[Packaging SVN for Archlinux] Not yet working.

Melodie · Post by **Melodie** » Mon Sep 24, 2007 8:27 pm

Melodie wrote:Updating to revision 11605 is done. Let's see next...

What do you think about the idea of coming back to the preceeding state for the music.cpp file before compiling the new revision ?

here's a way to check deadlock
change line 61 of music.cpp
from the single line:

pthread_mutex_init(&musicinfo_mutex, NULL);

to the four lines:

pthread_mutex_attr_t checkme;
pthread_mutexattr_init(&checkme)
pthread_mutexattr_settype(&checkme, PTHREAD_MUTEX_ERRORCHECK);
pthread_mutex_init(&musicinfo_mutex, &checkme);

So if I put

Code: Select all

pthread_mutex_init(&musicinfo_mutex, NULL);

back to replace the four lines, I'll be again in the same situation as yesterday in Ubuntu (or almost).

Post by **ace123** » Mon Sep 24, 2007 9:33 pm

problem is, that neither solves nor explains the actual problem... The ERRORCHECK is a debug mode that you aren't supposed to use in released software...

I'm going to make sure I check the return values of the pthread functions with ERRORCHECK and print an error if any of them don't make sense.

Melodie · Post by **Melodie** » Mon Sep 24, 2007 10:45 pm

Ok, I wait for more. Good night to all.

Post by **ace123** » Tue Sep 25, 2007 3:50 am

But did it work without the ERRORCHECK change (i.e. directly from SVN)?

You might also want to do an "svn revert" before the next time you update as you might get a conflict otherwise.
The syntax is "svn revert filename.cpp"... you need to specify each file to revert for safety reasons.

Post by **ace123** » Tue Sep 25, 2007 5:10 am

Okay, I committed a version with pthread error checking...

It will print an error like
ERROR IN PTHREAD FUNCTION pthread_mutex_unlock(&musicinfo_mutex): Operation not permitted (1)

Anyway I think I see why this is happening... If you skip (or the song changes for other reasons) *while* another song is loading, it will try to unlock the mutex again to ask for another song.

This is where the problem lies, because there is no way to tell the loader thread to drop whatever it's doing and load another song instead.

I see this if I mash Shift-S (skip track) a whole lot without waiting for the next song to load.

For now, we just need to disable any song change if one is already occurring. In the future we will have a functional sound system with streaming support so this won't be a problem.

Melodie · Post by **Melodie** » Tue Sep 25, 2007 7:05 am

Hello,
What's still strange (did you say 'bizarre' ?) is that under Ubuntu not only I got the sound (still same hardware, as well for GC as for sound card...) but mostover the CPU load was not 100% while playing. Well, it was raising 100% while loading yet, but not while playing.

(I'll check if the loudspeakers are high enough when the game starts... incase).

On the other hand, bizarre things came out, such as impossible to get to this one station, because no green square comes out, or just before, the square was there, but the 'd' control didn't respond, impossible to maneuver (decrease speed... I went straight to the goal : bong !

)

...

BTW, where are those stderr.txt etc... files when they might be needed ?

Post by **ace123** » Tue Sep 25, 2007 8:01 am

Try updating in the latest revision (11608). I think I fixed the multithreading issues.

Basically check your stderr or console output for a "ERROR IN PTHREAD FUNCTION". If you see any of those in the latest revision, I would be interested in the exact message (and when you can get it to happen).

Melodie · Post by **Melodie** » Tue Sep 25, 2007 8:22 am

ace123 wrote:Try updating in the latest revision (11608). I think I fixed the multithreading issues.

Basically check your stderr or console output for a "ERROR IN PTHREAD FUNCTION". If you see any of those in the latest revision, I would be interested in the exact message (and when you can get it to happen).

You were fast (it's morning here). I was at the 11607 and I found :

$ make
make all-am
make[1]: entrant dans le rÃ©pertoire Â« /home/joyce/Vegastrike/vegastrike Â»
depbase=`echo src/cmd/music.o | sed 's|[^/]*$|.deps/&|;s|\.o$||'`;\
g++ -DHAVE_CONFIG_H -I. -I./boost/1_33 -DHAVE_SDL=1 -DSDL_WINDOWING=1 -DHAVE_AL=1 -DHAVE_OGG -I/usr/include/python2.5 -DHAVE_PYTHON=1 -I./src -pipe -I/usr/include/SDL -D_GNU_SOURCE=1 -D_REENTRANT -pthread -MT src/cmd/music.o -MD -MP -MF $depbase.Tpo -c -o src/cmd/music.o src/cmd/music.cpp &&\
mv -f $depbase.Tpo $depbase.Po
src/cmd/music.cpp:86:2: erreur: #else aprÃ¨s #else
src/cmd/music.cpp:80:1: erreur: la condition dÃ©bute ici
src/cmd/music.cpp: In function 'void print_check_err(int, const char*)':
src/cmd/music.cpp:56: attention : deprecated conversion from string constant to 'char*'
make[1]: *** [src/cmd/music.o] Erreur 1
make[1]: quittant le rÃ©pertoire Â« /home/joyce/Vegastrike/vegastrike Â»
make: *** [all] Erreur 2
$

Then came... my boy-friend who modified a part of the music.cpp to add some comments that were missing, this way at this part of the file :

// <<<<<<< .mine
#else
pthread_mutex_init(&musicinfo_mutex, NULL);
//=======
//#else // _WIN32

#define ERRORCHECK_MUTEX
#ifdef ERRORCHECK_MUTEX
pthread_mutexattr_t checkme;
pthread_mutexattr_init(&checkme);
pthread_mutexattr_settype(&checkme, PTHREAD_MUTEX_ERRORCHECK);
checkerr(pthread_mutex_init(&musicinfo_mutex, &checkme));
#else // ERRORCHECK_MUTEX
checkerr(pthread_mutex_init(&musicinfo_mutex, NULL));
#endif // !ERRORCHECK_MUTEX

// >>>>>>> .r11607

I'll do a music.cpp-bak, update and invoke a diff, tu be sure about what's... up !

Post by **hellcatv** » Tue Sep 25, 2007 8:44 am

I'd say move your music.cpp into a faraway land and svn update
this should get you a clean music copy
also you could do
svn revert music.cpp

either way--the >>>>>>>>>>>>>>> and================ and <<<<<<<< means that you and ace123 changed the same lines... that's cus we were all here brainstorming about using the error checking code from pthread_mutex

Post by **ace123** » Tue Sep 25, 2007 8:44 am

What you got there was a conflict... because you modified the source code, and I also committed changes to it.

To discard your modifications and also prevent this from happening again (otherwise it will be flagged as modified/conflicted), you will want to do an "svn revert".

The syntax is "svn revert filename.cpp"... you need to specify each file to revert for safety reasons.
For example: "svn revert src/cmd/music.cpp"

Melodie · Post by **Melodie** » Tue Sep 25, 2007 9:15 am

ace123 wrote:What you got there was a conflict... because you modified the source code, and I also committed changes to it.

I'm not sure what you understood : I don't mess up with diverse revisions, I kept a copy of the file (rev 11607) my boy-friend modified just to check if the changes you did in the 11608 revision had the right comments at the right place, as the 11607 ended the make with errors due to theses comments missing.

1) I kept the copy
2) I did the svn up to get the 11608 revision in, noticed you also added changes to the music.h file additionnally to the music.cpp,

3) I invoked the 'make', which ended without error.

4) I just launched Vegastrike in gdb...

...and it's still slow with the cursor missing fluidness (it jerks as a grasshopper). There is no music either, even if I raise the loudspeakers a lot...

I thinkd I should try to find some people in Archlinux to try Vegastrike svn also, with diverse machines !

Melodie · Post by **Melodie** » Tue Sep 25, 2007 9:22 am

hellcatv wrote:I'd say move your music.cpp into a faraway land and svn update
this should get you a clean music copy
also you could do
svn revert music.cpp

either way--the >>>>>>>>>>>>>>> and================ and <<<<<<<< means that you and ace123 changed the same lines... that's cus we were all here brainstorming about using the error checking code from pthread_mutex

It's not about the >>>>>> or the ===== but the '//'. The compiler was failing on the >>>> and === lines that it was trying to interpret.

→ my boy-friend had not really touched the code, he just added the '//' comments.

// <<<<<<< .mine
#else
pthread_mutex_init(&musicinfo_mutex, NULL);
//=======
//#else // _WIN32

#define ERRORCHECK_MUTEX
#ifdef ERRORCHECK_MUTEX
pthread_mutexattr_t checkme;
pthread_mutexattr_init(&checkme);
pthread_mutexattr_settype(&checkme, PTHREAD_MUTEX_ERRORCHECK);
checkerr(pthread_mutex_init(&musicinfo_mutex, &checkme));
#else // ERRORCHECK_MUTEX
checkerr(pthread_mutex_init(&musicinfo_mutex, NULL));
#endif // !ERRORCHECK_MUTEX
// >>>>>>> .r11607

Melodie · Post by **Melodie** » Tue Sep 25, 2007 10:00 am

ace123 wrote:latest revision (11608).(...)

Basically check your stderr or console output for a "ERROR IN PTHREAD FUNCTION". If you see any of those in the latest revision, I would be interested in the exact message (and when you can get it to happen).

Not one "ERROR IN PTHREAD FUNCTION". Then the bt full says :

(gdb) bt full
#0 0xffffe410 in __kernel_vsyscall ()
No symbol table info available.
#1 0x4061da2c in sched_yield () from /lib/libc.so.6
No symbol table info available.
#2 0x4015c095 in ?? () from /usr/lib/libGL.so.1
No symbol table info available.
#3 0x40add6c7 in _nv000093gl () from /usr/lib/libGLcore.so.1
No symbol table info available.
#4 0x09a493a8 in ?? ()
No symbol table info available.
#5 0x09a914e0 in ?? ()
No symbol table info available.
#6 0x00000000 in ?? ()
No symbol table info available.
(gdb)

Hum... I'll redo it once from the beginning to be sure. After, I'll reboot to Ubuntu, in order to compare and see what happens there with the 11608 revision.

Please would you notice (I read again your preceeding answer, on page 9) that in Archlinux there is no trace for any music loading : not one file !

The error ../music/loading.ogg message appears *also* under Ubuntu juste after lauching (./vegastrike or (gdb) run), but in Ubuntu there is the music and the sound while playing and the mouse cursor *does not jerk*.

/And the Ubuntu vegastrike.config file is still with 32 bits, whereas Archlinux vegastrike.config still has the change done for 16 bits, which does not change anything it seems... .

Melodie · Post by **Melodie** » Tue Sep 25, 2007 10:14 am

I just made a diff between the two vegastrike.config files :

$ diff vegastrike.config /mnt/Ubuntu/home/joyce/Vegastrike/data4.x/vegastrike.config
23,25c23,25
< #set Video Voodoo
< #set Resolution 1280x1024
< #set Sound audio_off
---
> #set Video GeForce3
> #set Resolution 1024x768
> #set Sound audio_3d_windows
28c28
< #set Color win_16
---
> #set Color win_32
556c556
< 
---
> 
< 
> 
> <var name="colordepth" value="32"/>
567c567
< #end -->
---
> 
601c601
< 
604,605c604,605
< #end -->
< 1280x1024
---
> 
> 
---
> #end -->
642c642
< 
---
> 
---
> #end -->
698c698
< 
711c711
< #end -->
---
> 
742c742
< 
---
> 
---
> #end -->
750c750
< 
753c753
< #end -->
---
> 
759c759
< 
---
> 
---
> #end -->
767c767
< 
770c770
< #end -->
---
> 
819c819
< 
822c822
< #end -->
---
> 
829c829
< 
831,833d830
< #end -->
< 
< <var name="Positional" value="false"/>
834a832,834
> 
853c853
< 
855,857d854
< #end -->
< 
< <var name="Sound" value="false"/>
858a856,858
> 
867c867
< 
869c869
< #end -->
---
>

Now, I'll put the Archlinux's one away elsewhere, copy the one from Ubuntu, compile again and see what happens.

Melodie · Post by **Melodie** » Tue Sep 25, 2007 10:33 am

Melodie wrote: Now, I'll put the Archlinux's one away elsewhere, copy the one from Ubuntu, compile again and see what happens.

Same as before, except for the 1024x768 resolution wich is configured in Ubuntu : no sound, cursor jerks.

(gdb) bt full
#0 0xffffe410 in __kernel_vsyscall ()
No symbol table info available.
#1 0x4011ec11 in __lll_mutex_unlock_wake () from /lib/libpthread.so.0
No symbol table info available.
#2 0x4011bb18 in _L_mutex_unlock_108 () from /lib/libpthread.so.0
No symbol table info available.
#3 0x4011b808 in __pthread_mutex_unlock_usercnt () from /lib/libpthread.so.0
No symbol table info available.
#4 0x08a04c0d in Music::_LoadLastSongAsync ()
No symbol table info available.
#5 0x08a04e27 in Music::_GotoSong ()
No symbol table info available.
#6 0x08a051f2 in Music::GotoSong ()
No symbol table info available.
#7 0x08a0556c in Music::GotoSong ()
No symbol table info available.
#8 0x08a06914 in Music::SelectTracks ()
No symbol table info available.
#9 0x08a06c2b in Music::_Skip ()
No symbol table info available.
#10 0x08a07029 in Music::Listen ()
No symbol table info available.
#11 0x08a07063 in Music::MuzakCycle ()
No symbol table info available.
#12 0x08990284 in base_main_loop ()
No symbol table info available.
#13 0x08a9b6b3 in winsys_process_events ()
No symbol table info available.
#14 0x08a7af66 in GFXLoop ()
No symbol table info available.
#15 0x0882b333 in GameUniverse::Loop ()
No symbol table info available.
#16 0x08a9f52d in main ()
No symbol table info available.
(gdb)

Now, I upgrade kernel...

Melodie · Post by **Melodie** » Tue Sep 25, 2007 11:43 am

Melodie wrote:Now, I upgrade kernel...

Yyeesss !!!

The home compiled kernel was messing up somewhere. With a generic one, it works identical to Ubuntu ! \o/ !

(gdb) bt full
#0 0x4001c410 in __kernel_vsyscall ()
No symbol table info available.
#1 0x40630e61 in ?? () from /lib/libc.so.6
No symbol table info available.
#2 0x0868482b in micro_sleep ()
No symbol table info available.
#3 0x08990314 in base_main_loop ()
No symbol table info available.
#4 0x08a9b6b3 in winsys_process_events ()
No symbol table info available.
#5 0x08a7af66 in GFXLoop ()
No symbol table info available.
#6 0x0882b333 in GameUniverse::Loop ()
No symbol table info available.
#7 0x08a9f52d in main ()
No symbol table info available.
(gdb)

To be followed...

Post by **hellcatv** » Tue Sep 25, 2007 6:32 pm

what can often happen is that an nvidia driver can be configured with one kernel...then run with the other--if there's a kernel mismatch you need to remove the nvidia driver and reinstall

Post by **ace123** » Tue Sep 25, 2007 6:32 pm

As to the *comments* in the code, that's a bad idea, because when I commit the file without the ERRORCHECK enabled later on, it's going to create another conflict in that same area (SVN doesn't understand comments). So I would recommend that you revert the file at some point if you don't want that to happen again...

I'm talking about your Archlinux system:
When you say "no sound" do you mean the music doesn't work, or the sound effects do not work?
I see from your config file that audio should be on. Is your music is also set to enabled (say, High volume)?

At this point, everything ought to be working on all of your machines... so that's why I am a bit confused.

About your kernel, it makes sense that might change the behavior of pthreads. You were probably either using NPTL (Native POSIX Thread Library, which works in the kernel).

Edit: or, as hellcatv says, it might be because of the NVidia driver being installed incorrectly.

Melodie · Post by **Melodie** » Tue Sep 25, 2007 7:26 pm

Hello,
I answer both :
I had not done the 'svn revert src/cmd/music.cpp' before doing the 'svn up' to 11608, so I did it after, then I did 'svn -r 11607', then 'svn up' again, and then I did again the 'make'.

About the driver nvidia, I obviously had to compile it again, against the headers when I installed the generic kernel. Anyhow it's enabled in the xorg.conf of my system, so I wouldn't have a graphical interface if I hadn't, unless I would switch to generic 'nv' driver, which is not my goal.

$ uname -a
Linux squirrel 2.6.22-ARCH #1 SMP PREEMPT Thu Sep 20 19:43:47 CEST 2007 i686 AMD Athlon(tm) XP 2800+ AuthenticAMD GNU/Linux

$ lspci
(...)
01:00.0 VGA compatible controller: nVidia Corporation NV34 [GeForce FX 5500] (rev a1)

$ glxinfo
name of display: :0.0
display: :0 screen: 0
direct rendering: Yes
server glx vendor string: NVIDIA Corporation
server glx version string: 1.4
(...)

ace123 wrote:When you say "no sound" do you mean the music doesn't work, or the sound effects do not work?
I see from your config file that audio should be on. Can I assume your music is also set to enabled (say, High volume)?

Well, I told you all along : I had no sound at all under Archlinux, and the CPU load was 100% crazy, and had a mouse cursor jerking.

The reason : my fault (and or my boy-friend's fault)

One part of the home made kernel about the sound :

#
# Sound
#
CONFIG_SOUND=y

#
# Advanced Linux Sound Architecture
#
CONFIG_SND=m
CONFIG_SND_TIMER=m
CONFIG_SND_PCM=m
# CONFIG_SND_SEQUENCER is not set
# CONFIG_SND_MIXER_OSS is not set
# CONFIG_SND_PCM_OSS is not set
CONFIG_SND_RTCTIMER=m
# CONFIG_SND_DYNAMIC_MINORS is not set
# CONFIG_SND_SUPPORT_OLD_API is not set
# CONFIG_SND_VERBOSE_PROCFS is not set
# CONFIG_SND_VERBOSE_PRINTK is not set
# CONFIG_SND_DEBUG is not set

The same bloc of text in the kconfig26 of the generic kernel :

#
# Sound
#
CONFIG_SOUND=m

#
# Advanced Linux Sound Architecture
#
CONFIG_SND=m
CONFIG_SND_TIMER=m
CONFIG_SND_PCM=m
CONFIG_SND_HWDEP=m
CONFIG_SND_RAWMIDI=m
CONFIG_SND_SEQUENCER=m
# CONFIG_SND_SEQ_DUMMY is not set
CONFIG_SND_OSSEMUL=y
CONFIG_SND_MIXER_OSS=m
CONFIG_SND_PCM_OSS=m
CONFIG_SND_PCM_OSS_PLUGINS=y
CONFIG_SND_SEQUENCER_OSS=y
CONFIG_SND_RTCTIMER=m
CONFIG_SND_SEQ_RTCTIMER_DEFAULT=y
# CONFIG_SND_DYNAMIC_MINORS is not set
CONFIG_SND_SUPPORT_OLD_API=y
CONFIG_SND_VERBOSE_PROCFS=y
# CONFIG_SND_VERBOSE_PRINTK is not set
# CONFIG_SND_DEBUG is not set

The missing options of the first one must certainly be somewhere in the second one.

Melodie · Post by **Melodie** » Tue Oct 02, 2007 12:33 am

Hi,

Back to the home made kernel, with thoses supplementary options enabled:

CONFIG_SND_OSSEMUL=y
CONFIG_SND_MIXER_OSS=y
CONFIG_SND_PCM_OSS=y
CONFIG_SND_PCM_OSS_PLUGINS=y
CONFIG_SND_SEQUENCER_OSS=y

(Thks to jll_'s of Archlinux-fr's help for saying what, and... to my boy-friends for re-compiling...

)

The thing is : it works, I've go some sound, but quite ugly sound. Sort of saturated... what do you think about it ?

(release version of today).

Post by **loki1950** » Tue Oct 02, 2007 12:53 am

You may have to play with ALSA configuration a bit sample rate and buffer size's and such can you look at the other distro's you tried where sound worked for examples as starting points for tweaking.how do other music sources sound?

Enjoy the Choice

Melodie · Post by **Melodie** » Tue Oct 02, 2007 1:04 am

loki1950 wrote:You may have to play with ALSA configuration a bit sample rate and buffer size's and such can you look at the other distro's you tried where sound worked for examples as starting points for tweaking.how do other music sources sound?

Enjoy the Choice

Yes, might be an idea to compare (which files by the way ?)

Well 'till then a gdb message

(gdb) bt full
#0 0xffffe410 in __kernel_vsyscall ()
No symbol table info available.
#1 0x4062fe61 in ?? () from /lib/libc.so.6
No symbol table info available.
#2 0x08684a53 in micro_sleep ()
No symbol table info available.
#3 0x08990cc8 in base_main_loop ()
No symbol table info available.
#4 0x08a9c067 in winsys_process_events ()
No symbol table info available.
#5 0x08a7b91a in GFXLoop ()
No symbol table info available.
#6 0x0882bcaf in GameUniverse::Loop ()
No symbol table info available.
#7 0x08a9fee1 in main ()
No symbol table info available.
(gdb)

No idea what it says, but otherwise : theses klk people are burglers: they bought all the cargo for 5000 credits lower !!!

Post by **ace123** » Tue Oct 02, 2007 1:48 am

Yeah that GDB message is meaningless hehe.
The micro_sleep () at the top is just saying it's running.

Anyway I would be interested in the contents of the file called ".openalrc" in your home directory.

If that does not exist yet, try creating one with these contents:
(define devices '(alsa native))

If that doesn't work try:
(define devices '(native))
i.e. use OSS emulation instead of Alsa to see if that's the problem.

Or try:
(define devices '(arts esd alsa native))
Which will try the KDE Sound system or the Gnome Sound System respectively, before going to native sound.

They should change the sound system vegastrike uses.

Post by **hellcatv** » Tue Oct 02, 2007 6:56 am

there's also the possibility that the sample rate is too high for your system

you can go ahead and edit the rate listed in vegastrike.config from <var name="frequency" value="22050"/> to other values like 44100 or 11025

Post by **ace123** » Tue Oct 02, 2007 8:43 am

Or you can update your vegastrike.config to revision 11648...
Klauss just committed a more reasonable frequency variable.

Vega Strike Forums

[Packaging SVN for Archlinux] Not yet working.

Found out !