[3757] in testers
Re: Mary-kay-commandos: 8.2.6: Bug in ez
daemon@ATHENA.MIT.EDU (John Hawkinson)
Mon Jul 13 00:26:35 1998
Date: Mon, 13 Jul 1998 00:26:25 -0400
To: Calista E Tait <cat@MIT.EDU>
Cc: testers@MIT.EDU
In-Reply-To: "[3754] in testers"
From: John Hawkinson <jhawk@MIT.EDU>
| I loaded the file, I enabled basic styles, clicked on smaller about
| 4-6 times and the X server crashed. This is reporducible.
Running Xsun with -core and removing /core (A symlink, which root
won't follow under Solaris 2.6, eit!), I got a core, which is in
~coredumps/jhawk/Xsun.826.2.core.gz. [*]
If system:quality feels inclined to remove the other files in
~coredumps/jhawk, I wouldn't mind.
Anyhow, stack trace from adb is:
SIGSEGV: Segmentation Fault
Unable to open shared library X
$c
sunGXPolyGlyphBlt(0x1940a8,0x4,0xef180000,0x125,0x5,0xeffff57c) + 358
miPolyText8(0x1951b8,0x194038,0x101,0x125,0x7,0x1dba12) + 80
miSpritePolyText8(0x1951b8,0xef739f7c,0x1940cc,0xc5d40,0x5,0x1dba12) + 180
doPolyText(0x1dba17,0xeffffa98,0x1,0x1db9c0,0x0,0x0) + 418
PolyText(0x1e3898,0x1951b8,0x194038,0x1dba10,0x1dba18,0x101) + 74
ProcPolyText(0x194038,0x1e3898,0x1951b8,0x1dba00,0x12,0x8) + 1e0
Dispatch(0x1e3898,0xc167c,0xba550,0xba548,0xc1400,0xc1400) + 194
main(0x4,0x1,0xba85c,0xc1dfc,0x1,0x0) + 5ec
This seems peculiar since the current instruction doesn't seem messed up:
<pc?2i
sunGXPolyGlyphBlt+0x358: ld [%i2 + 0x10], %o1
andcc %o1, %o0, %g0
<i2/X
0xef180000: 0
<i2+10/X
0xef180010: 0
Actually attaching to the process is slightly different:
0t9607:A
$c
?(0x1984c0,0xa,0xef180000,0x102,0xb,0xeffff594) at eeed8638
?(0x19a9f0,0x198450,0x2f,0x102,0x12,0x1807ee) at ef4a90e8
?(0x19a9f0,0xef739f7c,0x1984e4,0xc5d40,0xb,0x1807ee) at ef4b3a54
doPolyText(0x1807f9,0xeffffa98,0x1,0x1807d4,0x0,0x0) + 418
PolyText(0x1a7ae0,0x19a9f0,0x198450,0x1807ec,0x1807fc,0x2f) + 74
ProcPolyText(0x198450,0x1a7ae0,0x19a9f0,0x1807dc,0x14,0x8) + 1e0
Dispatch(0x1a7ae0,0xc167c,0xba550,0xba548,0xc1400,0xc1400) + 194
main(0x4,0x1,0xba85c,0xc1dfc,0x1,0x0) + 5ec
.<pc/2i
0xeeed8638: ld [%i2 + 0x10], %o1
andcc %o1, %o0, %g0
.<i2/X
0xef180000: 61b009ff
.<i2+10/X
0xef180010: 18a12
.
Of course, it looks like everything is mapped from our perspective.
This is weird. Is it always like this on the SS5?
Using gdb claims the memory isn't mapped, though.
--jhawk
[*] - This is a core dump from a SPARCstation-5. It appears that
this isn't read properly by adb on an Ultra-1. I don't know how to
account for this other than it might be an adb bug. Developers should
probably be aware of this, however. gdb on the ultrasparc doesn't
seem to have a problem, but it's the gnu locked gdb, so it's for solaris
2.3, so...
I'll probably look into this and maybe report it to Sun, but
someone else should deal with the X server bug.
