home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
IRIX Installation Tools & Overlays 1999 May
/
SGI IRIX Installation Tools & Overlays 1999 May - Disc 2.iso
/
relnotes
/
x_eoe
/
ch4.z
/
ch4
Wrap
Text File
|
1999-04-19
|
14KB
|
463 lines
- 1 -
4. _B_u_g__F_i_x_e_s
+o _I_r_i_x _6._5 addresses a number of security issues with X
libraries; most of these are associated with buffer
overruns.
+o In previous releases, certain patterns of memory usage
within the X server could cause the system to run out
of swap space, such that the X server would ultimately
be killed by the kernel. This was fixed in _I_r_i_x _6._2.
Note that it is still possible to write an X client
that will cause the system to run out of memory/swap
space, for example by creating large pixmaps aaaadddd
iiiinnnnffffiiiinnnniiiittttuuuummmm without releasing them.
+o Input devices may kill the server during
initialization.
+o XReadDisplay does not work correctly for LG1 graphics
options.
+o Client programs get a BadValue error if they do a
read/modify/write of an event mask of an OpenGL window
with soft buffers.
+o On GR2/GR3 graphics, using XSetDashes with a non-zero
dash offset yields incorrect results.
+o An input device which is added after a device has been
removed during the same login session may be assigned
the same device ID as an existing device. This will
cause one or both devices to be unusable.
+o Image write operations could cause a context switch
crash on Impact graphics.
+o The X server could crash at reset for Reality Engine
graphics.
+o Calling XScreenSaverDisable can crash the X server.
+o Rendering failures for the 12-bit TrueColor visuals on
Impact graphics.
+o XReadDisplay was ignoring alpha on Impact graphics.
+o On Impact graphics, turning on/off video in a window
changed the window's visual.
+o Opaque stippled filled rectangles did not render
correctly on Impact graphics.
- 2 -
+o PEX rendering to a window crashes the X server if the
window is destroyed.
+o XWarpPointer produces erratic results with
PowerAnimator
+o On Impact graphics, text could be rendered with the
wrong color if the X server had previously rendered
24-bit TrueColor primitives.
+o Image read operations could cause a context switch
crash on Impact graphics.
+o XGetImage was reading back alpha planes on Impact
graphics.
+o XDrawString type operations were ignoring the GC's
logical function on Impact graphics.
+o Running OpenGL applications crashes the X server on a
High or Solid Impact system if the screen resolution is
1600x1200.
+o The X Input Extension DeviceButtonMotion event types
are not delivered.
+o Drawing wide lines with very large coordinates can
crash the X server.
+o XSGISetStereoBuffer did not work for 64-bit
applications.
+o PEX - problem with picking polylines.
+o Text strings rendered to pixmaps can be incomplete.
+o Rendering text to a pixmap using CID outline fonts can
generate incorrect results.
+o The X server could crash when running video tests or
XSGIVC applications.
+o On Impact graphics, GC rendering to an overlay window
using a GC that was created for an 8-bit normal plane
window can generate incorrect results.
+o Repeatedly calling XLoadFont/XUnloadFont for a CID font
can crash the X server.
+o Backing store support for OpenGL pbuffers can fail on
Impact graphics.
- 3 -
+o On Impact graphics, enabling the Impact Channel Option
should cause the non-ICO screen to be blanked.
+o On Impact graphics, the Impact Channel Option does not
blank/unblank when the screen saver is set up for
blanking.
+o Pbuffers save/restore code on Impact graphics was
attempting to lock down too much memory, resulting in a
catastrophic failure.
+o On Impact graphics, in the case of a dual head system
there was a 16K memory leak.
+o The X server could crash when backing store was
enabled.
+o XSGIvcListVideoFormatCombinations returns BadName
error value even when the name is correct.
+o Overlay Rendering is incorrect in Impact Graphics
systems, if the ICO hw is not installed but specified
as the default timing table to be loaded at the server
startup time.
+o Some PEX primitives render incorrectly, and some data
returned to a little-endian PEX client may not be
properly byte swapped.
+o On GR2/GR3 graphics, if you draw a stippled, filled
primitive with foreground color equal to the background
color (effectively yielding a solid fill), then switch
colors so that foreground and background are no longer
equal, subsequent stippled, filled primitives rendered
with this GC are rendered incorrectly as solid filled.
+o On Impact graphics, the system could hang when
resetting the X server and then blanking the screen.
+o When using the XSGI Stereo Extension, issuing GL calls
to a severely clipped window could yield incorrect
results if these calls occured after going into stereo
mode but before the first call to XSGISetStereoBuffer.
+o Calling XSGISetStereoBuffer for a non-GL window would
crash the X server.
+o The X server was ignoring the IgnoreLockMods settings
in /usr/lib/X11/xkb/X0-config.keyboard
- 4 -
+o On Indy and Indigo2 XL systems, Xsgi does not maintain
its pre-allocated cells, it is possible for a client to
be given one of the pre-allocated cells as read/write
during an XAllocColorCells.
+o On a dual-head Impact system, when dynamically
switching timing tables, a ghost cursor could appear
on the head where the new timing table was installed.
+o The matrix transformations of outline fonts specified
in X font names did not work.
+o Rendering clipped, zero-width lines to a pixmap could
crash the X server.
+o On Impact graphics, switching dynamically to a lower
screen resolution displayed the top left part of the
framebuffer, instead of the bottom left.
+o XDrawArc() with large width & height yields incorrect
results.
+o On NG1 graphics, repeated X server logins cause
vertical lines on monitor.
+o Calling XShapeCombineMask with a NULL source region
could cause the X server to crash.
+o XDrawArcs() draws does not assign proper pixels to line
when large widths are used as value.
+o On Impact graphics, performance of glXMakeCurrent was
poor when switching between multiple overlay windows.
+o Screen dimensions (in mm) returned for formats larger
than 1280x1024 were incorrect.
+o Xsgi does not set cursor bounds properly on
XWarpPointer.
+o Long font pattern can crash X server.
+o On Octanes in 4-screen OCO mode, cursor updates could
be delayed for as much as several seconds.
+o On Octanes during screen blanking, Xserver was not
properly programming one of the DACs used on the OCO.
+o On O2, after server reset (eg after a user would log
out and log back in again), the server would fail to
put monitor into powersave mode.
- 5 -
+o On O2, a GLX mixed program with GXxor X rendering gets
incorrect results.
+o NumLock stops working after loading a new keymap using
the setxkbmap command. If the personal keyboard
setting is changed using ipanel, the system runs
setxkbmap automatically on login, killing NumLock.
+o The group count is not updated when a keymap is loaded
using the setxkbmap command. This can cause real
problems when replacing a single-language keyboard
layout with a two language layout - because the group
count isn't updated, the second language is not
accessible.
+o NumLock stops working in Motif-based applications on
some European keyboards, after typing certain keys on
the main part of the keyboard (e.g., the letter 'a' on
a French keyboard).
+o On O2, when a software double buffered window is moved,
the back buffer content is not preserved, since back
buffer tiles are dynamically allocated and freed,
delayed free mechanism has to be added into CRM ddx
which won't free the back buffer tiles until moving
operation is done (i.e. until CopyWindow is serviced).
+o On O2, multiple requests to draw the same dashed
segment produce inconsistent results. This was most
readily seen with VkGraph.
+o On O2, for PseudoColor visuals, users could experience
more colormap flashing than normal.
+o On O2, StoreColors to an uninstalled DirectColor
colormap crashes the X server
+o On O2, rendering of large spans into a drawable with a
complex clip region is very slow.
+o On O2, the X server crashes after an X font server
dies.
4.1 _B_u_g__f_i_x_e_s__i_n__6_._5_._1
+o Starting with _I_r_i_x _6._5, the file /var/X11/xdm/Xservers
is now installed as a suggested upgrade, which means
that if the previous version was modified by the user,
the new version is installed as Xservers.N (see
_U_p_d_a_t_i_n_g _C_o_n_f_i_g_u_r_a_t_i_o_n _F_i_l_e_s in the _v_e_r_s_i_o_n_s man page).
(560630)
- 6 -
+o In the Traditional Chinese locale (zh_TW.ucns), the
output method does not properly display characters in
codeset 2, from the CNS 11643.1986-2 character set.
+o pound,$ and brace key don't work with de_CH & fr_CH
keyboard (562278)
+o Getting ">" when typing "<" on a text field in a java
applet. (573309)
+o On O2, running snapshot over 12-bit pseudoColor windows
could cause the X server to crash (574967)
+o On O2, Colormap needs to be initialized to black when
the X server is reset to minimize visual problems with
displaying uninitialized colormap values (577226)
+o DM pbuffers were erroneously being reported as
supported on Impact, RealityEngine and InfiniteReality
graphics. (598882)
+o Security problems with xterm (600678), libXaw (600686),
libXt (600690, 609039, 609104), libX11 (607710, 608693,
608806, 608998, 609085, 614197).
+o Support for the O2 Dual Channel Display option,
including support for DDC monitor communications with
the dual channel monitors (607902)
+o The Xt default error handler could end up in an
infinite loop if it's trying to report an out-of-memory
error condition (609236)
+o XReadDisplay (and the Iris GL command readdisplay) was
reading back the wrong information for overlay windows
on XZ/Extreme graphics (610999)
+o Security problem with fonts.dir file (612888)
4.2 _B_u_g__f_i_x_e_s__i_n__6_._5_._2
+o Security problems with libXaw (627441), libXmu (627136,
627543), libX11 (625464, 627631, 628917, 628942).
+o Security problem with xdm (639086).
+o Added some monitors to /var/X11/Xvc/moninfo (611033)
+o On O2, XReadDisplay returned pixel values with lower
intensities than actual for 15-bit TrueColor visuals
(624305)
- 7 -
+o Keyboard bell wouldn't work for the second head of a
dual-head Octane. This was a regression from Irix 6.5,
introduced in Irix 6.5.1 (628562).
4.3 _B_u_g__f_i_x_e_s__i_n__6_._5_._3
+o Fix so NDS slave display can be specified in
configuration file, and configuration file settings can
be overridden by command line options. (429053)
+o The xwsh clone option on its popup menu is inoperative
if the xwsh was started up in VT100 mode (640704).
+o The X server on Impact systems does not release free
memory back to the system (641338).
+o The X server sometimes does not release free memory
back to the system (641347).
+o Clicking on the root window when running xshowcmap
-inter can result in a protocol error (651090)
+o When running an X connection through LBX, the values
for BlackPixel and WhitePixel are swapped, leading to a
"reverse video" look for apps that use these values
(651755)
+o lbxproxy ignores the DISPLAY environment variable
(655286).
+o Euro Symbol and ISO8859-15 character map were added to
several fonts. (651016)