mcb@hpfcdc.HP.COM (Michael Berry) (05/23/89)
RELEASE 5.3 Defect Fix Update for s500 HP-UX This update contains updates to five major "subsystems" of HP-UX. This document describes each fix included in the release with the description keyed to the STARS report number for the corresponding defect. The list groups together all bug fixes for a given subsystem. For each subsystem the bug fixes are sorted by SR number. This document includes information on ALL updated subsystems, including separate products. There is no separate "Update_info" document for the Release 5.3 FORTRAN compiler. While all customers will receive a copy of this document, only customers who purchases a software update service for FORTRAN will receive the new compiler software. *************************************** ** SUBSYSTEM: COMMANDS / LIBRARIES ** *************************************** The following is the list of defects which we have fixed in Release 5.3 HP-UX. SR # Description 1650-038745 autobkup (1M) The "autobkup" utility fails when switching from the first to the second tape in an HP 35401A autochanger. (There is also a new "bkserver" program in Release 5.3 to address this problem.) 4000-090696 mount (1M) The "mount" command failed on device file names (such as "/dev/fd") that did not conform to the new naming convention (e.g. "/dev/dsk/0s0"). 4000-090704 autobkup (1M) There is a problem with switching to the second tape in the magazine of an HP 35401A. RESOLUTION: Duplicate of SR # 1650-038745 4000-090712 sh (1) The Bourne shell fails with a core dump if there is no environment when it starts up. RESOLUTION: Duplicate of SR # 5000-221259 4000-090720 LP models: dumb, hp2225a, hp2631g, hp2686a, hp2934a The model scripts contain C-language style comments. RESOLUTION: Duplicate of SR # 5000-231837 4000-090738 Some print jobs lost on heavily loaded system. RESOLUTION: Duplicate of SR # 5000-186981 4000-090746 mailx (1) It is not possible to "reply" to mail with an address of the form "user@localhost". RESOLUTION: Duplicate of SR # 5000-248765 4000-090753 tcio (1) The "tcio" command no longer uses "/dev/console" when invoked from "cron". Some users preferred this to the new behavior. (This release includes a "otcio" program which is "tcio" with the OLD semantics of writing to "/dev/console".) 4000-090761 grep (1) The 5.2 version of "grep" runs much slower than the 5.1 version of the same command. (The 5.1 versions had all data stored in a single data segment. The 5.2 version used two data segments. This slowed down access to its data structures.) 4000-092437 libm.a [pow.o] The double-precision power function "pow" incorrectly reports an overflow condition for underflow errors. REQUIRES: Re-link (if application accesses the "pow" function either directly or indirectly. (Note: the FORTRAN compiler will generate a call to this routine for expressions that involve exponentiation.) The following command will help you determine if your application uses "pow". $ nm a.out | grep pow 5000-186981 cancel(1), disable(1), enable(1), lp(1), lpstat(1), accept(1M), lpadmin(1M), lpmove(1M), lpsched(1M), lpshut(1M), reject(1M) It was possible to lose print jobs under certain circumstances, such as canceling a (different) job on a heavily loaded system. This was the result of a bug in the kernel support for FIFO files. 5000-221259 sh (1) It is not possible to boot from a recovery system tape built with the Release 5.2 version of "mkrs". (The version of the Bourne shell placed on the tape fails if there is no environment when it starts up; this is indeed the case when booting from a recovery system.) 5000-227769 mount (1M) The "mount" command issued an error message if invoked with the "-a" option. 5000-232181 vi (1) Interactive substitution no longer works in either "vi" or "ex". (The editors were trying to access the NLS message catalog file "info.cat" which was missing from Release 5.2 HP-UX.) 5000-231837 LP models: dumb, hp2225a, hp2631g, hp2686a, hp2934a These LP spooler model scripts will not work as shipped with Release 5.2 HP-UX. Each script started out with a version stamp in the form of a C-style comment instead of the Bourne shell syntax for a comment. (We corrected those syntax errors in this release.) 5000-238204 cron (1M) The "cron" daemon runs scheduled jobs twice on the day of a transition from Daylight Savings Time to Standard Time. See the updated "cron" Reference page for a discussion of this. 5000-241281 shutdown (1M) If you halt a system using "shutdown" with the "-h" option, you will see an error message about "init process termination". (There is a bug in the "reboot" program. The Release 5.3 version of "shutdown" uses "/etc/stopsys" to halt the system.) 5000-248765 mailx (1) It was not possible to respond (using the "r" command) to messages from users whose mail address was of the form "user@localhost". (The variable "optim" has been added to mailx so that when it is set, the "r" command generates optimized addresses. This feature is disabled by default.) 5000-423525 lpr (1) The "lpr" script failed if invoked with the "-r" option. (There was a syntax error in the script.) *************************************** ** SUBSYSTEM: Kernel ** *************************************** Release 5.3 includes one new capability (described in detail below) in the kernel. In addition, this release contains fixes for defects in the CIPER printer device driver (CIPERLP), 9-track mag tape device driver (HP7974.opt), and the SDF file system code in the operating system. SR # Description 1650-044610 When using a CIPER printer (e.g. HP 2608S or 256x) other HP-IB peripherals may hang when the CIPER printer is off-line. (The printer will go off-line if it runs out of paper, or the paper jams, or in case of a ribbon malfunction, or if the platen gate is open.) The CIPER driver now relinquishes the HP-IB card resource so that other peripherals can use it. It periodically checks the printer status and resumes spooling when the printer is ready. 4000-090738 Some print jobs lost on heavily loaded system. RESOLUTION: Duplicate of SR # 5000-186981 4000-090779 Using the "rename(2)" system call from s300 HP-UX (Release 6.0) or s800 HP-UX (Release 2.0) over an RFA connection to a Series 500 can corrupt part of the file system on the s500. (Note: commands that use include "ar", "compress", "ld", "mv", and "sh".) The corruption will appear as either two or more files (i.e. directory entries) referencing the same i-node, but with the link count set to one, or one or more directory entries pointing to a free i-node (i.e. the file contents are gone). 4000-090787 The HP 7974/78 9-track tape device driver prevents user program from reading past logical end-of-tape (EOT). This only happened under special circumstances such as backspacing over a tape mark and then reading past that same mark. 4000-094151 When several processes are simultaneously reading the same directory it is possible for one of these read operations to fail. This in turn could lead to a failure in higher level application code. The error occurs when reading a directory entry which spans two different data blocks in the file system. 5000-417931 The 9-track tape device driver (for streaming drives like the HP 7974 and 7978) now supports access of the new HP 7979 and 7980 tape drives. However, there is no support for enabling the data compression feature of the HP 7980XC via the device special file minor number. (You will need to activate it from the front panel of the tape drive.) 5000-186981 It was possible to lose print jobs under certain circumstances, such as canceling a (different) job on a heavily loaded system. *************************************** ** SUBSYSTEM: STARBASE Graphics ** *************************************** The following is the list of STARBASE graphics defects which have been fixed in Release 5.3 of HP-UX for the Series 500. The "REQUIRES" item in the defect list below indicates whether you need to relink your STARBASE program in order to pick up the fix. SR # Description 1650-049932 When calling "polyline2d" with a large array of coordinate info, STARBASE will "skip" data points. (Internally, STARBASE breaks up large arrays of coordinate info into buffers of 100 data points and sends each buffer to the display driver. There was an error in this segmenting code.) REQUIRES: Relink 5000-221341 Using the Model 550, when tracking on a second attached graphics station, the cursor echo appears on the FIRST attached graphics display. I.e. it is not possible to have independent tracking at two different "seats". REQUIRES: Rerun (with new STARBASE daemon installed) *************************************** ** SUBSYSTEM: DOCUMENTATION ** *************************************** Release 5.3 HP-UX includes updates to the following on-line HP-UX Reference entries. SR # Description 5000-196816 spell (1) The description of how to add words to the spelling list was in error. The new reference page contains a correct example of how to do this. 5000-238204 cron (1M) The reference page now discusses what happens on days when there is a transition from Daylight Savings Time to Standard Time. There was also an error in the Release 5.2 "Update_info" document in the description of how to use the "mvdevs" utility. This program converts device file names from the old H-P naming convention (e.g. "/dev/hd") to the newer System V naming convention (e.g. "/dev/dsk/0s0"). The text of the "Update_info" file included this sentence: ... (There should be NO character device for the root volume.) Ignore this statement, which came from a similar description for use with HP-UX on the Series 300. (On the Series 500, we do indeed invoke "fsck" with a character mode device file name as argument.) *************************************** ** SUBSYSTEM: FORTRAN ** *************************************** The following is the list of defects which have been fixed in Release 5.3 of the FORTRAN subsystem on the Series 500. We recommend that users recompile their FORTRAN programs to pick up all of these defect fixes. You can resolve some problems by relinking existing object files (i.e. those with ".o" suffixes) using the new FORTRAN I/O ("-lI77") and math ("-lF77") libraries. (The "fc" command will accomplish this when invoked with a list of object files rather than FORTRAN source files.) The "REQUIRES" item in the defect list below indicates whether you can pick up a fix by relinking or whether you must recompile source. SR # Description 1650-014969 Output of negative INTEGER*2 using octal format causes infinite loop RESOLUTION: Fixed in Release 5.1 HP-UX 1650-022749 <CONTROL>-D does not always terminate input RESOLUTION: User misunderstanding See the termio(7) manual page in the HP-UX reference. 1650-042200 Calling putenv(3) causes following invocation of sh(1) (using system(3)) to fail RESOLUTION: User misunderstanding The string passed to putenv(3) must be in a statically allocated variable, such as a COMMON variable, or a local variable in the main program. When setting up the variable make sure it is terminated with a null (by concatenating CHAR(0) to the end). 1650-045492 Core dump when compiling with "+Vd" and "-K" options REQUIRES: Recompile RESOLUTION: Duplicate of SR # 5000-292136 1650-069518 Implied-DO loops with COMMON variables cause data corruption REQUIRES: Recompile RESOLUTION: Duplicate of SR # 5000-292136 4000-035477 Printing 3-digit integer with format I6.2 causes infinite loop RESOLUTION: Fixed in Release 5.1 HP-UX 4000-090936 Wrong output from NAMELIST of INTEGER*2 variables REQUIRES: Recompile 4000-091215 Overflow errors on numeric input not detected REQUIRES: Relink 4000-091231 Illegal numeric input not detected REQUIRES: Relink 4000-091256 Wrong number of bytes written on unformatted output of BYTE expressions and LOGICAL*4 expressions REQUIRES: Recompile 4000-091389 Collapsed implied-DO loops can cause incorrect code to be generated for subsequent ordinary DO loops REQUIRES: Recompile 4000-091397 Collapsed implied-DO loops with nested implied-DO loops may generate incorrect code REQUIRES: Recompile 4000-091421 Wrong answer from ASINH intrinsic when argument is zero REQUIRES: Relink 4000-091538 Errors on READs from internal files cause nulls to be appended to standard output REQUIRES: Relink 4000-091579 EOF's not detected on READs from internal files REQUIRES: Relink 4000-091595 EOF's not detected on READs from internal files REQUIRES: Relink RESOLUTION: Duplicate of SR # 4000-091579 4000-091629 AINT, ANINT, DNINT functions give wrong answers for arguments larger than 2**31 - 1 (2147483647) REQUIRES: Recompile 4000-091637 NAMELIST output of BYTE and LOGICAL*1 variables is wrong REQUIRES: Recompile and relink 4000-091645 Erroneous compile errors caused by NAMELIST statements REQUIRES: Recompile 4000-091652 Error on a too large FORMAT statement not detected; FORMAT doesn't work properly either REQUIRES: Recompile 4000-092395 Hex output - truncation of leading F's not done correctly REQUIRES: Relink 4000-093070 "-K" option causes segmentation violation on formatted READs REQUIRES: Relink 5000-146514 Calling malloc(3) from FORTRAN doesn't work RESOLUTION: User misunderstanding If the memory requested from malloc is to be treated as a COMMON block, then the length requested from malloc must include an extra four bytes for the length word of the COMMON block (see page 33 in the Fortran Reference Supplement for the Series 500). 5000-162677 Tabs to the left do not work on terminal input REQUIRES: Relink 5000-204040 Incorrect input leaves a <CARRIAGE-RETURN> in the input stream REQUIRES: Relink 5000-221614 Can't use HP-UX file descriptor numbers > 2 REQUIRES: Relink 5000-226019 Unformatted WRITEs of INTEGER*2 output wrong data REQUIRES: Relink RESOLUTION: Duplicate of SR # 5000-230656 5000-226571 Unformatted READ of integer*2 array gets wrong data REQUIRES: Recompile 5000-226589 Unformatted WRITEs of INTEGER*2 output wrong data REQUIRES: Relink RESOLUTION: Duplicate of SR # 5000-230656 5000-228312 CLOSE statement with "ERR=" clause doesn't trap "out of paper" error REQUIRES: Relink 5000-230656 Unformatted WRITEs of INTEGER*2 output wrong data REQUIRES: Relink 5000-235275 Unformatted WRITEs of INTEGER*2 output wrong data REQUIRES: Relink RESOLUTION: Duplicate of SR # 5000-230656 5000-270678 Implied-DO loops may cause compiler to get a segmentation violation REQUIRES: Recompile 5000-292136 Implied-DO loops with SAVE variables cause data corruption REQUIRES: Recompile