Non-root applications sometimes use all available swap space, which can result in a hung system or cause critical system daemons to die.
Workaround: Because the VM subsystem reserves a small amount of swap space that is in the form of physical memory for root processes, it enables system administrators to log in and kill offending user processes that have used all available swap space. Therefore, system hangs are prevented and critical system daemons can keep running.
Because non-root processes cannot reserve swap space against this pool, some non-root processes may no longer be available with ENOMEM whenever you type the df or swap -s command and a message is displayed indicating that a small amount of swap space is still available.
If a system that you need to upgrade has overlapping slices on any of the attached disks (excluding the backup slice), the upgrade configuration process fails. If you are using Disk Space Reallocation, the auto-layout process fails.
Workaround: Change or delete the overlapping slices by using the format command.
Note that changes to slices will cause data loss on both the slice being changed and on any slices that is being overlapped by the changed slice. Any data on the affected slices should be backed up before slice manipulation is attempted.
The above-listed message is displayed as is or a variation of this message may be displayed whenever you reboot after a cpr Suspend.
The criteria for using cpr are:
root must not be a logging filesystem
The file system selected for the cpr state file (see power.conf(4)) must not be a logging file system.
The NFS protocol reports the modification, access, and meta data change times on files as 32-bit unsigned value that counts the number of seconds elapsed since January 1, 1970, as an unsigned value. This gives NFS a theoretical upper limit to the Year 2106.
Starting with the Solaris 7 release, the Solaris NFS client and server no longer mistakenly render negative times on files as unsigned values. Thus an NFS client that conformed to the NFS protocol specification would interpret that a file last modified in 1969 was in fact modified in 2106.
In the Solaris 7 release, the NFS client and server implementation has been corrected to adhere to the protocol. Whenever the NFS server encounters a file with a negative time, it refuses to provide attributes for it to the NFS client.
Therefore, no longer will files with negative times be supported over an NFS mount. When the NFS client tries to access a file with negative times, the client typically gets an "overflow" error.
Also, one can have a tar (see the tar(1) command), or cpio (see the cpio(1) command) archive that has negative times stored in the archive. When extracting this archive over NFS, it no longer works if using a Solaris 7 NFS client.
Workaround: The workaround to the first problem is to log onto the NFS server that has the file (the nfsstat(1M) utility, invoked with the -m option, identifies the server), and use the touch(1) command to change the file times to the current date and time.
Binary compatibility of existing Motif binaries is not affected. Many parameters in the Motif Application Programming Interface (API) are of type XmString, which is an opaque type. Motif source code that uses char* where the Motif API expects an XmString parameter is always incorrect. Such incorrect Motif code is more likely to core dump when compiled in the Solaris 7 operating environment.
Workaround: Use an older version of Motif.
If your system's LC_TIME environment variable is set to a locale different from the locale that is set in the LC_MESSAGES environment variable, then the date prompt displays a format that is not compatible with what the CDE Mailer expects.
Workaround: Before you start your vacation mail, make sure that your system's LC_MESSAGES and LC_TIME environment variables are set to the same locale.
The Copy/Paste functions do not work properly if you try to copy and paste any text that includes Hebrew and Arabic characters and numbers, such as 1, 2, 3, 4, etc.
If you are using dtmail in either Hebrew or Arabic locales and you have attachments with Hebrew or Arabic names, these names are displayed in reversed order in the dtmail attachment.
If you are using dtcm in either Hebrew or Arabic locales, the name of the Month in the Calendar Manager icon on the front is incorrectly reversed.
Dtmail may truncate the body of an email message if the message body contains non-printable characters.
The following warning message is displayed:
Message contains a binary code which is invalid as a character in the current locale. For this reason part of this message may be truncated. If you run CDE in the same locale that this message came from, you will be able to view the entire message. |
If an email message is created in one locale and then emailed to another locale, email may be truncated because it contains a binary code that is invalid as a character in the locale to which the email was sent.
For example, this bug may occur when:
email composed in the ko locale and emailed to the ko_UTF.8 locale
email composed in the zh locale and emailed to the zh.GBK locale
Workaround: Perform the following steps:
Exit CDE.
Log into CDE in the locale where the email was composed
or
quit dtmail and set the environment variable LANG to the correct locale; then restart dtmail.
The same data that is sent as an attachment is not truncated when viewed.
If you open dtmail in the Arabic locale on your Common Desktop Environment, the Arabic date and time is neither correctly shaped nor correctly laid out. Also, if you insert Arabic characters in a new message dialog, it results in a row of Arabic characters that is displayed without being correctly shaped or laid out.
If you select ar, the Arabic locale, from the login screen, the dtlogin is displayed with a very large font. The default font is also very large after signing on.
This bug only affects Motif programs that have been compiled on systems that run the Solaris 2.5.1 operating environment and also contain a custom Motif widget that subclasses the XmText widget. Such a program may cause a core dump when it is executed on a system with the Solaris 2.6 or Solaris 7 operating environment running.
Workaround: Recompile the program on a system with either the Solaris 2.4, 2.5, 2.5.1, 2.6 or Solaris 7 operating environment running.
You are unable to view any appointments in the Calendar Manager in Arabic when you select dtcm in the Common Desktop Environment.
Workaround: Perform the following steps:
Select ar, the Arabic locale, from the dtlogin screen.
Start dtcm to open the Appointment Editor.
Add a new appointment that includes Arabic text.
Close the Appointment Editor.
The new appointment is now displayed in the Calender Manager.
If you deploy centralized calendar services where users have access to their Calendar from a remote centralized server, then the Calendar requires about ten times the total size of Calendar files in swap space to work reliably.
For example, if users want to have access to Calendar from the same server and the Calendar files' size totals 50 MB, then Calendar requires 500 MB of swap space to work reliably.
If these space limitations are not considered during the installation of Solaris, users may have unreliable access to the Calendar.
Workaround: Increase swap space to enable the Calendar to work reliably.
If you are currently using the MWM, upgrade to CDE desktop. The CDE desktop provides superior functionality and the transition is not difficult. The following information is provided if you cannot upgrade from the MWM to the full CDE desktop.
If you have developed applications using the Motif Development kit that shipped with the 2.5 Software Developer's Kit, you can now use dtwm, the Windows Manager, that is part of CDE. The Windows Manager can be used just like the MWM. The Windows Manager also supports resource files for the MWM. The following actions can be taken to make the Windows Manager appear similar to an existing MWM configuration.
Turn off the dtwm front panel using the following settings:
Dtwm*useFrontPanel: False |
This resource can be set in .Xdefaults, or the existing resource can be changed from True to False in the file /usr/dt/app-defaults/C/Dtwm
The dtwm option -name allows the use of most existing mwm resources. The -name option is documented in the dtwm man page.
MWM customers may or may not depend on the following file: $HOME/.mwmrc. There is a dtwm resource config file that can be set to make dtwm read the file $HOME/.mwmrc. This resource is documented in the dtwm man page.
The root menu for dtwm and mwm is not the same. The dtwm root menu is fully configurable. You can make any changes in the root menu. See the man pages for dtwm and dtwmrc.
The CDE Mailer needs twice the size of a mailbox in free memory to open the mailbox. For example, if you have a 150 MB mailbox, you need at least 300 MB of free memory to open the mailbox.
Workaround: Increase your swap space to enable your mailbox to be opened.
Java applications dump core when there is not enough swap space. The application indicates the following bus error:
not enough space |
Workaround: Exit other applications running on the system or add more swap space. See the System Administration Guide, Volume I for details on adding swap space.
The Java Thread.suspend() method may cause Java application programs to hang. It is strongly recommended that you do not use the suspend() and resume() methods.
Hangs may occur because a thread holding a lock is suspended and the thread responsible for resuming this thread needs this lock. This is a general problem with threaded programming because of improper usage of these primitives, which leads to application deadlocks.
Workaround: Use other appropriate synchronization methods, such as wait() and notify().
java.sql.Timestamp accepts a two-digit year but does not allow for years beyond 1999 (it interprets a two-digit date as 19xx). This problem does not occur when a four-digit year is specified.
The java.sql.Timestamp class extends the java.util.Date class where several methods have been deprecated. To ensure forward compatibility, use the java.text.SimpleDateFormat methods.
Random bus error or segmentation fault occurs in an application or support library.
Workaround: Use java_g or java -debug to run this program. Also, try turning the JIT off:
setenv JAVA_COMPILER NONE |
Any of these workarounds change the execution profile of the VM.
Numerous X-requests in Java applications sometimes cause a slow startup in 8-bit color (TrueColor or PseudoColor) environments. This is especially apparent on slower systems. A large part of the startup delay is due to the computation of an optimal dithered colormap by the awt library during initialization.
Workaround: Adjust the colormap size computed at initialization by using the environment variable VIRTCUBESIZE. Set it to a power of 2 between 4 and 32. The default value is 32.
setenv VIRTCUBESIZE 8 |
When -nojit is used, the following warning is displayed:
Warning: JIT compiler "none" not found. Will use interpreter |
Workaround: Set -Djava.compiler=NONE
The following warning message is displayed when you try to compile a 64-bit XViewTM or OLIT application:
Creating 64-bit applications in {XView, OLIT} is not supported. |
To prevent user programs from encountering a bug that causes certain 64-bit user applications to stall when executing instructions loaded between virtual addresses 0 (0x0) and 2G (0x80000000), the system automatically ensures that 64-bit applications do not load instructions in that range. This does not reduce the amount of physical memory available to applications, nor does it prevent user data from being loaded in that range. 32-bit applications are not affected.
A user's multi-threaded applications can hang when such the application causes an _XRead to be issued from Xlib.
Workaround: You can either:
Not use multiple threads
or
Restrict all access to Xlib routines to only one of the multiple threads.
The photo CD I/O device does not support an xil_get_pixel operation.
Workaround: If you need to obtain the value for a particular pixel of a photo CD image, copy the photo CD image to a memory image and then call xil_get_pixel on the memory image.
The ImageViewer (sdtimage) can freeze in limited circumstances during the manipulation of an image on a screen. This problem only occurs when XIL is using the X Shared Memory (xshm) display pipeline. A thread deadlock may be the cause of the problem between XIL and xshm.
The xshm pipeline is only used when the workstation framebuffer does not have an accelerated XIL display driver available, such as the TCX and ZX framebuffers. Existing accelerated XIL drivers for cg6(GX), cg14(SX), ffb(Creator) and afb(Elite) framebuffers do not cause this problem.
Workaround: If this problem occurs, use the Image Tool (/usr/openwin/bin/imagetool) as a substitute for sdtimage.
When an XGL Gcache decomposes a polygon with edges, some edges are rendered incorrectly for xgl_gcache_multi_simple_polygon().
Workaround: Set the XGL_GCACHE_USE_APPL_GEOM to TRUE. Alternately, use xgl_gcache_polygon() instead of xgl_gcache_multi_simple_polygon() and turn off the XGL_GCACHE_SHOW_DECOMP_EDGES when creating the GCACHE context.
With XViewTM (and possibly other tool kits) calling the application's resize_proc() before sending the resize request to the server, the window size returned by DGA is behind by one resize event.
Workaround: Calling XSync() before calling xgl_window_raster_resize() fixes the problem.
AnswerBook2 server administrative access authentication fails if you use a CGI-based http server instead of the default NSAPI server.
Workaround: Perform the following steps:
Become root on the document server.
Turn off administrative access control for the document server by typing the following command:
# /usr/lib/ab2/bin/ab2admin -o access_off |
If you still need to control access to administrative functions on the document server, use available commands on your http server to restrict access to the following URL:
/ab2/@Ab2Admin |
3Com EtherLink III 3C905B cards in a Compaq ProLiant 6500 can fail to generate interrupts. The card sends packets but does not receive them.
Workaround: There is no known workaround for this problem. However, because some slots appear to have this problem occur more frequently than other slots, moving the card to another PCI slot might help. Also, successive reboots of the system have succeeded in getting the card out of this state.
Some PCI motherboards contain DMA chipsets that are unable to support 100-Mbps Fast Ethernet.
Workaround: See "100-Mbps Ethernet Performance" in the Solaris 7 (Intel Platform Edition) Device Configuration Guide for additional information and recommended solution.