[as-devel] Bug Reports, 8 different ones! (long)
Frederick Bruckman (fb@enteract.com)
Mon, 30 Aug 1999 05:17:49 -0500 (CDT)
After updating to 1.7.143, I took the opportunity to experiment with
the configurations. This exposed has exposed certain problems that I
hadn't noticed before. I figure the least I can do is share these with
all of you.
I have two systems. One is a Mac Quadra 840AV running NetBSD-1.4.1,
using Ken Nakata's Mac68k_Color_Xserver, with 128M ram. This is a
BigEndian machine. The other is in two boxes, a 486DX4100 loghost and
NFS server exporting /usr, and a 486DX266 Xserver and NFS client, both
running NetBSD-1.4.1-release. The loghost has 48M ram, the Xserver
only 24M. This is still enough that the loghost only rarely touches
swap, the Xserver never. The Xserver is running XF86_S3 from XFree86
3.3.3.1, as distributed with the NetBSD release.
1) The gradients are wrong on the Quadra. I get a nice gold or copper
by shading grey to a purplish color. Sounds good to me, but when I
port the same scheme to the LittleEndian machine, copper and gold
become, (not surprisingly) purple!
2) Other than that, the gradients are fine on the Quadra, but the TCP
setup has other problems with them. Pager, Winlist, and Wharf will not
accept "BackGradient 1", but [2-4] are OK. The menus will accept no
gradients at all. In each case, the offender dumps core, and leaves a
stack trace thusly:
#0 0x4002802e in draw_gradient_line_dither ()
#1 0x40031060 in _GLOBAL_OFFSET_TABLE_ ()
#2 0x40028824 in L313 ()
#3 0x4002a072 in draw_gradient ()
#4 0x400258e6 in mystyle_make_icon ()
#5 0x40025987 in mystyle_make_icon_overlay ()
#6 0x4591 in CreateIconPixmap ()
#7 0x307a in main ()
3) The TCP setup will also not accept "MenuMiniPixmaps 1". Here's the
stack trace from afterstep.core...
#0 0x4004402e in draw_gradient_line_dither ()
#1 0x4004d060 in _GLOBAL_OFFSET_TABLE_ ()
#2 0x40044824 in L313 ()
#3 0x40046072 in draw_gradient ()
#4 0x400418e6 in mystyle_make_icon ()
#5 0x40041987 in mystyle_make_icon_overlay ()
#6 0x40041a8a in mystyle_make_pixmap_overlay ()
#7 0x16ee0 in setup_menu_pixmaps ()
#8 0x16c9b in configure_menu ()
#9 0x16a38 in do_menu ()
#10 0x13090 in ExecuteFunction ()
#11 0x10bec in HandleButtonPress ()
#12 0xf6ae in DispatchEvent ()
#13 0x12e30 in ExecuteFunction ()
#14 0x14a0a in ComplexFunction ()
#15 0x131a6 in ExecuteFunction ()
#16 0x3cd5 in main ()
That coredump might be from 1.7.130 or 1.7.137, not sure, but it does
look related to #2, anyway.
4) Following the suggestion in the man page of setting a gradient in a
"MyStyle" and then inheriting it doesn't seem to work. Tried it with
Wharf, but it's simply ignored. No errors in ~/.xsession-errrors.
5) I can't use the Pager to push a window partway off the top, but it
will go partway off the bottom, left or right. In 1.7.130-137 it would
go partway off the top, but wouldn't go partway off the bottom. In any
case, the windows can be moved, in all versions, as far off of the
page as I care to put them by using the "Move" menu item.
6) I would like to decry removal of support for the Shape Extension!
That worked fine in 1.7.137, both systems. I have discovered the
transparent options for BackPixmap, but the results aren't nearly as
satisfying. The icons don't pack closely, at best, and there are
annoying artifacts when the icons are moved, or when an app is started
iconified.
7) Perhaps this is just my goof, but I'm not getting how to "restore
apps to the correct window". The "StartsOnDesk N" in the database is
no longer respected. Moreover, the geometry saved in the
.workspace_state file is still relative to the page, so that all the
apps restart on page 0.
Lastly, one old bug:
8) The geometry in the .workspace_state is always the last option.
This breaks commands such as
xterm -n lynx -T lynx -geometry 96x30+0+18 -e lynx -color &
xterm -n pine -T pine -bg grey90 -fg blue4 -geometry 96x30+0+18 -e pine
because for "xterm -e foo", the "-e" must be the last option. I don't
want to tell you what happens when you try
xterm -e bash -ic "bash-function"
(same thing). Perhaps '-geometry' could simply be the first '-'ed opt?
Would that break something else?
I hope you all find this report useful. If there's something I can do
to help, please ask.
Regards