONVERSION OF FIXED DEC(31,29)TO PIC'99'

PROBLEM DESCRIPTION:
Incorrect output is received with either VA PL/I 2.2.1 or
Enterprise PL/I for z/OS 3.1.0 due to an error in conversion
of FIXED DEC(31,29) to PIC'99'.

LOCAL FIX:
recompile without LIMITS(FIXEDDEC(31)) option.

PROBLEM SUMMARY:
****************************************************************
* USERS AFFECTED: All. *
****************************************************************
* PROBLEM DESCRIPTION: Compile ERROR in CONVERSION of *
* FIXED DEC(31,29) to PIC'99'. *
****************************************************************
* RECOMMENDATION: *
****************************************************************

------

APAR: IY39847 COMPID: 5765F7000 REL: 810
ABSTRACT: -QSMP=OMP GIVES AN ERROR CODE 40 WITH A TESTCASE

PROBLEM DESCRIPTION:
When compiling the customer's testcase with -qsmp=omp, we get
an error code 40.
The error was reproduced with latest level of V6.

------

APAR: IY39848 COMPID: 5765F7000 REL: 810
ABSTRACT: ERROR WHEN FILENAME > 4 CHARS

PROBLEM DESCRIPTION:
*************************************************************
!* DEFECT ABSTRACT : Error when filename > 4 chars
!* CUSTOMER : IBM
!* PMR NUMBER : 87527,999,000
!* COMPILER INVOCATION : xlc
!* COMPILER OPTION(S) : -c -O3 -qpdf1 -qtocmerge
!* -bimport:/s/import/test.imp
!* OS OPTION(S) :
!* OS LEVEL : AIX4L, 5L
!* COMPILER LEVEL : V600 latest
!* MACHINE TYPE/MODEL :
!*
**************************************************************
NOTE: I am opening this defect against tobey because the
compiler seems to be failing in xlCcode component from the -v
option. However, you may find this is not the case once have
done addition investigation. Thank you.
PROBLEM DESCRIPTION:
I have already informed couple of problems related to the
combination of -O3, -qpdf1, and -qtocmerge. This is another
problem which is related to those options along with -bimport.
This problem is really tricky and related to the location of
import file.
I have a standalone test-case which can be used to recreate
this problem on developer's environment. (Please see below)
Please use the test-case as it is. don't change the function
names and order in the structure. If you do so, you may not
recreate this issue.
Here I am going to elaborate the location of the import file.
Please follow the following steps.
$ su root
$ mkdir asar
$ chown <user>:<group> asar
$ ln -s /asar /s
now log out from root previlage and login as <user>
$ mkdir -p /s/imp
$ touch /s/imp/test.imp
Please follow the abobe mentioned steps proerly. Otherwise you
may not
able to recreate this problem.
$ cd /home/changw/PMRs/87527/
$ ls
hok.c
$ xlc -c hok.c -O3 -qpdf1 -qtocmerge -bimport:/s/imp/test.imp
here is the error message I got in my environemnt
1586-346 (S) An error occurred during code generation. The
code generation return code was 40.
However, the compiler doesn't hit the error when I changed the
file name from hok.c to t1.c
$ xlc -c t1.c -O3 -qpdf1 -qtocmerge -bimport:/s/imp/test.imp
$
After further testing, I found that if the filename is greater
than 2 characters (or 4 if you include the ".c"), the compiler
crashes.
Here's the -qdebug=except and -v output.
$ xlc -qnolm -c -qdebug=except hok.c -O3 -qpdf1 -qtocmerge
-bimport:/s/imp/test.imp
 EXCEPTION terminates program: no handler for exception.
 Exception of type: TRAP at 1011C2AC. Regs:
 00000000 2FF21130 3029C694 00000000 57617B88 00000008 00000001
00000000
 10009BBD 10009BBD 00000000 5761AC94 1011C29C DEADBEEF 00000007
2FF223F4
 2FF22414 A0001B38 00010B0C 00000008 00000002 46EAE0A8 30017DE0
416B0C08
 30015BA0 3000DA90 00000026 5770E688 A00001C0 46EADE48 00000000
3002C1A0
 resume: cr: lr: ctr: xer: fpscr: msr:
mq:
 1011C2B0 22422842 1011C29C D0062F00 00000000 00002000 0000D0B2
00000000
 Trap condition: 0 = 0 arithmetic compare
 Trap instruction: TI 4,R00,0
 Traceback:
 Line ? Disp 00000034
Dsa: 2FF21130
 Line ? Disp 000000C0
Dsa: 2FF221B0
 Line ? Disp 0000028C
Dsa: 2FF22230
 Line ? Disp 00000128
Dsa: 2FF22290
 Line ? Disp 00000120
Dsa: 2FF22320
 --- End of call chain ---
1586-346 (S) An error occurred during code generation. The code
generation return code was 217.
$ xlc -qnolm -c -v hok.c -O3 -qpdf1 -qtocmerge
-bimport:/s/imp/test.imp
exec:
/.../torolab.ibm.com/fs/projects/vabld/run/matrix/600/aix/daily/
021125//xlcentry(/.../torolab.ibm.com/fs/projects/vabld/run/matr
ix/600/aix/daily/021/exe/xlcentry,-D_AIX,-D_AIX32,-D_AIX41,-D_AI
X43,-D_AIX50,-D_AIX51,-D_IBMR2,-DWER,-qansialias,-qnolm,-O3,-qpd
f1,-qtocmerge,-bimport:/s/imp/test.imp,-ohok.ok.c,/tmp/xlcW0oLwK
Ea,/tmp/xlcW1oLwKEb,/dev/null,hok.lst,/dev/null,/tmp/xlcW2oEc,NU
LL)
exec:
/.../torolab.ibm.com/fs/projects/vabld/run/matrix/600/aix/daily/
021125//ipa(/.../torolab.ibm.com/fs/projects/vabld/run/matrix/60
0/aix/daily/021125/eipa,-comp,-v,-qansialias,-qnolm,-qpdf1,-qipa
=level=0,-qtocmerge,/tmp/xlcW0oLw,/tmp/xlcW1oLwKEb,hok.o,hok.lst
,/tmp/xlcW2oLwKEc,NULL)
exec:
/.../torolab.ibm.com/fs/projects/vabld/run/matrix/600/aix/daily/
021125//xlCcode(xlCcode,-qdebug=NLDSTM,/tmp/ipao_I4qa,/tmp/ipao_
I4qb,hok.o,hok.lst,//ipao_I4qc)
1586-346 (S) An error occurred during code generation. The code
generation rrn code was 40.
unlink: /tmp/xlcW0oLwKEa
unlink: /tmp/xlcW1oLwKEb
unlink: /tmp/xlcW2oLwKEc
TESTCASE:
Location: /:/projects/vacsup/testcase/aix/aix/87527/hok.c
Here is the test-case, say hok.c
void dummy0(void *par);
void dummy1(void *par);
void dummy2(void *par);
void dummy3(void *par);
void dummy4(void *par);
void dummy5(void *par);
void dummy6(void *par);
void dummy7(void *par);
void dummy8(void *par);
void dummy9(void *par);
void dummya(void *par);
void dummyb(void *par);
void dummyc(void *par);
void dummyd(void *par);
void dummye(void *par);
void dummyf(void *par);
void dummyg(void *par);
void dummyh(void *par);
void dummyi(void *par);
void dummyj(void *par);
void dummyk(void *par);
void dummyl(void *par);
void dummym(void *par);
void dummyn(void *par);
void dummyo(void *par);
void dummyp(void *par);
void dummyq(void *par);
void dummyr(void *par);
void dummys(void *par);
void dummyt(void *par);
void dummyu(void *par);
void dummyv1(void *par);
void dummyw(void *par);
void dummyx1(void *par);
void dummyy12(void *par);
void dummyz1(void *par);
struct dummystruct {
void (*dummy)( void *par);
} dummystructtab[] = {
dummy1, dummy2, dummy3, dummy4, dummy5,
dummy6, dummy7, dummy8, dummy9, dummya,
dummy3, dummy0, dummyb, dummyc, dummyd,
dummye, dummyf, dummyg, dummy1, dummyh,
dummyi, dummyj, dummyk, dummyl, dummym,
dummy3, dummyn, dummyo, dummy1, dummyp,
dummyq, dummyr, dummys, dummyt, dummyu,
dummyv1, dummyw, dummyx1, dummyy12, dummyi,
dummyz1
};
WORKAROUND:
Don't use the option combination stated above.
EXPECTED RESULTS:
- Clean compile.
ACTUAL RESULTS:
1586-346 (S) An error occurred during code generation. The
code generation return code was 217.
**************************************************************
**************************************************************

LOCAL FIX:
Don't use the option combination stated in the description.

PROBLEM SUMMARY:
Same as submitter's text

PROBLEM CONCLUSION:
Description from development:
Some code in USYM_f.c was assuming that the import filename was
null-terminated in the string pool. It isn't (or at least there
is no requirement for it to be). Cope with that and the
problem vanishes. Change the readImportFile routine to accept
a length parameter for the filename.

------

APAR: IY39849 COMPID: 5765F7000 REL: 810
ABSTRACT: QUERY RE. PERFORMANCE IMPACT OF -QDPCL

PROBLEM DESCRIPTION:
Query re. impact on performance of -qpdcl

PROBLEM CONCLUSION:
Development found the problem

------

APAR: IY39850 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL COMPILER ERROR IN IPA

PROBLEM DESCRIPTION:
A problem has been encountered where thre is an ICE in IPA with
options -O3 -qpdf1 -q64 -qlist combined. Build fails with ICE
and the following set of error messages:
1586-000: (E) Error in message set 1, unable to retrieve message
491.
1586-000: (E) Error in message set 2, unable to retrieve message
35.
1586-000: (E) Error in message set 2, unable to retrieve message
35.
1586-000: (E) Error in message set 2, unable to retrieve message
34.
1586-000: (E) Error in message set 2, unable to retrieve message
33.
1586-000: (E) Error in message set 2, unable to retrieve message
32.
1586-000: (E) Error in message set 2, unable to retrieve message
31.
It has been established that failure to pull out messages from
the message catalog is not the cause of ICE. The issue is under
investigation.

LOCAL FIX:
Use option -qipa=noobject.

PROBLEM SUMMARY:
Fixed in VAC++ V6. Will ship with the next PTF. Currently
scheduled for Dec 2002.

PROBLEM CONCLUSION:
Fixed in VAC++ V6. Will ship with the next PTF. Currently
scheduled for Dec 2002.

------

APAR: IY39851 COMPID: 5765F7000 REL: 810
ABSTRACT: "UNABLE TO RETREIVE MSG FROM .CAT" WHEN USING -O4 OR -O5 (IPA)

PROBLEM DESCRIPTION:
When compiling the customer's testcase (.i file) with any of
the following commands, we get "unable to retrieve error
messages from .cat".
Compiling commands:
xlC -c test.i -O4
xlC -c test.i -O5
xlC -c test.i -O3 -qipa
xlC -c test.i -O3 -qipa=level=0
Testcase:
located on DFS /:/projects/vacsup/testcase/aix/aix/30195
The output:
1586-000: (E) Error in message set 1, unable to retrieve
message 491.
1586-000: (E) Error in message set 2, unable to retrieve
message 35.
1586-000: (E) Error in message set 2, unable to retrieve
message 35.
1586-000: (E) Error in message set 2, unable to retrieve
message 34.
1586-000: (E) Error in message set 2, unable to retrieve
message 33.
1586-000: (E) Error in message set 2, unable to retrieve
message 32.
1586-000: (E) Error in message set 2, unable to retrieve
message 31.
1586-491 (
Notes:
1- Customer got an extra error messages :
          (E) Message system initialization, unable to open
catalogs:
vacimsg.cat, /usr/vac/exe/default_msg/vacimsg.cat.
2- I checked with Packaging and Test teams, customer has the
right size and date of the catalog file.
3- Compiling the testcase with -O or -O3, testcase succeeds
cleanly
4- Compiling with -O4 -qnoipa gives an object file, but
with this:
    1500-030: (I) INFORMATION: Uintah::ParticleVariable<class
SCIRun::Point>::ParticleVariable(): Additional optimization
may be attained by recompiling and specifying MAXMEM option
with a value greater than 8192.
    1500-036: (I) The NOSTRICT option (default at OPT(3)) has
the potential to alter the semantics of a program. Please
refer to documentation on the STRICT/NOSTRICT option for
more information
5- The size of the .i file is <1MB, and with -O4 or -O5 takes
about 45 minutes on a nice machine with 4GB RAM.

------

APAR: IY39852 COMPID: 5765F7000 REL: 810
ABSTRACT: INCORRECT OUTPUT(INFINITE LOOP) WITH -O3 AND -QPDF1.

PROBLEM DESCRIPTION:
When compiling test case with -O3 and -qpdf1, the compiler
generates an executable that contains a infinite loop, which
causes it to hang.
$ xlC -qdebug=except -O3 -qpdf1 -qnolm t2.cpp
      19 1500-010: (W) WARNING in main: Infinite loop.
 Program may not stop.
$ a.out
[hangs]
Following the logistics of the test case, there should not
be any infinite loops. This may be caused by incorrect
code rearrangement during the optimization phase.
This problem seems to during in -O2 as well.
I did not run the program through any debuggers because it
involves optimization.
ENVIRONMENT:
AIX4L,5L
VAC++ 600
TESTCASE:
// test.cpp
#define FALSE 0
#define TRUE 1
int main(void)
{
int status;
if ( FALSE )
 do {
  printf("Should not be here\n");
 } while ( *(&status) == 0 && 1 );
if ( FALSE ) ;
else if ( TRUE )
{
 do_again:
 printf("You are in else-if block\n");
 if ( FALSE )
  goto do_again;
}
if ( FALSE )
 goto do_again;
printf("Exit from the function\n");
return(0);
}
COMPILE COMMAND:
xlC -O3 -qpdf1 test.cpp
WORKAROUND:
Don't compile with -O3 and -qpdf1 together.
EXPECTED RESULTS:
Clean compile without warning message.
An executable without infinite loops.
ACTUAL RESULTS:
19 1500-010: (W) WARNING in main: Infinite loop. Program may
not stop.
An executable that hangs.

LOCAL FIX:
Don't compile with -O3 and -qpdf1 together.

PROBLEM SUMMARY:
Same as submitter's text.

PROBLEM CONCLUSION:
Development checked in a fix that resolved customer's problem.

------

APAR: IY39855 COMPID: 5765F7000 REL: 810
ABSTRACT: INTERNAL APAR FOR XLF COMPILER PTF 8.1.0.3

PROBLEM DESCRIPTION:
Internal APAR for XLF Compiler PTF 8.1.0.3

------

APAR: IY40096 COMPID: 5724C3505 REL: 310
ABSTRACT: OUTBND CALL BROKEN AS A RESULT OF SUPPORT MULTI CHANNEL IDS ATT

PROBLEM DESCRIPTION:
Outbnd call broken as a result of support Multi Channel IDs

PROBLEM SUMMARY:
Outbnd call broken as a result of support Multi
 Channel IDs ATT fileset

PROBLEM CONCLUSION:
corrected code handling outbound call
handling

------

APAR: IY40120 COMPID: 5724C3505 REL: 310
ABSTRACT: OUTBND CALL BROKEN AS A RESULT OF SUPPORT MULTI CHANNEL IDS DMS

PROBLEM DESCRIPTION:
Outbnd call broken as a result of support Multi Channel IDs

PROBLEM SUMMARY:
Outbnd call broken as a result of support Multi
 Channel IDs ATT fileset

PROBLEM CONCLUSION:
corrected code handling outbound call
handling

------

APAR: IY40121 COMPID: 5724C3505 REL: 310
ABSTRACT: OUTBND CALL BROKEN AS A RESULT OF SUPPORT MULTI CHANNEL IDS EURO

PROBLEM DESCRIPTION:
Outbnd call broken as a result of support Multi Channel IDs

PROBLEM SUMMARY:
Outbnd call broken as a result of support Multi
 Channel IDs ATT fileset

PROBLEM CONCLUSION:
corrected code handling outbound call
handling

------

APAR: IY40122 COMPID: 5724C3505 REL: 310
ABSTRACT: OUTBND CALL BROKEN AS A RESULT OF SUPPORT MULTI CHANNEL IDS INS

PROBLEM DESCRIPTION:
Outbnd call broken as a result of support Multi Channel IDs

PROBLEM SUMMARY:
Outbnd call broken as a result of support Multi
 Channel IDs ATT fileset

PROBLEM CONCLUSION:
corrected code handling outbound call
handling

------

APAR: IY40126 COMPID: 5724C3505 REL: 310
ABSTRACT: DEVICE DRIVER CRASH IN SIMPLE_LOCK

PROBLEM DESCRIPTION:
WVR Device Driver crashed in simple_lock.

PROBLEM SUMMARY:
WVR Device Driver crashed in simple_lock

PROBLEM CONCLUSION:
Add an extra check into the calculation of
the channel pointer by checking for NULL's along the way

------

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]