[#5522] Lib DirectFB-1.4.0 crashes when run DirectFB demo on SVN trunk
Submitted By: Sonic Zhang
Open Date
2009-09-11 06:47:53 Close Date
2009-12-23 02:21:47
Priority:
Medium Assignee:
Barry Song
Jie Zhang
Status:
Closed Fixed In Release:
N/A
Found In Release:
2010R1 Release:
Category:
N/A Board:
Processor:
BF548 Silicon Revision:
Is this bug repeatable?:
Yes Resolution:
Fixed
Uboot version or rev.:
Toolchain version or rev.:
svn trunk r3626
App binary format:
Summary: Lib DirectFB-1.4.0 crashes when run DirectFB demo on SVN trunk
Details:
root:/>
root:/> df_input &
[1] 408 df_input
~~~~~~~~~~~~~~~~~~~~~~~~~~| DirectFB 1.4.0 |~~~~~~~~~~~~~~~~~~~~~~~~~~
(c) 2001-2009 The world wide DirectFB Open Source Community
(c) 2000-2004 Convergence (integrated media) GmbH
----------------------------------------------------------------
(*) DirectFB/Core: Single Application Core. (2009-09-11 10:38)
(*) Direct/Thread: Started 'VT Switcher' (412) [CRITICAL OTHER/OTHER 0/0] <12288>...
(*) Direct/Thread: Started 'VT Flusher' (413) [DEFAULT OTHER/OTHER 0/0] <12288>...
(*) DirectFB/FBDev: Found 'bf54x-lq043' (ID 0) with frame buffer at 0x03e01000, 382k (MMIO)
(*) Direct/Thread: Started 'Keyboard Input' (414) [INPUT OTHER/OTHER 0/0] <12288>...
(*) DirectFB/Input: Keyboard 0.9 (directfb.org)
(*) Direct/Thread: Started 'Linux Input' (415) [INPUT OTHER/OTHER 0/0] <12288>...
(*) DirectFB/Input: bf54x-keys (1) 0.1 (directfb.org)
(*) Direct/Thread: Started 'Linux Input' (-1) [INPUT OTHER/OTHER 0/0] <12288>...
(*) DirectFB/Input: ADXL34x accelerometer (2) 0.1 (directfb.org)
(*) DirectFB/Graphics: Generic Software Rasterizer 0.6 (directfb.org)
(*) DirectFB/Core/WM: Default 0.3 (directfb.org)
(*) FBDev/Mode: Setting 480x272 RGB24
(*) FBDev/Mode: Switched to 480x272 (virtual 480x272) at 24 bit (RGB24), pitch 1440
(*) FBDev/Surface: Allocated 480x272 24 bit RGB24 buffer (index 0) at offset 0 and pitch 1.
t*) FBDev/Surface: AllocateJu48mp27 t24o itNUGBLL b afedd(ireexss)
Te*)ptiriot/n tecoacnt Lexdet '
tCUimRRemENtaT onPRf OCDiESctS:Fo
C
OMM=df_input PID=408
CPU = 0
TEXT = 0x00370000-0x00374560 DATA = 0x00376560-0x003777e8
BSS = 0x003777e8-0x00820000 USER-STACK = 0x0083feb0
return address: [0x(null)]; contents of:
ADSP-BF548-0.2 525(MHz CCLK) 131(MHz SCLK) (mpu off)
Linux version 2.6.30.6-ADI-2010R1-pre-svn7295 (root@eight) (gcc version 4.1.2 (ADI svn)) #9
SEQUENCER STATUS: Not tainted
SEQSTAT: 0000002d IPEND: 0008 IMASK: ffff SYSCFG: 0006
EXCAUSE : 0x2d
physical IVG3 asserted : <0xffa0069c> { _trap + 0x0 }
RETE: <0x00000000> /* Maybe null pointer? */
RETN: <0x00860000> [ /lib/libpthread.so.0 + 0x0 ]
RETX: <0x00000480> /* Maybe fixed code section */
RETS: <0x008e5ef0> [ df_input + 0x1ef0 ]
PC : <0x00000000> /* Maybe null pointer? */
DCPLB_FAULT_ADDR: <0x008e7854> [ df_input + 0x3854 ]
ICPLB_FAULT_ADDR: <0x00000000> /* Maybe null pointer? */
PROCESSOR STATE:
R0 : 00375864 R1 : 00000002 R2 : 009ea274 R3 : 00375864
R4 : 00000008 R5 : 0033d8fc R6 : 008e7858 R7 : 009ea274
P0 : 009ea274 P1 : 00000000 P2 : 009ea10c P3 : 008e7860
P4 : 008ee638 P5 : 0033d808 FP : 0083f9fc SP : 0085ff24
LB0: 008d7cbd LT0: 008d7cb0 LC0: 00000000
LB1: 0323e3c1 LT1: 0323e3c0 LC1: 00000000
B0 : 00000000 L0 : 00000000 M0 : ffffffc8 I0 : 00000000
B1 : 00000001 L1 : 00000000 M1 : 00000000 I1 : 00000000
B2 : 00000000 L2 : 00000000 M2 : 00000000 I2 : 00000000
B3 : 00000000 L3 : 00000000 M3 : 00000000 I3 : 00000001
A0.w: 00000666 A0.x: 00000000 A1.w: 00000666 A1.x: 00000000
USP : 0083f9d8 ASTAT: 02003025
Hardware Trace:
0 Target : <0x00004fc0> { _trap_c + 0x0 }
Source : <0xffa00630> { _exception_to_level5 + 0xa4 } CALL pcrel
1 Target : <0xffa0058c> { _exception_to_level5 + 0x0 }
Source : <0xffa00440> { _bfin_return_from_exception + 0x18 } RTX
2 Target : <0xffa00428> { _bfin_return_from_exception + 0x0 }
Source : <0xffa004e4> { _ex_trap_c + 0x74 } JUMP.S
3 Target : <0xffa00470> { _ex_trap_c + 0x0 }
Source : <0xffa006f6> { _trap + 0x5a } JUMP (P4)
4 Target : <0xffa0069c> { _trap + 0x0 }
Source : <0x008e502a> [ df_input + 0x102a ] JUMP (P1)
5 Target : <0x008e5022> [ df_input + 0x1022 ]
Source : <0x008e5eec> [ df_input + 0x1eec ] CALL pcrel
6 Target : <0x008e5ed0> [ df_input + 0x1ed0 ]
Source : <0x009228dc> [ /usr/lib/directfb-1.4-0/systems/libdirectfb_fbdev.so + 0x28dc)
7 Target : <0x009228d4> [ /usr/lib/directfb-1.4-0/systems/libdirectfb_fbdev.so + 0x28d4]
Source : <0x009268da> [ /usr/lib/directfb-1.4-0/systems/libdirectfb_fbdev.so + 0x68daL
8 Target : <0x009268c8> [ /usr/lib/directfb-1.4-0/systems/libdirectfb_fbdev.so + 0x68c8]
Source : <0x009268b4> [ /usr/lib/directfb-1.4-0/systems/libdirectfb_fbdev.so + 0x68b4P
9 Target : <0x009268a0> [ /usr/lib/directfb-1.4-0/systems/libdirectfb_fbdev.so + 0x68a0]
Source : <0x008d352e> [ /usr/lib/libdirectfb-1.4.so.0 + 0x5352e ] CALL (P1)
10 Target : <0x008d34d2> [ /usr/lib/libdirectfb-1.4.so.0 + 0x534d2 ]
Source : <0x008d34fc> [ /usr/lib/libdirectfb-1.4.so.0 + 0x534fc ] IF CC JUMP
11 Target : <0x008d34d3> [ /usr/lib/libdirectfb-1.4.so.0 + 0x534d3 ]
Source : <0x008d35f4> [ /usr/lib/libdirectfb-1.4.so.0 + 0x535f4 ] JUMP.S
12 Target : <0x008d35e8> [ /usr/lib/libdirectfb-1.4.so.0 + 0x535e8 ]
Source : <0x008d350f> [ /usr/lib/libdirectfb-1.4.so.0 + 0x5350f ]
13 Target : <0x008d34de> [ /usr/lib/libdirectfb-1.4.so.0 + 0x534de ]
Source : <0x008d34d0> [ /usr/lib/libdirectfb-1.4.so.0 + 0x534d0 ] JUMP.S
14 Target : <0x008d347e> [ /usr/lib/libdirectfb-1.4.so.0 + 0x5347e ]
Source : <0x008d3442> [ /usr/lib/libdirectfb-1.4.so.0 + 0x53442 ] IF CC JUMP
15 Target : <0x008d3408> [ /usr/lib/libdirectfb-1.4.so.0 + 0x53408 ]
Source : <0x00890716> [ /usr/lib/libdirectfb-1.4.so.0 + 0x10716 ] JUMP (P1)
Follow-ups
--- Vivi Li 2009-10-14 05:13:19
I also find it crash in bf527-ezkit when run command "df_dok
--noresults".
--- Mike Frysinger 2009-10-14 06:15:48
i recently (last week) tested DirectFB 1.4.2 on the BF548-EZKIT and didnt notice
any problems with it
--- Sonic Zhang 2009-10-14 06:25:01
Did you run df_dok and df_input, not all directfb demoes crash.
--- Mike Frysinger 2009-10-15 16:01:16
df_dok is generally what i run when checking DirectFB. df_input doesnt crash
either.
root:/> df_dok
~~~~~~~~~~~~~~~~~~~~~~~~~~| DirectFB 1.4.2 |~~~~~~~~~~~~~~~~~~~~~~~~~~
(*) DirectFB/Core: Single Application Core. (2009-10-15 19:40)
(*) Direct/Thread: Started 'VT Switcher' (257) [CRITICAL OTHER/OTHER 0/0]
<12288>...
(*) Direct/Thread: Started 'VT Flusher' (258) [DEFAULT OTHER/OTHER 0/0]
(*) DirectFB/FBDev: Found 'bf54x-lq043' (ID 0) with frame buffer at 0x03e00000,
382k (MMIO 0x00000000, 0k)
(*) Direct/Thread: Started 'Keyboard Input' (259) [INPUT OTHER/OTHER 0/0]
(*) Direct/Interface: Loaded 'PNG' implementation of 'IDirectFBImageProvider'.
(*) Direct/Interface: Loaded 'FT2' implementation of 'IDirectFBFont'.
(*) Direct/Interface: Loaded 'GIF' implementation of 'IDirectFBImageProvider'.
Benchmarking 256x252 on 480x262 RGB24 (24bit)...
Anti-aliased Text 3.040 secs ( 39.078
KChars/sec) [100.0%]
Anti-aliased Text (blend) 3.236 secs ( 12.237
KChars/sec) [100.3%]
Fill Rectangle 3.332 secs ( 15.489
MPixel/sec) [100.0%]
Fill Rectangle (blend) 3.252 secs ( 1.983
Fill Rectangles [10] 4.136 secs ( 15.597
Fill Rectangles [10] (blend) 32.480 secs ( 1.986
Fill Triangles 3.120 secs ( 14.473
Fill Triangles (blend) 3.300 secs ( 1.954
Draw Rectangle 3.012 secs ( 6.241
KRects/sec) [100.0%]
Draw Rectangle (blend) 3.032 secs ( 1.022
Draw Lines [10] 3.032 secs ( 23.416
KLines/sec) [100.3%]
Draw Lines [10] (blend) 3.176 secs ( 4.722
KLines/sec) [100.0%]
Fill Spans 3.056 secs ( 14.776
Fill Spans (blend) 3.284 secs ( 1.964
Blit 3.180 secs ( 16.229
Blit 180 3.372 secs ( 13.392
Blit colorkeyed 3.448 secs ( 13.096
Blit destination colorkeyed 3.580 secs ( 7.208
Blit with format conversion 3.208 secs ( 8.043
MPixel/sec) [100.3%]
Blit with colorizing 3.208 secs ( 6.032
Blit from 32bit (blend) 4.080 secs ( 1.581
Blit from 32bit (blend) with colorizing 4.408 secs ( 1.463
MPixel/sec) [100.2%]
Stretch Blit 3.000 secs ( 12.262
Stretch Blit colorkeyed 3.000 secs ( 11.658
--- Vivi Li 2009-10-16 02:25:11
On regression machine, we still use 09r1-rc10 toolchain. Now I use a latest 4.3
toolchain to build it, and got this:
--
root:/> df_dok --noresults
(*) DirectFB/Core: Single Application Core. (2009-10-16 12:28)
(*) Direct/Thread: Started 'VT Switcher' (49) [CRITICAL OTHER/OTHER 0/0]
(*) Direct/Thread: Started 'VT Flusher' (50) [DEFAULT OTHER/OTHER 0/0]
(*) DirectFB/FBDev: Found 'bfin-t350mcqb' (ID 0) with frame buffer at
0x03f003c0, 225k (MMIO 0x00000000, 0k)
(*) Direct/Thread: Started 'Keyboard Input' (51) [INPUT OTHER/OTHER 0/0]
(*) Direct/Thread: Started 'PS/2 Input' (52) [INPUT OTHER/OTHER 0/0]
(*) DirectFB/Input: IMPS/2 Mouse 1.0 (directfb.org)
../../DirectFB-examples-1.2.0/src/df_dok.c <1502>:
(#) DirectFBError [dfb->CreateImageProvider( dfb,
DATADIR"/biglogo.png", &provider )]: No (suitable) implementation
foun!
(!!!) *** WARNING [Application exited without deinitialization of DirectFB!]
*** [../../../DirectFB-1.4.2/src/core/core.c:862 ]
(!!!) *** WARNING [still objects in 'Layer Region Pool'] ***
[../../../DirectFB-1.4.2/lib/fusion/object.c:241 in fusion_object]
(!!!) *** WARNING [still objects in 'Layer Context Pool'] ***
[../../../DirectFB-1.4.2/lib/fusion/object.c:241 in fusion_objec]
(!!!) *** WARNING [still objects in 'Surface Pool'] ***
[../../../DirectFB-1.4.2/lib/fusion/object.c:241 in fusion_object_pool]
--- Mike Frysinger 2009-10-16 02:58:08
that means you didnt enable libpng support in your build. if you dont enable
support for the various images/fonts, then directfb cant load them. that isnt a
bug.
--- Vivi Li 2009-10-16 03:13:28
I see. I will try again.
What kind of toolchain do you use, 4.3 or 4.1?
--- Mike Frysinger 2009-10-16 03:22:06
just gcc-4.3 afaik
--- Barry Song 2009-10-21 06:05:54
With libpng, I run directfb, I didn't find any problem:
(1)
(*) DirectFB/Core: Single Application Core. (2009-10-21 08:52)
(*) Direct/Thread: Started 'VT Switcher' (414) [CRITICAL OTHER/OTHER 0/0]
(*) Direct/Thread: Started 'VT Flusher' (415) [DEFAULT OTHER/OTHER 0/0]
(*) DirectFB/FBDev: Found 'bf54x-lq043' (ID 0) with frame buffer at 0x03e01000,
(*) Direct/Thread: Started 'PS/2 Input' (416) [INPUT OTHER/OTHER 0/0]
(*) Direct/Thread: Started 'Keyboard Input' (417) [INPUT OTHER/OTHER 0/0]
(*) Direct/Thread: Started 'Linux Input' (418) [INPUT OTHER/OTHER 0/0]
(*) DirectFB/Input: bf54x-keys 0.1 (directfb.org)
Anti-aliased Text 3.044 secs ( 39.027
Anti-aliased Text (blend) 3.204 secs ( 12.359
Fill Rectangle (blend) 3.248 secs ( 1.986
Fill Rectangles [10] 4.132 secs ( 15.612
Fill Rectangles [10] (blend) 32.448 secs ( 1.988
Fill Triangles 3.116 secs ( 14.492
Fill Triangles (blend) 3.296 secs ( 1.957
Draw Rectangle 3.004 secs ( 6.225
Draw Rectangle (blend) 3.020 secs ( 1.026
Draw Lines [10] 3.028 secs ( 23.117
Fill Spans 3.060 secs ( 14.757
Blit 3.156 secs ( 16.352
Blit colorkeyed 3.456 secs ( 13.066
Blit destination colorkeyed 3.576 secs ( 7.216
Blit with format conversion 3.200 secs ( 8.064
Blit with colorizing 3.204 secs ( 6.040
Blit from 32bit (blend) 4.036 secs ( 1.598
Blit from 32bit (blend) with colorizing 4.356 secs ( 1.480
Stretch Blit 3.000 secs ( 12.283
MPixel/sec) [ 99.6%]
Stretch Blit colorkeyed 3.000 secs ( 11.708
(2)
[1] 410 df_input
root:/> (*) Direct/Thread: Started 'VT Switcher' (414) [CRITICAL OTHER/OTHER
0/0] <12288>...
Vivi, can you double-check?
--- Vivi Li 2009-11-12 22:58:39
directfb test fail for all regression machines on which we run this test,
including bf548-ezkit/bf527-ezkit/bf537-stamp.
You may check bug 5180 for some reference.
--- Barry Song 2009-12-15 02:07:10
Yes. Vivi, You are right.
I have confirmed that using same toolchain, same sources code:
1. compile on my computer, all tests pass;
2. compile on your computer, system always crash and tests fail.
Then I will check what's the difference.
--- Barry Song 2009-12-16 05:31:45
The binaries of directfb library and applications are different on right and
wrong machines.
A main problem is that wrong machines will fail this:
for ac_header in dlfcn.h
do
as_ac_Header=`$as_echo "ac_cv_header_$ac_header" | $as_tr_sh`
if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" =
set"; }; then
{ $as_echo "$as_me:$LINENO: checking for $ac_header" >&5
$as_echo_n "checking for $ac_header... " >&6; }
$as_echo_n "(cached) " >&6
fi
ac_res=`eval 'as_val=${'$as_ac_Header'}
$as_echo "$as_val"'`
{ $as_echo "$as_me:$LINENO: result: $ac_res"
>&5
$as_echo "$ac_res" >&6; }
Then "#define HAVE_DLFCN_H 1" and "#include
<dlfcn.h>" will not be given to programs.
Cleary dlfcn.h exists in the toolchaine:
/opt/uClinux/bfin-linux-uclibc/bin/../bfin-linux-uclibc/runtime/usr/include
both on right and wrong machines.
But wrong machine thinks it doesn't exist when running configure.
--- Mike Frysinger 2009-12-16 06:08:09
attach the config.log file. if you're testing trunk, you might want to try
doing this first:
echo > vendors/config/bfin-unknown-uclinux
make lib/DirectFB_{clean,only}
--- Robin Getz 2009-12-16 12:26:39
BTW - is the test machine really building things as root? (the message says is
was built "root@eight"?)
-Robin
--- Barry Song 2009-12-17 01:04:50
After test machine finds dlfcn.h, the binary is still different. For example:
10.99.29.61:
uclinux61-bf548-std:/home/test/work/cruise/checkouts/uclinux-dist.barry/lib/DirectFB/build-DirectFB-1.4.2/src/display/.libs
# ls -l
total 920
-rw-r--r-- 1 root root 128660 2009-12-17 14:14 idirectfbdisplaylayer.o
-rw-r--r-- 1 root root 24472 2009-12-17 14:14 idirectfbpalette.o
-rw-r--r-- 1 root root 46696 2009-12-17 14:14 idirectfbscreen.o
-rw-r--r-- 1 root root 51100 2009-12-17 14:14 idirectfbsurface_layer.o
-rw-r--r-- 1 root root 132044 2009-12-17 14:14 idirectfbsurface.o
-rw-r--r-- 1 root root 70408 2009-12-17 14:14 idirectfbsurface_window.o
-rw-r--r-- 1 root root 454280 2009-12-17 14:14 libdirectfb_display.a
lrwxrwxrwx 1 root root 25 2009-12-17 14:14 libdirectfb_display.la ->
../libdirectfb_display.la
10.99.29.131:
linux-cddk:/home/test/uclinux-dist/lib/DirectFB/build-DirectFB-1.4.2/src/display/.libs
-rw-r--r-- 1 root root 128352 2009-12-17 13:23 idirectfbdisplaylayer.o
-rw-r--r-- 1 root root 24276 2009-12-17 13:23 idirectfbpalette.o
-rw-r--r-- 1 root root 46472 2009-12-17 13:23 idirectfbscreen.o
-rw-r--r-- 1 root root 50820 2009-12-17 13:23 idirectfbsurface_layer.o
-rw-r--r-- 1 root root 131680 2009-12-17 13:23 idirectfbsurface.o
-rw-r--r-- 1 root root 70100 2009-12-17 13:23 idirectfbsurface_window.o
-rw-r--r-- 1 root root 452600 2009-12-17 13:23 libdirectfb_display.a
lrwxrwxrwx 1 root root 25 2009-12-17 13:23 libdirectfb_display.la ->
Test machine gets a bigger binary.
--- Barry Song 2009-12-17 01:21:27
Jie,
the toolchain path in two machines are both:
/opt/uClinux-svn3672/bfin-uclinux/bin:/opt/uClinux-svn3672/bfin-elf/bin:/opt/uClinux-svn3672/bfin-linux-uclibc/bin
Binary path:
10.99.29.61(fail):
/home/test/work/cruise/checkouts/uclinux-dist.barry/user/DirectFB-examples/build-DirectFB-examples-1.2.0/src
10.99.29.131(pass):
/home/test/uclinux-dist/user/DirectFB-examples/build-DirectFB-examples-1.2.0/src
--- Jie Zhang 2009-12-21 19:34:02
Should be fixed now.
--- Mike Frysinger 2009-12-21 19:50:49
i dont see any reasoning behind this change ? why would marking functions
hidden make it work on one system when it works fine on so many others ?
also, declaring variables hidden is a bad idea and known to break across
systems (search the binutils archive). many arches will generate copy relocs
with hidden data which means some units will have their own storage copies.
--- Jie Zhang 2009-12-21 21:44:29
Those functions are local to their DSO. If they are not hidden, one library will
reference the symbol in another library which has the same name, which is wrong.
DirectFB loads its DSOs using readdir. So the load order depends on the order in
which readdir returns the file. It seems that test machine arranges file order
differently than others
I don't understand your second paragraph. I think one principle of DSO is only
export what are intended and hide everything else.
--- Vivi Li 2009-12-23 02:21:49
From the test results on regression machines, Jie's patch works.
So close this bug.
Files
Changes
Commits
Dependencies
Duplicates
Associations
Tags
File Name File Type File Size Posted By
No Files Were Found