Closed Bug 58339 Opened 24 years ago Closed 22 years ago

Flash plugin (presumably 5.0 r47 - r50 only) hangs mozilla, when it's trying to play audio, while audio device is active. (Reproducible with some [e.g. es1371] sound driver)

Categories

(External Software Affecting Firefox Graveyard :: Flash (Adobe), defect, P2)

Product:
External Software Affecting Firefox Graveyard
Component:
Flash (Adobe)
Platform:
Other
Other
Type:
defect

Tracking

(Not tracked)

Status:
VERIFIED FIXED

People

(Reporter: bzbarsky, Assigned: rubydoo123)

References

(
URL
)

Details

(4 keywords, Whiteboard: GET THE FLASH 6 BETA -- IT'S FIXED THERE)

Attachments

(4 files)

[workaround] mozilla hangs on startup loading flash plugin while /dev/dsp is inuse
2.25 KB, patch
Details | Diff | Splinter Review
test mp3-file
15.92 KB, application/octet-stream
Details
the .so file for flash 4.0r12 which doesn't have this problem
738.57 KB, application/octet-stream
Details
flash player
2.31 KB, application/octet-stream
Details 
This is trunk build 2000102708 on Linux.

I have installed the Macromedia Flash plugin by just copying the .so into the
plugins directory.

When I start mozilla with the 20001026 builds or when I go to a page that
requires a plugin (www.macromedia.com or about:plugins) with the 2000102708
build I get the following messages in the console:

> Registering plugin 0 for: "*","All types",".*"
> Registering plugin 0 for: "application/x-shockwave-flash","Shockwave Flash","swf"
> Registering plugin 1 for: "application/futuresplash","FutureSplash Player","spl"

If the audio device is active (mp3 player is running, wav player is running),
mozilla stops right before registering the shockwave plugins.  It becomes
completely unresponsive.  Menus don't work, shortcut keys don't work, mouse
clicks don't work.  It looks like a browser crash.  As soon as the mp3 player is
paused, the plugin registration finishes and after that there are no problems.

Given that many people nowadays have mp3s playing pretty continuously and many
will install the Flash plugin, this bug could generate a lot of bad vibes.

Even just a timeout on the plugin registration would help a lot...
I just thought I should add that the device is active, but _not_ locked.  I am
using esd, so multiple programs can write to the audio device at the same time
and the audio streams just get superimposed.
 Esd only allows esd-aware applications to multiplex the audio device; they
have to send their audio to esd instead of directly to the audio hardware.
If applications open the audio device directly it doesn't help (indeed,
as esd itself has the underlying audio device open, they aren't likely to
get anywhere). I suspect that the Flash plugin has not been compiled to talk
to esd and wants direct access to the audio device.

Well, it is rather likely that Flash has no esd support.  However using the same
exact Flash plugin with Netscape 4.75, I can visit flash pages while playing
music and they get displayed fine.

Under the same conditions Mozilla locks up until I pause the music.
Confirming. I thought that the plugin is simply not functional for Mozilla since
I have xmms playing MP3s nearly permanently. Mozilla simply locks up when
started or when accessing a page that requires the flash plugin. This can be
solved by temporarily stopping the audio playback, but most people just will
think that Mozilla hangs and kill it (like I did until now).

I think this is a quite critical bug as
a) NS6 users
b) Mozilla + Flash plugin users
with audio device in use will have this "Browser crashes on Flash pages"
experience. Adjusting Priority and Severity accordingly.

If NS6/Linux ships with the Flash plugin, this bug should definitely be
mentioned in the release notes.
Severity: normal → major
Status: UNCONFIRMED → NEW
Ever confirmed: true
Priority: P3 → P2
I just tried this with a recent branch build (Nov 7 linux build) and the build
does not start if the audio device is busy.  It just looks like it locked on
startup.

adding relnoteRTM keyword.
Keywords: relnoteRTM
*** Bug 60217 has been marked as a duplicate of this bug. ***
build id: 2000121606
Debian Linux Woody 2.2.17 with an SBPCI128 enabled using the es1371 module, with
esd for sound in X. Using Helix Gnome and Enlightenment.

for me it doesn't work either way, even if I've got no sound (CD, MP3, wav,
etc...) When I lauch mozilla in an eterm the last line I get is:

Registering plugin 20 for:
"application/x-java-bean;jpi-version=1.3.0_01","Java(tm) Plug-in", ""

Then nothing. Mozilla doesn't start. I've also got the spellchecker, psm, and
talkback plugins installed. When I remove all other plugins but the two files
for flash, same thing happens. Last line is:

Registering plugin 0 for: "*", "All types",".*"

and nothing again; mozilla freezes before lauching completely.

The only way to make mozilla start is to delete libflashplayer.so from my plugin
directory.
This bug is a major annoyance.  To the user it just looks like the browser has
hung at startup.  adding keywords.
Keywords: mozilla0.9, mozilla1.0, nsbeta1
*** Bug 65496 has been marked as a duplicate of this bug. ***
*** Bug 67665 has been marked as a duplicate of this bug. ***
*** Bug 70066 has been marked as a duplicate of this bug. ***
Since no one seems to be working on this bug I thought I post a workaround:

Use a soundserver like artsd (KDE2) or esd (Enlightenment/GNOME). Most if not
all GNOME/KDE2 applications support "their" soundserver, so this should be no
problem. Output plugins for both soundservers are available for XMMS.
Applications that try to access the sound device directly (like Mozilla with the
flash plugin) have to be started with a preceding artsdsp/esddsp:

artsdsp mozilla

Et voila, the linux flash plugin finally becomes usable with Mozilla.
Moving to m0.9.1
Target Milestone: --- → mozilla0.9.1
Unfortunately this doesn't work for me. If I load a page with flash content (not
about:plugins ) I get the following error message:

mcop warning: user defined signal handler found for SIG_PIPE, overriding

Then mozilla doesn't repsond any more.
Sorry, it works.

When running KDE simply use: artsdsp mozilla
When not running KDE: first start artsd, then artsdsp mozilla

By the way, the plugin for xmms is called:
xmmsarts
Keywords: hang
Keywords: flash, relnote
*** Bug 73618 has been marked as a duplicate of this bug. ***
The problem is that the flash plugin tries to open /dev/dsp without the
O_NONBLOCK flag.  It is not so easy to modify the flash binary file to give the
right flag, but an LD_PRELOAD hack seems to work. Take a look at
http://throb.netspace.org/~bperk/flashhack.c for a rough example.

I think this is a flash bug, not a mozilla bug.


Brian, I have no doubt that this is a flash bug.  But Netscape 4.x manages to 
work around it somehow.  When used with 4.x, the same exact plugin binary not 
only does not block startup but renders flash content correctly without sound if 
I go to a flash/sound page with the audio device busy...

How did 4.x do this? Would using signal handlers prevent the crash? Is there an
updated flash plugin for Linux?
*** Bug 78278 has been marked as a duplicate of this bug. ***
*** Bug 78326 has been marked as a duplicate of this bug. ***
Reassign to Serge. If you can't make 0.9.1, please bump up. I think there a tons 
of dups of this one.
Assignee: av → serge
*** Bug 79361 has been marked as a duplicate of this bug. ***
Keywords: mostfreq
*** Bug 80573 has been marked as a duplicate of this bug. ***
*** Bug 80432 has been marked as a duplicate of this bug. ***
*** Bug 80585 has been marked as a duplicate of this bug. ***
sorry for delay,i removed flash pluings and that resolved problem. verified that bug 80432 is a duplicate of this bug
Reassigning to dr.
Assignee: serge → dr
*** Bug 80860 has been marked as a duplicate of this bug. ***
Status: NEW → ASSIGNED
*** Bug 81285 has been marked as a duplicate of this bug. ***
TM to 0.9.2 per PDT triage (it's OK to check it in by Friday or after 0.9.1
branch is made).
Target Milestone: mozilla0.9.1 → mozilla0.9.2
Note: this sometimes happens on Windows for me with Java. For example, the AIM 
'ding' went off the I started Java and crashed, hard. If this happens on 
Windows, it should be made 0.9.1. Can anyone reproduce on Windows?
Priority: P2 → P1
*** Bug 82933 has been marked as a duplicate of this bug. ***
*** Bug 83680 has been marked as a duplicate of this bug. ***
worksforme.

Setup:
- Mozilla gcc295 nightly 2001-05-27
- Debian unstable
- XMMS using OSS plugin (! - not ESD)
- Flash from Netscape Comm. 4.77 package copies to Mozilla's plugin dir

Reproduction:
1. Play in XMMS
2. Start Mozilla
3. Visit flash site, visit <about:plugins>.

Actual Result:
- All fine.
Broken still for me.

Setup:
- Mozilla gcc295 nightly 2001-05-31
- Mandrake 7.2
- XMMS using OSS plugin (! - not ESD)
- Flash from Macromedia download

Reproduction:
1. Play in XMMS
2. Start Mozilla
3. Visit flash site, visit <about:plugins>.

Actual Result:
- Startup stalls until XMMS is stopped, about:plugins stalls until XMMS is stopped.


From the about:plugins page

Shockwave Flash

File name: /home/thaynes/mozilla/plugins/libflashplayer.so
Shockwave Flash 5.0 r47

So are we to assume this is limited to this particular flash plugin (5.0 r47) or
is it more widespread?


Toby: worksforme with flash plugin v5.0 r47 and ESD for output. I believe it has
to do with how your audio device(s) is/are configured... I don't know anything
about this, though. *helpwanted*
Keywords: helpwanted
This could depend on the sound card.  Some sound cards allow multiple sound
streams to be written to the audio device at once and deal.  Some do not. 
Nevertheless, this is _never_ a problem with NS 4.x and the _same_ Flash plugin.
*grumble* I can't seem to reproduce this no matter how much I seem to mess with
my sound settings... Peter: I have no clue what to do with this. Obviously
people are seeing this left and right, but I'm stumped!
Assignee: dr → peterlubczynski
Status: ASSIGNED → NEW
I see this sometimes on Windows with Java but I have no idea why.
Maybe Serge knows about a quirk in 4.x UNIX plugins? Passing the buck, feel free
to reasign.
Assignee: peterlubczynski → serge
Ok, at Peter's request, I'm going "works for me too" :)

- Mozilla cvs build from 2001-06-07 using gcc 2.95.3
- RedHat 6.2 + updates
- XMMS using OSS
- Flash from netscape 4.77 rpm (Shockwave Flash 4.0 r12)
- Kernel 2.4.4-ac5
- ymfpci sound card driver

Same scenario as everyone else.  Playing mp3s, go to flash-enabled site (
http://www.joecartoon.com ), mouse over a link and it plays the sound .


 

Here is my experience:

Setup:
Mozilla 2001060708 trunk
RedHat 7.0
XMMS with OSS output driver
Flash version v5.0 r47 downloaded from Macromedia

While XMMS is playing, Mozilla will not start, about:plugins hangs, and
http://www.joecartoon.com hangs.  As soon as XMMS stops, everythings works fine.

If I use the version of Flash (4.0 r12) from Netscape 4.77, I do not have the
same problems.  There is no sound on http://www.joecartoon.com, but it does
display. 

If I use Flash 5.0 r47 with Netscape 4.77, I can start Netscape and view
about:plugins while XMMS is playing, but http://www.joecartoon.com will not
display until XMMS is stopped.

So something changed in Flash between 4.0 r12 and 5.0 r47, but Netscape still
handles the change more gracefully than Mozilla does.
This isn't just a problem of stalling the app for a little while, sometimes
mozilla just completely hangs.  See my comments to bug 68773 (which has been
marked as a dupe of this one) for test cases and reproducibility.  Flash 4
plug-in seems to be okay, but in our case Flash 4 is not an option.  On the same
machine where mozilla has "issues" with the Flash 5.0r47 plugin, Nestcape 4.76
does not (i.e. no stalling, multiple flash movies can play audio simultaneously,
etc)...  It would be great to figure out why 4.x can handle this gracefully.  

I'm not an in-depth Linux hacker, but would be happy to try to provide data from
our set-up if someone can tell me what data and how to go about collecting it.

Does anybody have a good contact at Macromedia who might be able to help look
into what's going on on the plug-in side?
I also hang on loading www.joecartoon.com while the audio device is being used.

Setup:
Kernel 2.4.4 with ALSA drivers for sbawe64
Mozilla from today's tip.
Flash 4 (I thought I had v5 installed...)

To reproduce:
Fire up xawtv, and use 'ps ax' to confirm 'esd -nobeeps' is running; then load
the www.joecartoon page. Boom. Turn off xawtv and restrart, and the flash loads
fine.
Get this: I cannot crash a debug build running the same test. The flash works
fine in it, and the sounds play fine together with xawtv audio continuing.
Totally weird.

Chris, this is still a problem. Many people see it, it's mostfreq and has 9 votes.

cc:ing a contact at Macromedia for any input.

Shrirang/Arun, do you know of anyone else?

NOTE: This is a mozilla0.9.2 bug!!!!
kiwi@macromedia.com , can you suggest anything?  We're somewhat stumped.  Note
that this is quirky behavior in Flash plugin as well that somehow 4.x
accomodated. Serge, do you know what 4.x did?  Figuring out that kludge would
help us!

I was unable to reproduce this on my redhat7.0 box with Creative EMU10K1 
PCI Audio Driver. I've tried mozilla091&08,  flash 4 & 5, xmms & realaudio, all 
audio streams I was trying to open simultaneously, even from different user's 
account, mixed just fine. But I'm truly believe, some blocking system calls can 
cause this problem. I'm trying to find out any reproducible test cases maybe old 
linux box to debug this, otherwise it's difficult to say anything:(
I just realized that my problems with the Flash and Mozilla was due
esd's (esound) authentication system.

Mozilla (PC Linux 2001060808) hangs in plugin registration if flash plugin
cannot contact to an esd server although the server is running. The esd
server refused the connection due an incorrect authentication key. The server
stores the authentication key in $HOME/.esd_auth file at the startup of the
server.
Ok, I have an idea: might some drivers (like the EMU10K1) automatically mix
sound streams before sending the output to the sound card? Maybe certain cards
can do this in hardware. There seems to be a common thread among the
'worksforme' users: they are using non-OSS drivers, I think. Maybe either the
driver or hardware is handling multiple audio streams transparently.
On my system:
- Linux 2.4.5
- Soundblaster 16 using the OSS 'sb' driver
- Shockwave Flash 5.0 r47
- Mozilla 0.9

If I'm using xmms with the esound plugin, flash works fine, sound and all. If I
use the OSS plugin, flash works without sound. Since I usually use esd to mix
sound streams, I didn't notice this until I tested it.

HTH
Setting to P2.

Anyone have a solid, reproducable test case for this?
Priority: P1 → P2
Sure, I can reproduce this 100%

1. Get a motherboard with AC97 sound on it.
2. Use the corresponding OSS/ALSA sound driver, I use ALSA.
3. Play some MP3 with XMMS or noatun or any other application.
4. Start Mozilla with Flash 5 R47 plugin. It hangs until the audio device is not
busy any more.

If you can't reproduce this it is probably because your audio device supports
hardware mixing. My other system has a SB AWE 64 and Mozilla hangs there too if
the audio device is in use. Many of the Athlon boards have AC97 sound on board,
I bet you won't have any problem reproducing this bug with them.
OK, here's another one, just to confuse things even more (sorry) !!

I am using mozilla 0.9.1, linux 2.4.5, emu10k1, xmms with oss, and flash plugin
v4.0 r12

If I start playing in xmms, then go to joecomics, I can hear both xmms and
joecomics. However, most of the time when I then leave joecomics and load
another page, mozilla crashes (and I can't get crash data because talkback is
broken !). The crash is not always immediate, sometimes it's on the second page
I load, and once I had to go to joecomics twice to get it to crash.



Sorry, I mean joecartoon, not joecomics. And sorry for the spam. Oh dear...

Bumping to 0.9.3.
Target Milestone: mozilla0.9.2 → mozilla0.9.3
*** Bug 87291 has been marked as a duplicate of this bug. ***
*** Bug 87446 has been marked as a duplicate of this bug. ***
*** Bug 88395 has been marked as a duplicate of this bug. ***
I can reproduce the bug on suse 7.1 / gcc 2.95.2 / kernel 2.4.5

The only way i can get it to work fine is by setting xmms to use the eSound
output plugin and starting mozilla with 'esddsp mozilla' (as described in an
erlier comment).

If this helps, the exact same setup at home works fine. I have a SB!Live Value
at home and Ensoniq ES1371 [AudioPCI-97] here at work.
If any of these hangings is due communication problems with esd (esound)
the bug 85772 might bring some light to it.
*** Bug 90025 has been marked as a duplicate of this bug. ***
We are using debian, but are not running esd on our system at all.  We do use
AC97 audio with the alsa drivers.  Clicking refresh on a page containing a Flash
movie with audio (e.g. http://harrypotter.warnerbros.com ) will reliably bring
mozilla to a grinding halt on our system...
Each version of the Flash plugin seems to be different.

Shockwave Flash 5.0 r45 is statically linked with libesd. 
Could you report for other versions? 
 
chauveau@tolkien:~/.moz_plugins$ ldd libflashplayer.so 
        statically linked
chauveau@tolkien:~/.moz_plugins$ strings libflashplayer.so | grep esd
esd_open_sound
esd_get_server_info
esd_free_server_info
esd_play_stream
esd_close
...
esd_audio_flush
/.esd_auth
/tmp/.esd/socket
esd %s
- esd_subscribe_all_info: not yet implemented!
- esd_update_info: not yet implemented!
- esd_unsubscribe_info: not yet implemented!
/usr/etc/esd.conf
%s/.esd.conf







 
cc Troy Evans. Troy can you please help here ?Thnx!
strings libflashplayer.so | grep esd

produces nothing

ldd libflashplayer.so

statically linked


for v 4.0 r12
Shockwave Flash

File name: /home/thaynes/mozilla/plugins/libflashplayer.so
Shockwave Flash 5.0 r47

Mime Type Description Suffixes Enabled
application/x-shockwave-flash Shockwave Flash swf Yes
application/futuresplash FutureSplash Player spl Yes

This one is statically linked as well and also has the esd symbols internally.

That intrigues me - doesn't the licensing for ESD prohibit static linking
against the esd libraries? libesd is LGPL so the Flash player should only be
dynamically linked to the libesd libraries in order to keep it a separate work. 

Where's my flash player source code under the GPL?!


*** Bug 90674 has been marked as a duplicate of this bug. ***
why do plugins get loaded early?
... and seemingly on the main thread?
... and serially?

FIXME?
Keywords: arch
>why do plugins get loaded early?
because of 
http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/modules/oji/src/nsJVMManager.cpp#597
if line #601 is true all plugins are loaded:
-----
#0  PR_LoadLibraryWithFlags (libSpec={type = PR_LibSpec_Pathname, value = {
        pathname = 0x81adf58 "/home/serge/plugins/linux/libflashplayer5.so",
mac_named_fragment = {
          fsspec = 0x81adf58, name = 0x806a220 " \022&@@\023&@"},
mac_indexed_fragment = {
          fsspec = 0x81adf58, index = 134652448}}}, flags=0) at prlink.c:443
#1  0x4161d465 in nsPluginFile::LoadPlugin (this=0xbfffe920,
outLibrary=@0xbfffe91c)
    at nsPluginsDirUnix.cpp:323
#2  0x41612d20 in nsPluginHostImpl::ScanPluginsDirectory (this=0x81ac760,
pluginsDir=@0xbfffea00, 
    compManager=0x806a220, layoutPath=0x81ac958, checkForUnwantedPlugins=0)
    at nsPluginHostImpl.cpp:3849
#3  0x416131c8 in nsPluginHostImpl::LoadPlugins (this=0x81ac760) at
nsPluginHostImpl.cpp:3937
#4  0x416122b1 in nsPluginHostImpl::GetPluginFactory (this=0x81ac760, 
    aMimeType=0x415c62b1 "application/x-java-vm", aPlugin=0xbfffec4c) at
nsPluginHostImpl.cpp:3650
#5  0x415b8630 in nsJVMManager::StartupJVM (this=0x81ab458) at
nsJVMManager.cpp:602
#6  0x415b8bee in nsJVMManager::MaybeStartupLiveConnect (this=0x81ab458) at
nsJVMManager.cpp:783
#7  0x415bec2f in nsJVMManager::StartupLiveConnect (this=0x81ab458,
runtime=0x80e1a00, 
    outStarted=@0xbfffed48) at nsJVMManager.h:128
#8  0x40f0372b in nsJSEnvironment::nsJSEnvironment (this=0x81aae98) at
nsJSEnvironment.cpp:1527
#9  0x40f03115 in nsJSEnvironment::GetScriptingEnvironment () at
nsJSEnvironment.cpp:1446
#10 0x40f03ba4 in NS_CreateScriptContext (aGlobal=0x818a5d8, aContext=0x8185b90)
    at nsJSEnvironment.cpp:1574
#11 0x40efbb24 in nsDOMSOFactory::NewScriptContext (this=0x818a478,
aGlobal=0x818a5d8, 
    aContext=0x8185b90) at nsDOMFactory.cpp:123
#12 0x40e48531 in ?? () from
/usr/builds/6.x/092/mozilla/dist/bin/components/libdocshell.so
#13 0x40e4a73c in ?? () from
/usr/builds/6.x/092/mozilla/dist/bin/components/libdocshell.so
#14 0x401e7dae in nsGetInterface::operator() (this=0xbffff110, aIID=@0x4095c8f0, 
    aInstancePtr=0xbffff060) at nsIInterfaceRequestor.cpp:37
#15 0x4094a81c in ?? () from
/usr/builds/6.x/092/mozilla/dist/bin/components/libnsappshell.so
#16 0x4094df64 in ?? () from
/usr/builds/6.x/092/mozilla/dist/bin/components/libnsappshell.so
#17 0x4093ba7c in ?? () from
/usr/builds/6.x/092/mozilla/dist/bin/components/libnsappshell.so
#18 0x40939e60 in ?? () from
/usr/builds/6.x/092/mozilla/dist/bin/components/libnsappshell.so
#19 0x4093a252 in ?? () from
/usr/builds/6.x/092/mozilla/dist/bin/components/libnsappshell.so
#20 0x805952b in main1 (argc=1, argv=0xbffff514, nativeApp=0x0) at
nsAppRunner.cpp:1132
#21 0x805a44f in main (argc=1, argv=0xbffff514) at nsAppRunner.cpp:1464

Looks like the only way to do lazy plugin loading is to come up with the concept 
similar to the component registry. We need to get an info about all available 
plugins at start up and on Unix we need to actually download all of them to be 
able to query. XPCOM plugins should be fine in this department.
I have a similar problem (maybe it should be filed as a new bug?).  Anytime
flash starts to load, the browser completely exits, with no warning.  It also
takes all browser windows with it that were running from the same process.  This
has gotten progressively worse with the newer builds.  Before flash did work. 
Then it usually worked, but sometimes a certain point in the movie would cause
the crash.  Recently all sites with flash are crashing (even the zdnet flash
ads).  So I don't think this is caused by the sound.

I'm using the newest nightly build (1001071606), complete fresh install.  Tried
running as root and as regular user with same results.  Can reproduce by
visiting the before mentioned site www.harrypotter.com.  Immediately crashes as
soon as flash starts to load.  My system is a P3 933, 448MB RAM, 30gig drive,
running Redhat 7.1 with the 2.4.5 kernel and XFree86 4.1 (everything else is
standard).  Results of running mozilla on a flash site:

[root@main mozilla]# ./mozilla
./run-mozilla.sh ./mozilla-bin
MOZILLA_FIVE_HOME=/usr/local/mozilla
  LD_LIBRARY_PATH=/usr/local/mozilla:/usr/local/mozilla/plugins
     LIBRARY_PATH=/usr/local/mozilla:/usr/local/mozilla/components
       SHLIB_PATH=/usr/local/mozilla
          LIBPATH=/usr/local/mozilla
       ADDON_PATH=/usr/local/mozilla
      MOZ_PROGRAM=./mozilla-bin
      MOZ_TOOLKIT=
        moz_debug=0
     moz_debugger=
Error: Object "drawingArea" does not have windowed ancestor
[root@main mozilla]#
*** Bug 87931 has been marked as a duplicate of this bug. ***
*** Bug 91589 has been marked as a duplicate of this bug. ***
Target Milestone: mozilla0.9.3 → mozilla0.9.4
*** Bug 92305 has been marked as a duplicate of this bug. ***
*** Bug 92428 has been marked as a duplicate of this bug. ***
Just to add another test case, mozilla hangs here too if I start it while xmms
is running. Stopping the play makes mozilla continue. If I start the player
again and try to view the about:plugins page, it hangs again.
Setup:
- Multimedia audio controller: Ensoniq ES1371 [AudioPCI-97] (rev 08) (not OSS,
using a module called es1371 which comes with the standard kernel it seems)
- kernel 2.4.5
- no esd or something like that
- Shockwave Flash 5.0 r47, the same I use with Netscape 4.77, where this
situation doesn't happen- mozilla 0.9.2
*** Bug 93748 has been marked as a duplicate of this bug. ***
Depends on: 91804
*** Bug 88072 has been marked as a duplicate of this bug. ***
Guys, this is a bug in the flash plugin.  You have to register the plugins and
for whatever reason flash is opening the sound device on registration and will
block until it can open it.  We have to register the plugins when we start up so
we can find out what mime types they support.

I'll buy that it's a Flash plugin bug, but why does the exact same thing work
just fine under Netscape 4.x on Linux and not under Mozilla?
Guys, apparently artsd from KDE 2.2b1 (and possibly 2.2 final, which should be
out next Monday) is enough to trigger this problem (preventing Mozilla from
starting if the Flash plugin files are there). No xmms or anything, just artsd,
which is always running in KDE. Doing a "killall artsd" inside KDE allows
Mozilla to start again.

As far as I know, this didn't happen in KDE 2.1.1 (currently still the latest
released version, but that should change next Monday).
While it is true that artsd in kde2.2beta1 will cause mozilla to not start with
the flash plugin, I have been running mozilla successfully with the artsd
wrapper without any problems. It can be invoked like this
artsdsp mozilla
and then all sound is directed through the kde sound server without any
noticeable problems.
why kill arts ?try running the following comandartsdsp -m /opt/mozilla/mozillainstead of /opt/mozilla/mozillaof course don't forget to replace the path if nessessaryalso the -m option is only needed in some cases.for gnome users i think the have to use esddsp.
It certainly appears that arts and esd can also get involved and cause problems
here, but the issue happens within mozilla itself even if neither of these two
audio systems is running.  Just go to any flash movie with audio.  Once the
movie starts playing, click the refresh button.  On our machine, this invariably
locks mozilla up completely.  Based on past comments, I believe this is because
our system is using AC97 audio with alsa drivers.  I haven't yet been able to
get our in-house audio expert's time to try to track down what it is that the
flash plugin is doing that causes the system so much grief, but the problem
*does not occur* on the exact same system using netscape 4.76.  The two apps
must do things differently when refreshing a page or registering/initializing
plug-ins or something.

It would seem that mozilla itself already has the audio device open to be
playing the first instance of the movie, and then gets in its own way when
trying to refresh the page and gets itself confused or stuck.
*** Bug 94527 has been marked as a duplicate of this bug. ***
*** Bug 92142 has been marked as a duplicate of this bug. ***
mozilla doesn't start if in kde-configuration centre full duplex mode for sound 
card is activated. If full duplex is activated, I have to kill artsd, then moz 
starts.
If I run artsd and in KDE-configuration centre full duplex mode is not 
activated, moz starts without problem.
suse 7.1
kernel 2.4.0
kde2.1
flash plugin 5 r47

i have a via on board sound card.

but i didn't try whether it will crash on a flash-site.
To work with artsd start mozilla with the command "artsdsp ./mozilla" and it
will work.

*** Bug 95229 has been marked as a duplicate of this bug. ***
There has just been a post to the opera-linux mailing list along the lines of,
"Great, I have thrown Netscape 6 into the rubbish bin now that Opera 5 works
with Flash AND doesn't crash with it." (heavily paraphrased.)

We really need some experienced people looking into this, else when that guy
comes back to Netscape 6.x to try it again later on, it will still crash on
Flash loading and frustrate people even more.
Based on James Green's comments ( Aug 14, 2001 15:30 ) I am adding this as a
blocker to the Mozilla advocacy tracking bug.
Blocks: advocacybugs
See also the negative comments here: 

http://slashdot.org/article.pl?sid=01/08/20/1446205&mode=thread

about Mozilla and Flash.

*** Bug 96216 has been marked as a duplicate of this bug. ***
*** Bug 96274 has been marked as a duplicate of this bug. ***
moving to 0.9.5
Target Milestone: mozilla0.9.4 → mozilla0.9.5
*** Bug 99405 has been marked as a duplicate of this bug. ***
Hi,

I've created a patch for mozilla to behave (regarding Flash 5.0 plugin) exactly
the way Netscape 4.x does - where ppl say it "works":

The flash 5.0 plugin from Macromedia will, with this patch applied, now _not
block_ the startup of Mozilla anymore. But the Flash 5.0 plugin _will_ freeze
mozilla as soon as a webpage using the flash plugin is loaded and /dev/dsp is
inuse. The same thing happens to Netscape 4.x with Flash 5.0 loaded.

This patch makes life with the Flash Plugin easier - as the startup of mozilla
isn't blocked by playing music anymore and one point of a possible flash-freeze
(the startup - which occures more often than loading a flash site ;) ) is
omitted. -- But this is, however, no solution to the Buggy Flash 5.0r47 plugin.

Note to Macromedia Programmers: Flash 4.x (which doesn't have this sort
breakage) just disables the sound if /dev/dsp is in use! This is a very easy
solution and doesn't require many changes.

It seems to me (this bug is ONE YEAR OLD!) that Macromedia will wait till Flash
6.x to correct (or not?) this bug - so please apply this Workaround in the given
or changed and reviewed form to the Mozilla tree. (patch is in "cvs diff -u" format)

Best wishes,
domi.
Keywords: patch, review
Thanks Dominik, good job.
I'm using almost the same approach to fix bug 74938, 
it already has r=, and it'll be checked in as soon as it gets sr=
About your patch:
I wouldn't call it workaround, it's a solid fix,
only 
+        plugin =  new ns4xPlugin(&callbacks, pLibrary, pfnShutdown, nsnull);
looks suspicion, it'll leak.
If we decide to take this patch, it's likley the branch would benefit from this fix.
Keywords: nsbranch
To my understanding this patch is not needed if the fix for bug 74938 goes in.
It is essentially the same.
*** Bug 100271 has been marked as a duplicate of this bug. ***
*** Bug 100376 has been marked as a duplicate of this bug. ***
If you are running esd on debian systems, it helps to link
/tmp/.esd -> /var/run/esound
since the flashplugin looks at /tmp/.esd/socket for a valid
esd socket
*** Bug 101106 has been marked as a duplicate of this bug. ***
serge checked in bug 74938 on the trunk yesterday... Can someone who was seeing 
the problem test today's builds?

The recent nightly build (I used
ftp://ftp.mozilla.org/pub/mozilla/nightly/latest/mozilla-i686-pc-linux-gnu.tar.gz)
does solve this bug on my system. For my testcases see Bug-Report 101106. Due to
the lack of KDEx.y.z-stuff on my system, I can't comment on the problems people
have with artsd.

Debian/Progeny 1.0
WindowMaker
Mozilla 0.9.4+ - BuildID 2001092206
ES1371-kernel-module on SB PCI64
XMMS using OSS
Flash-Plugin 5.0 r47

Please excuse by bad commenting-style, I'm new to Bugzilla and I don't know what
to do with those buttons and status-options here around. :)
Progeny Debian
Gnome 1.4 + Sawfish
Mozilla 0.9.4+ - BuildID 2001092206
Kernel 2.4.9-ac7 + alsa 0.9.0b7 (sb16 module)
XMMS using OSS
Flash-Plugin 5.0 r47

I confirm, this bug has been solved on my system too. Without any use of esd or
artsd (i don't have any of them installed).
It is true that Mozilla will *start* when audio is playing.  That was true
in the 0.9.4 release as well.  The current issue (see bug 100376) is that
visiting any page containing flash content will hang the browser if audio
is playing (until the audio process is killed).

I still find this is true on build id 2001092206; my stats:
  - x86/Linux 2.2.19
  - gtk 1.2.8
  - glibc 2.1.2
  - using built-in linux sound driver module es1371

> The current issue (see bug 100376) is that visiting any page containing
> flash content will hang the browser if audio is playing (until the audio
> process is killed).
It is impossible with the Flash 5.0 r47 to do archieve simultanous playback of
sound from mozilla side. It does not work in Netscape 4.x and does not work with
Mozilla and _WONT_ ever -- Until Macromedia provides a new + working version.

The Flash 4.0 Plugin "worked" in Netscape, cause it simply disables the
flash-sound playback if already sound was playing. (My preferred solution for
the Flash 5.0 mess)

best,
domi.
*** Bug 101310 has been marked as a duplicate of this bug. ***
what are the chances of getting this fixed for the 0.9.4 branch?
Whiteboard: [ETA ?], [PDT]
Serge,
is this now fixed with your checkin of the patch in bug 74938 to the 0.9.4 branch?


Yes, mozilla with flash plugin won't hang on startup if audio is playing,
so I'm going to resolve this bug as fixed, but we have to reopen 
bug 100376, as a tracker of known problem. Andrei, could you reopen that one, 
please?
Well, I was wrong, this bug is valid one,
there are a lot of comments about mozilla hangs on startup in this bug report, 
and this startup problem has been eliminated by fix for bug 74938, but the main 
problem how it described in Sammary:
"Registering flash plugin while audio device is active fails and completely 
stops browser" still exists:(
Status: NEW → RESOLVED
Closed: 23 years ago
Resolution: --- → FIXED
*** Bug 102331 has been marked as a duplicate of this bug. ***
Serge, so it should be reopened?
I'm using Mozilla 0.94 and had alot of problems with this bug.  
Based on comments listed here regarding artsdsp, I have modified the
/usr/local/bin/mozilla script which I wrote to launch mozilla.

#!/bin/sh
cd /usr/local/mozilla
/opt/kde2/bin/artsdsp /usr/local/mozilla/mozilla

This seems to work around the problem for me.
Regarding the artsdsp solution, it does not work for me.  I have
arts-0.5.4 (I do not have kde2, so I have the standalone version
of arts), and running mozilla under artsdsp's control does not
have any effect (it tries to open /dev/dsp just the same).

Moreover, I can't see how it *could* work: libflashplayer.so is
*statically* linked.  artsdsp works by using LD_PRELOAD to
interpose its own libc wrapper, and presumably catch the open(2)
call.  But since the plugin is statically linked, it contains
within it only the raw system call linkage.  (Michael, do you
have a different version of libflashplayer.so -- maybe a version
which is dynamically linked? ("ldd libflashplayer.so")  The one
I have, the latest it seems, is 5.0r47.)

As far as external solutions go (given a statically linked
plugin), the possibilities seem to be:

  - Modify /dev/dsp itself.  I posted a patch on bug 100376 to
    make the 'es1371' module pretend O_NONBLOCK is always passed.
    Alternatively, if someone could write a software audio
    mixer proxy device for Linux, that would do it.

  - Modify libflashplayer.so to pass O_NONBLOCK.  This requires
    finding the call site in the binary and patching it.

  - Run mozilla under ptrace(2), intercept all open(2) calls, and
    add O_NONBLOCK.  This would probably make debugging mozilla
    all the more difficult.

Of these solutions, I like the binary patching approach best, but
I don't know enough about the plugin registration process to find
the place open is called.  (I should mention that O_NONBLOCK is
not a perfect solution, again see bug 100376, but it is much
better than hanging.)

Of course, the real solution is to get Macromedia to release a
non-broken plugin.  Does Netscape/AOL have any pull whatsoever
with Macromedia?  Clearly all attempts to communicate at the
engineer to engineer level are ineffective.

Michael O'Brien (and anyone else): This is fixed on the trunk and branch. It is
*not* fixed in the 0.9.4 milestone, which was released before this bug was fixed
by the checkins in bug 74938.  The recent duplicate was also from 0.9.4, so this
bug should not be re-opened.

If anyone wants to download a build that has the fix in it, you can grab a
nightly or wait until 0.9.5 comes out.

The link for the nightlies: http://ftp.mozilla.org/pub/mozilla/nightly/latest/

I can verify that the problem originally reported now works on linux trunk build
2001092906.  Starting Mozilla with XMMS playing music works fine.  Viewing flash
pages while XMMS is playing music works fine for me also.

Setting vbranch keyword - someone please verify this on the branch and update to
verified fixed.
Keywords: vbranch
Unfortunately, for me the latest nightly (2001092906) does not solve the
problem.  I *can* now start Mozilla while /dev/dsp is in use, but I *cannot*
view Flash pages while /dev/dsp is in use.

System:
  - x86/Linux 2.2.19
  - glibc 2.1.2
  - using built-in linux sound driver module es1371
  - Flash plugin 5.0r47


I'm reopening this as a tracker of the flash plugin problem, reproducible on 
linux with e.g. es1371 sound driver, and as an attempt to describe this problem 
more precisely I'm changing the summary description to:
"Flash plugin (presumably 5.0 r47 only) hangs mozilla, when it's trying to play 
audio, while audio device is active. (Reproducible with some [e.g. es1371] sound 
driver)"
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Summary: Registering flash plugin while audio device is active fails and completely stops browser → Flash plugin (presumably 5.0 r47 only) hangs mozilla, when it's trying to play audio, while audio device is active. (Reproducible with some [e.g. es1371] sound driver)
Whiteboard: [ETA ?], [PDT]
*** Bug 102638 has been marked as a duplicate of this bug. ***
what are the chances this will make the 094 branch?
Whiteboard: [Need ETA]
After fix mozilla startup problem (bug 74938) this one is became pure flash bug, 
not a mozilla one, and I have know idea what we can do here to guard us from 
poor implemented plugins...
I'm changing TM to "---", and I would like to keep this one open as known 
problem of flash (5.0 r47) plugin.
Target Milestone: mozilla0.9.5 → ---
Attached file test mp3-file — 

    
    

    
  
I've encountered some strange effects. Simply play that short mp3-file, that
I've attached down this page in a loop (important, loop it). Its length is one
second. Then try to start the unpatched Mozilla 0.9.4. It starts, though it
needs some time. Then try to watch flash-files with sound. You will see, that
there is some scheduling going on. When the mp3-file stops, there is a short
brake, because xmms has to "rewind". In this brake, the sound of the flash-file
is played.

Oh, very important, too: Try this if you're using the
es1731-kernel-sound-module. I think this driver might be a part of the problem.

What are your results ?

--
moritz

    
    

    
  
I've forgot to say, that the flash-file should not have permanent sound. Those
flash-files from joescartoon.com (I know, it's dumb) work fine.

    
    

    
  
removing nsbranch keyword
Keywords: nsbeta1, 
          
            nsbranch, 
          
            patch, 
          
            review, 
          
            vbranch

    
    

    
  
*** Bug 103744 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 105406 has been marked as a duplicate of this bug. ***

    
    

    
  
I can reproduce this problem in build 2001102210, Shockwave Flash 5.0 r47 with
http://www.mobizinc.com.  I have noticed it in the last few daily builds (since
18 October, I think).

I run the unstable Debian distribution, kernel version 2.2.12.

    
    

    
  
*** Bug 107862 has been marked as a duplicate of this bug. ***

    
    

    
  
I've ran strace with netscape 4 and with mozilla.

Netscape 4 with its original Flash plugin:
3319  access("/dev/dsp", W_OK)		= 0
3319  open("/dev/dsp", O_WRONLY|O_NONBLOCK) = -1 EBUSY (Device or resource busy)

.. the plugin kept checking, this is how it should work.

But then I realized that it was not a fair test because the libflasher.so I
was using with mozilla was surely newer, I replaced it and I got a sligthly
different dump:

Netscape 4 with new Flash plugin:
3582  open("/dev/dsp", O_WRONLY)        = ? ERESTARTSYS (To be restarted)
3582  --- SIGALRM (Alarm clock) ---
3582  gettimeofday({1004579476, 744873}, NULL) = 0
3582  sigreturn()                  	= ? (mask now [])
3582  open("/dev/dsp", O_WRONLY)	= ? ERESTARTSYS (To be restarted)
3582  --- SIGALRM (Alarm clock) ---

This second time, Netscape 4 *did* hang. So it's not that mozilla can't work
with Flash and Netscape 4 can. The new plugin is to blame.

So I ran mozilla with the old plugin shipped with netscape 4. Guess what? It
worked... it didn't even make me wait. It just did:

3989  access("/dev/dsp", W_OK)          = 0
3989  open("/dev/dsp", O_WRONLY|O_NONBLOCK) = -1 EBUSY (Device or resource busy)

..and flash seemed to give up on sound.



    
    

    
  
*** Bug 108326 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 110844 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 112424 has been marked as a duplicate of this bug. ***

    
    

    
  
Hi,

I have a "Ensoniq ES1371 [AudioPCI-97]" Soundcard, where the sound drivers in
linux don't allow simultanous playback of 2 sound streams.

But I finally managed to play mp3s and stuff while surfing flash sites!: The
Flash r5 Plugin appearently supports ESD the e-sound deamon, which comes with
GNOME :-) And esound mixes the 2 streams by itself :-)

Fortunately RealPlayer, xmms, vlc and xine have esound support too (!) so
switching to ESD was only some clicks :-))) and now everything works like a
charm. yoyoyo.

best,
dominik.


    
    

    
  
Hey, is there a timeout for plugins ? If so maybe you could check, if the 
plugin times out, and /dev/dsp is blocking, pop up an error message or 
something ?


    
    

    
  
At the very least, should not plugins be run in a separate thread,
so a misbehaved plugin can't hang the whole browser?

    
    

    
  
I am experiencing a similar problem under Solaris (SPARC).  I try running
Mozilla (0.9.6) on a remote machine, for which I do not have permission for
/dev/audio, and after being unable to open the device, I get a SEGV.

open("/optdist/mozilla/./plugins/libflashplayer.so", O_RDONLY) = 41
[...]
access("/dev/audio", 2)                         Err#13 EACCES
open("trace.txt", O_WRONLY|O_APPEND|O_CREAT, 0666) Err#13 EACCES
    Incurred fault #6, FLTBOUNDS  %pc = 0xFEB04A08
    siginfo: SIGSEGV SEGV_MAPERR addr=0x0000000C
    Received signal #11, SIGSEGV [caught]
    siginfo: SIGSEGV SEGV_MAPERR addr=0x0000000C


    
    

    
  
Problem do not occurs anymore. I'm not sure why.I've installed Mozilla Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.6) Gecko/20011120.I've installed the plugin Shockwave Flash 5.0 r47 (had to download it).I think I didn't do anything else (same hardware, same kernel, same driver = P3/AC97, 2.4.14, elsa).Now I can run XMMS and play stuff and I can run Mozilla on a web site with flash.No problem anymore.I have to try at home.Anyone have a clue about what happened ?

    
    

    
  
I've got some hangs when I was running Mozilla+Flash and the XMMS playing on
/dev/dsp. But don't known why, I have a /dev/dsp1 device too.
And then I changed the XMMS configuration to start playing on /dev/dsp1. And now
I can run Sound on Flash, and listen XMMS.

** Info about my system:
[assdebuger@josephine mozilla]$ l /dev/ds*
crw-rw----   1 assdebug audio     14,   3 Dec  2  1999 /dev/dsp
crw-rw----   1 assdebug audio     14,  19 Dec  2  1999 /dev/dsp1
[assdebuger@josephine mozilla]$


about:
Mozilla 0.9.6+
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.6+) Gecko/20011206

[assdebuger@josephine mozilla]$ uname -a
Linux josephine.domain.com 2.4.16 #3 Qua Nov 28 01:24:04 BRST 2001 i686 unknown
[assdebuger@josephine mozilla]$

    
    

    
  
I found (r42 works really good with Netscape/Mozilla)

    
    

    
  
and (r41 crashed netscape 6.0 for me a long time ago)

    
    

    
  
Confirmed with mozilla 0.9.7, linux 2.4.8 and mandrake 8.1 and flash 5.0r47
I'm NOT running the artsd daemon and I'm using xmms with the OSS-driver.
The kernel module I'm using is maestro3 (which again uses ac97_codec and more).
Starting mozilla works, also when playing music in xmms. Xmms uses device /dev/dsp
Things go bad when I'm trying to access any page that utilises flash and xmms is
playing: mozilla totally hangs.
When I close xmms, mozilla is responding again, the page loads just fine and
there are no other problems.
Flash 4.0 works just fine, except that there's no sound.
I also tried something new - I don't think the other submitters tried this: try
flash 5.0 with communicator 4.7x - communicator HANGS TOO, untill I close xmms,
then it continues.
So: this is a flash bug, communicator DID NOT handle this situation any better
than mozilla does. 
Other people that reported that netscape handled this properly probably were
using mozilla with the 5.0 plugin and communicator with the 4.0 plugin.


Why not just ask the Macromedia people to have the flash plugin handle this
better: no /dev/dsp - no sound, just continue loading, don't hang the browser.


    
    

    
  
If I have the plugin installed, opening any page with macromedia content
instantly crashes mozilla (0.9.7 milestone, built from sources). I have no sound
daemon, but the /dev/dsp device is free (I'm not playing any sound). Setup is
more-or-less-stock Mandrake Linux 8.1, kernel version 2.4.17. Sound card (if it
matters, but it would seem otherwise) is a NeoMagic MagicMedia 256ZX. 

    
    

    
  
I would like to echo Francesco Chemolli's comments. I have completely given up
on installing flash plugins into recent versions of Mozilla. Each time I hist a
page with flash in it if I have the plugin loaded, the browser crashes. I am
displaying the browser remotely but I am not playing any sound either locally or
remotely (no sound card in the machine I have Mozilla running from).

    
    

    
  
Flash plugin crashes mozilla on remote display is bug 58937.
Francesco Chemolli, if you are run into different problem than bug 58937,
please open a new bug preferably with talkback's incident ID in it.
Thank you.

    
    

    
  
*** Bug 121378 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 121621 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 118615 has been marked as a duplicate of this bug. ***

    
    

    
  
the item for this bug in the release notes bug really sucks. Could someone
write a nice summary of the current state of the bug plus any known workarounds?

The current item reads: "On Linux, there may be problems with ESD Audio and
Flash." I thought it looked like using esd was a workaround.

    
    

    
  
*** Bug 123116 has been marked as a duplicate of this bug. ***

    
    

    
  
It seems the mozilla team fixed this bug as much as they could.

Unfortunately, it seems not possible to provide a binary patch to the flash
plugin since 
open(...,O_WRONLY) produces a shorter binary than
open(...,O_WRONLY|O_NONBLOCK). 
I've contacted beta_flashlinux@macromedia.com and asked them to release a fixed
version with the correct non blocking open(2) call.

IMHO, comment #135 from Nicolás Lichtmaier (thank you!) provides enough evidence.

    
    

    
  
*** Bug 124496 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 123874 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 91804 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 106204 has been marked as a duplicate of this bug. ***

    
    

    
  
It has been solved by using esound daemon.
Before logging in my host, run `/usr/bin/esd` as a daemon, make an script and
run every boot-time. locate in /etc/init.d/ as in debian distirbution.
in redhat may be /etc/rc.d/rc.local? or not.
If you use mpg123, use esd-supported mpg123.(like mpg123-esd in debian package)

But, if you want to play midi-file using timidity, you should kill running-esd.
Otherwise, the sound leaks out.

    
    

    
  
*** Bug 127148 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 126989 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 127823 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 127859 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 100376 has been marked as a duplicate of this bug. ***

    
    

    
  
We will look at addressing this for the next version of Macromedia Flash Player 
for Linux, we currently don't have a public timeline

Troy

    
    

    
  
*** Bug 128709 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 130513 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 130663 has been marked as a duplicate of this bug. ***

    
    

    
  
Flash 6.0 is out now.. if you wanna test that one for this problem.

http://www.macromedia.com/shockwave/download/index.cgi?P1_Prod_Version=ShockwaveFlash

    
    

    
  
cuz84d@netscape.net: What OS are you on?  Flash 6.0 has been out for Windows for
a while, but the latest _Linux_ release is still 5.0r47 (which this bug is about).

    
    

    
  
I didn't realize that it was out *that* long.  I noticed after I posted here,
that   Troy here is from macromedia.. so I thought too that first he was
refering to v6.0, but I didn't know they didn't release a linux update yet, one
cannot tell just by going to their webpage, where the page will detect your OS.
 I'm running W2k anyway.

    
    

    
  
*** Bug 131628 has been marked as a duplicate of this bug. ***

    
    

    
  
I would like to say something on this, about the reproducibility. Some sound
drivers lie about weather the soundcard is in use or not, so the driver can
allow multiple apps to play sound, and use hardware accelerated mixing. SBLive
among others is ones that can do this. I was wondering, though, does the same
flash plugin have problems on netscape 4.x? Because if it does, this isnt a
Mozilla bug, but instead Macromedia's fault for not checking weather the
soundcard is in use before trying to use it.

    
    

    
  
See comment 135, comment 147 and comment 157.
Whiteboard: [Need ETA] → [need ETA]

    
    

    
  
If it wasn't Mozilla's bug, this bug would have been closed as invalid. There
was a time when, with the same plugin, Moz wouldn't freeze.

    
    

    
  
Thanks Boris. But, hmm. I was thinking, if we started a thread for each plugin
(not just flash) this would make it so that a plugin couldnt lock, crash, or
otherwise foobar mozilla, right?

    
    

    
  
A thread-plugin could still crash mozilla by (for example) corrupting memory or
segfaulting.  The only real protection you would get from using a thread for
each plugin is in allowing mozilla to operate when a thread waits or loops forever.

    
    

    
  
Thats better than nothing, right?

    
    

    
  
Bug also appears when starting the aol instant messenger ( java ) at page
<a href="http://toc.oscar.aol.com" type="image/png"> toc.oscar.aol.com </a> the
applet freezes just as it tries to start audio.

    
    

    
  
I say we should build a threading plugin system. Atleast then, we have partial
protection from misbehaving plugins, which could be quite usefull (and less
annoying when a baddly written sound plugin locks moz)

    
    

    
  
Hey, thats pretty cool. Maybe that should be a priority to flesh that out.
(That, and send someone to hit the various offending plugin authors with
cluebats... you never lock yourself if you cant access sound, thats just very
retarded behavior)

    
    

    
  
*** Bug 134351 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 134353 has been marked as a duplicate of this bug. ***

    
    

    
  
Seems to be plugin related after all ??
Can anybody confirm that this problem can be fixed by installing "Shockwave
Flash 5.0 r48" ???
http://download.macromedia.com/pub/shockwave/flash/english/linux/5.0r48/flash_linux.tar.gz

This seems to have solved part of the problem, I am still having porblems with
java applications that use audio. 
See: http://toc.oscar.aol.com

    
    

    
  
With Flash 5.0 r48 it still does not work .. (tested with artsd as sound daemon)

    
    

    
  
Tis true! The new flash plugin doesnt work. Confirmed using any oss using app.
*sigh!*

    
    

    
  
*** Bug 135120 has been marked as a duplicate of this bug. ***

    
    

    
  
hey, this sucks... what does Macromedia says?
about the threading model:don't do something that could cause even more
problems... we will soon have threads bugs...

    
    

    
  
see comment #168 for macromedia's last word..

    
    

    
  
I'm having this problem with nightly build 2002040606 under linux and any 0.9.9
I've tried if I use flash 5.0 r47 whether the plugin plays sound or not
(www.macromedia.com and www.neopets.com both lock it up) and whether the sound
is in use or not.  I am now using flash 4.0 r12 and it works just fine
everywhere.  I have not been able to find it available for download anywhere, so
I've attached the .so and the .class file for people to try.  

    
    

    
  


  
the .so file for flash 4.0r12

    
    

    
  

      
      
      
      

        Attached file
          
        flash player
        — 
      

    


  
the .class file for flash 4.0r12

    
    

    
  
Skiingyac: are you using esd (gnome/enlightenment), arts (kde), or any other
sound daemon that locks up the sound card even when it's not in use?  Everyone
else has had problems only when the sound card is free.

As for a threaded plugin system, that sounds like complete overkill.  The
problem is that there's a bug in this particular plugin, not that mozilla needs
a threaded plugin system.  There's absolutely no justification for the
additional complexity.

    
    

    
  
Er, I mean when the sound card is in use.

    
    

    
  
not to get anyone's hope up but macromedia has a new version of flash for linux
up ... (still 5 though)

    
    

    
  
The new Flash plugin (5.0 r48) behaves in the same way.

    
    

    
  
 Hey, i can't download the *.so and *.class from flash 4 posted here (comments
195 and 196). It downloads the cgi script !!! Is this a new bug ?!

    
    

    
  
On Windows it's only a file name which is misleading, so you can save it and
then rename to what is should be.

    
    

    
  
 Oh, i see... the file is too big to be a CGI script... but what were the
original file names (for the so and class) ? libflashplayer.so and
ShockwaveFlash.class ?

    
    

    
  
*** Bug 136059 has been marked as a duplicate of this bug. ***

    
    

    
  
I _USED_ to have this problem with Flash, Linux and Mozilla. I now have Flash
with sound, MP3s and Real Player all playing at the same time on the same soundcard.

Steps to fix this:
1. Stop all programs that might be using sound.
2. As root, go in to /tmp and rm -rf .esd. Delete the dir and the socket= in it.
    run /etc/rc.d/init.sound/sound
3. Start up XMMS and configure it to use esd. You might have to install
gnome-audio rpm AND xmms-esd rpm. I use KDE and it still works.
4. Start up mozilla and go to flash.com, on my PC it renders fine and I can move
round flash sites while playing mp3s and both sounds being played together.
5. You can also configure Real Player to use ESD.

FYI, I am running mandrake 8.2, es1371, KDE 3, Mozilla 0.9.9+ and Shockwave
Flash 5.0 r48

    
    

    
  
Please stop spamming this bug with "use a multiplexer" workarounds.

    
    

    
  
*** Bug 136261 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 134282 has been marked as a duplicate of this bug. ***

    
    

    
  
Confirming - the new Flash plugin (5.0 r48) behaves in the same manner.
This is a pretty major bug to have been around for nearly 2 years.  Someone
gonna fix it, or what?

    
    

    
  
> Someone gonna fix it, or what?

Or what.  The possible fixes are:

1) Rewrite Mozilla's plugin architecture to run plugins on separate threads
2) Have Macromedia add "|O_NONBLOCK" to their open() call

#1 is hard, will take lots of time, and is not happening any time soon.  It may
   happen eventually.
#2 is outside of Mozilla's control and is a bug in the plugin.



    
    

    
  
Well, I think the threads for plugins is a better idea anyhow. It prevents
stupid plugins (aka flash) to freeze moz, and also allows moz to keep running at
a pretty good speed without having a cpu intensive plugin slow everything up. I
think its a good idea. It might also be a good idea to have (in addition to
threads) a sandbox for plugins, so they cant accidently corrupt mozilla's memory
space, or play havok with various operating systems like win9x. I belive MSIE
already has a system like that in place. (Though, knowing microsoft, its messed up.)

    
    

    
  
In my Amiga days, I would solve this by binary patching Macromedia's plugin. I
lack the knowledge to do it on Linux, but for someone handy with disassembler
and x86 it should be a quick job. :)

    
    

    
  
IMHO this bug should be marked as a duplicate of bug #62460

I don't think I should make this change. Someone else?

All of Mozilla's plugins crash too much, and Mozilla really should do something
similar to Konquorer (such as comment #210 "#1" and/or comment #211) to isolate
problems with plugins from crashing Mozilla entirely. There is no point in
trying to address every bug in every plugin.

Obviously a fix will not make 1.0. However, I would urge the Mozilla developers
to make this bug a top priority once 1.0 gets out the door. 99% of my crashes
come from Mozilla's plugins.

    
    

    
  
First, can someone change the topic line of this bug? This effects all
non-multiplexing drivers, which is basically all drivers except alsa (atleast I
think alsa mixes sound sources now) and sblive's oss driver which uses hardware
to mix. 

Second, All bugs are marked duplicate of this one, Matt (well, assuming I didnt
misread something somewhere)

Third, Mozilla is a multiplatform browser, and probably what konq does doesnt
apply to us. I still stand by the fact we should defenetly thread mozilla for
plugins (maybe for tabs too, so long rendering times dont temporarly lock moz)
and maybe a plugin sandbox (which could be used to isolate java, as well) if it
seems viable (as I said earlier, microsoft chose to set up a sandbox for various
stuff, so its probably a good idea security wise.) I also agree whatever we do,
it should be a high priority. Im also sensing we dont have enough time before
1.0, so if we actually dont, then we should make it the first bug/feature
request/whatever we do.

    
    

    
  
> "Second, ALL bugs are marked as a duplicate of this one"

Huh? I hope not!

I think that bug #58339 is really bug #62460. Thus this bug should be marked as
a duplicate of the other one.

WRT your third argument, this bug is for Moz on Linux. Konquorer handles
NETSCAPE  plugins on linux much better than Mozilla does. From discussions on
the two bugs mentioned above it initially seems that Moz might be able to
eventually implement a similar solution.

I would recommend that all such "plugin doesn't work" bugs be marked as a
duplicate of #62460 so Moz can focus on the underlying problem.

    
    

    
  
Regardless of how it's done, I think something should be done to keep any plugin
from causing problems like this.  Having to stop any audio software every time I
want to load a page using flash is extremely annoying, and should be unnecessary.
Most users will not put up with this sort of behaviour, and will just use a
different browser.

    
    

    
  
putting in the post 1.0 milestone
Target Milestone: --- → mozilla1.0.1

    
    

    
  
> Or what.  The possible fixes are:
> 
> 1) Rewrite Mozilla's plugin architecture to run plugins on separate threads
> 2) Have Macromedia add "|O_NONBLOCK" to their open() call
> 
> #1 is hard, will take lots of time, and is not happening any time soon.  It may
>    happen eventually.
> #2 is outside of Mozilla's control and is a bug in the plugin.

Look, guys, we're all arguing over the wrong issues here.  The *real* goal is to
fix the macromedia plugin bug *soon*, so that mozilla stops freezing when users
visit one of the 2 million flash webpages, and neither of these options will do
that.

I suggest a third option: add a special case in the mozilla plugin code to
prevent flash from freezing mozilla on linux when the sound device is busy.

We know for sure that when the sound card device is busy, flash will hang
mozilla.  So why don't we just add a check before starting the flash plugin to
make sure that the sound device isn't busy?  If the device IS busy, then don't
start flash.  It's as simple as that.

We can even pop open a dialog box saying "the sound card is busy so we can't run
flash on this page, sorry".  That way the user's will know exactly what's going
on, and will be able to fix it.

This bug is serious, it's been one of the top-reported bugs for months now!  All
we need is to give people something to hold on to.  Let's stop bickering over
theoretical and fix the problem at hand.

    
    

    
  
Yeah, but what the hardware is doing is none of our buisness. Infact, what the
plugin is doing, we dont care. Long as it doesnt get in our way, who cares if it
accidently locked itself. The problem is, it is locking, and it is getting in
our way. We have to fix that problem first, because that effects more than just
flash.
And with displaying the flash dialog... that could be nice, but if macromedia
just does their job for once, we wouldnt be having this problem in the first
place! I seriously think macromedia should just fix the bug, (a couple of
possible patches have already been posted) and get on with the show.


    
    

    
  
> a couple of possible patches have already been posted

A couple of possible suggestions from people who don't know the code have been
posted...

The suggestion in comment 218 would make a decent interim solution, actually...

    
    

    
  
Sounds kinda lame to me. But if the users want it, let em have it.


    
    

    
  
I agree a dialog is a big improvement over an inexplicable hang.

Keep in mind though that /dev/dsp being in use doesn't guarantee a hang;
some people have artsd or esd working, and some people have multiplexing
sound cards.  Perhaps a checkbox along the lines of "let flash try anyway"
would be good.


    
    

    
  
Yeah, I agree.

    
    

    
  
Adding dependency on the mentioned bug because resolving that bug would resolve
this one too.

Additional note to comment #222 - if we would do what the Flash plugin should do
(use open call with O_NONBLOCK) then we would detect exactly only the case when
the plugin would hang, so no checkbox is necessary. 

It is necessary to be sure to use this check only on the Flash plugin not other
plugins. And we probably should make the check only optional by a hidden pref so
we can easily switch it off if the bug in the plugin is fixed.


Depends on: 62460

    
    

    
  
It's true that attempting to open with O_NONBLOCK takes care of those
with multiplexing sound drivers.  However if the plugin is using artsd
or esd, my understanding (can't confirm because it doesn't seem to
work for me!) is the plugin doesn't try to open() /dev/dsp at all.
Mozilla can't anticipate how the plugin will try to get access to the
sound device (/dev/dsp vs e.g. /tmp/.esd/socket), therefore I think a
checkbox is useful.

No longer depends on: 62460

    
  
Depends on: 62460

    
    

    
  
*** Bug 138042 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 134258 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 139035 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 139073 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 134297 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 136517 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 138368 has been marked as a duplicate of this bug. ***

    
    

    
  
**** TO GET FLASH WORKING READ THIS ****
You need to use the flash 4.0 :~(.
Get the two flash player attachments.
The bigger one is .so and the smaller one is .class if you get some weird name
like attachment.cgi like i did.
Install them should work great.
****************************************

I am using Flash 4.0 now... Thanks so much for posting that.  Everything works
fine.  I wonder what the difference is  (what i will not be able to view because
of the older version.)  its more important for me to listen to 24/7 neil young
while browsing though. :)

Thanks so much
Jon Taylor.


    
    

    
  
There is a new version of the flash plugin wich, in combination with mozilla
1.0rc1 seems to hang the system less oftem.

    
  
Blocks: 139820

    
    

    
  
*** Bug 139952 has been marked as a duplicate of this bug. ***

    
    

    
  
Changed platfrom and os to other: note that these issues are present on linux 
and Sun
OS: Linux → other
Hardware: PC → Other

    
    

    
  
*** Bug 140721 has been marked as a duplicate of this bug. ***

    
    

    
  
i agree that mozilla should add a check and warning for this problem, it will have 2 major results:

1- people will not hang their mozilla by accident, and will know why it hang (and maybe give the tip to use esd or others), that it wasnt a mozilla problem, but a flash one

2- as people will be aware that the problem is from the flash, macromedia will not ignore this that easily and will probably fix this faster (or else will make then look very bad)

after they fix this, we can check the checkbox for installed builds, remove the code from the tree and add a line in the release notes saying to upgrade to the fixed flash plugin version

now this is critical, newbies to mozilla will not know why this happend and will blame mozilla... this is unneeded bad press for the all mozilla project

    
    

    
  
...but this *is* a Mozilla bug, so it is OK to blame the current plug-in
handling of  Moz, IMHO...
Every crash is a Moz bug. Moz should be able to handle external code so that it
doesn't crash itself... see Netscape (as also someone mentioned here previously
- see above).

So I don't think the Mozilla Project should blame other companies for writing
software which hangs Mozilla until Mozilla cannot handle poorly written external
code (plug-in).

And remember, that not only the Flash plugin crashes Mozilla, but also java_vm.
Should we then expect Sun to write a better plug-in and we should bear the
crashes until that point...? I don't think so ;)
(And these are only the two most popular plugins...)

The esd solution is also not a 100% one.

I believe Moz could be a little more precautious when launching plugin code... ;)


    
    

    
  
Acroread crashes Mozilla too, not just Flash and Java.

I know that writing a protected plugin environment would be a lot of work (just
how much I'm not sure), but Mozilla really does need to target this for
_sometime_ in the future. The sooner the better ... IMHO this is the biggest
outstanding bug in Mozilla for Linux.

Don't forget that Konquorer does implement something similar when it uses
Netscape plugins. Perhaps we can look at their implementation, and even use
parts of it (correctly wrt GPL etc of course).

It's pretty sad IMHO when someone else uses Netscape plugins better than Mozilla...

    
    

    
  
In my opinion it's not the plugin that hangs mozilla or the system, I think it's
just that mozilla has some bad sound code built in.

BTW: Know that the the flash plugin is a java-class, so this might afterall be a
java problem. The plugin consists of two components, namely :
ShockwaveFlash.class and libflashplayer.so.
Has anybody tried debugging these ?

    
    

    
  
> In my opinion it's not the plugin that hangs mozilla or the system, I think it's
> just that mozilla has some bad sound code built in.

Sorry, but your opinion is wrong...  just killing the program once it's hung
shows that control is in libflashplayer.so at that point.

ShockwaveFlash.class is used for LiveConnect, which is not really supported in
Mozilla (in other words, Mozilla completely ignores this file).

    
    

    
  
Hum... a new revision of the plugin is out for Linux (5.0r48), and now I can
access flash sites while using XMMS, ie, it does not try to open/block the sound
device anymore.
The URL to download it is
http://www.macromedia.com/shockwave/download/index.cgi?P1_Prod_Version=ShockwaveFlash

Please confirm :)

    
    

    
  
Hum... a new revision of the plugin is out for Linux (5.0r48), and now I can
access flash sites while using XMMS, ie, it does not try to open/block the sound
device anymore.
The URL to download it is
http://www.macromedia.com/shockwave/download/index.cgi?P1_Prod_Version=ShockwaveFlash

Please confirm :)


    
    

    
  
20020503 Shockwave Flash 5.0 r48

Xan, sadly no luck, still hangs here, probably you don't have esd or the like
running, btw do you use alsa?

    
    

    
  
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.9) Gecko/20020313  Shockwave
Flash 5.0 r48
No luck here, either.  I still have to stop whatever I'm playing on xmms to
continue loading the page and using mozilla.
I'm using alsa drivers.

    
    

    
  
Wow, weird. I'm using OSS drivers and 5.0r47 doesn't work but 5.0r48 does. I'll
investigate why it does work in my box :)

    
    

    
  
*** Bug 142265 has been marked as a duplicate of this bug. ***

    
    

    
  
Hi

I have opened another ticket that was a duplicate of this one.
This is what I have tested.

Configuration:
- kernel 2.2.20 + openwall patches + alsa 0.5.x + i810 audio drivers
- Slack 8.0 based distro (glibc 2.2.3)

I have tested the Flash Plugin 5.0 r48, while having /dev/dsp busy and it hangs
with all this browsers: Mozilla (Mozilla/5.0 (X11; U; Linux i686; en-US;
rv:1.0rc1) Gecko/20020417), Netscape 6.x, Netscape 4.x, Opera 6 (statically linked).

Then I have used the Flash Player from the Netscape 4.77 distribution (version
4.0 r12) and it doesnt hang with any of the tested browsers (Mozilla, Netscape
6, Netscape 4, Opera 6).

So one "solution" now is to use the older plugin (at least for me).




    
    

    
  
What's the word on the flash 6.x plug-in?


    
    

    
  
This is the same behaviou I have been observing of other applications besides
mozilla/netscape, if you start realplayer-one while the audio device is busy,
you get the same problem. Most other software simply says that the audio device
is busy, and there fore cannot be accesed. Gqmpeg just refuses to play the audio
but does start up.
These are just observations of some testing I did, I hope it is usefull.

    
    

    
  
Debian 2.4.18
Mozilla 2002051121
Shockwave Flash 5.0 r47 plugin
ALSA

I am able to run sound apps and load a flash page (www.flash.com) at the same
time without mozilla locking up. I am using a Sony Vaio laptop with it's built
in soundcard. But, my floormate, who is running the same program versions and is
using  identical sound config but on a Turtle Beach soundcard can't load flash
pages while running a sound app (for the record he tried both r47 and r48).

I have no idea why it would make a difference but hopefully it might help those
of you who, unlike me, know what they are doing.

    
    

    
  
*** Bug 145536 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 146810 has been marked as a duplicate of this bug. ***

    
    

    
  
It seems to me that having Mozilla set an alarm() before initializing the plugin
would solve this problem. The stuck open() call would return EINTR, and things
would go on as planned.

Am I right?

    
    

    
  
*** Bug 143462 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 148337 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 148828 has been marked as a duplicate of this bug. ***

    
    

    
  
Hi all,
  Found a very simple way of sorting this problem out well it works for me, the
problem is with the aRts Sound Server, but u do not need to disable it, simply
click where it says Autosupend if idle for and set the timing to 1 second that
should fix it unless u activate flash within 1 second of the sound server.  Not
sure what else uses the sound server but all by sound works fine will using flash.

Hope that works on other systems, if no try disabling what you have running the
arts server all the time.

    
    

    
  
Not all of us are running the arts daemon, but nice try anyway ;)

    
    

    
  
Adding myself to track this intresting bug.  Anyone know the URL for aRts?

    
    

    
  
arts - www.kde.org

    
    

    
  
*** Bug 150075 has been marked as a duplicate of this bug. ***

    
    

    
  
Two questions concerning this bug:

1. I'm using KDE3 on Linux-Mandrake 8.2, and when i encounter this bug, 
   there is seemingly no way to kill Mozilla. Even killing all processes 
   containing "moz" using kill -KILL does not rid me of my Mozilla windows.
   And then when you close KDE, X somehow doesn't terminate correctly and 
   *Linux even freezes* when shutting down the computer. Is there a 
   workaround for this?

2. This type of problem would seem to warrant an option to turn off 
   plugins in the browser, as is now possible in Mail/News. Is there 
   already an enhancement request filed to add this capability?



    
  

        Attachment #78038 -
        Attachment description: flash player → the .so file for flash 4.0r12 which doesn't have this problem

    
    

    
  
[NOT A MOZILLA ISSUE THIS IS A FLASH 5 ISSUE -- NOT MOZILLA][SEE COMMENT #210
for fixes, attached patch for workaround or use Flash 4 in other attachment][SEE
COMMENT #168 from Macromedia]
Whiteboard: [need ETA] → [NOT A MOZILLA ISSUE THIS IS A FLASH 5 ISSUE -- NOT MOZILLA][SEE COMMENT #210 for fixes, attached patch for workaround or use Flash 4 in other attachment][SEE COMMENT #168 from Macromedia]
Target Milestone: mozilla1.0.1 → Future

    
    

    
  
*** Bug 138669 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 153221 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 152543 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 142699 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 118182 has been marked as a duplicate of this bug. ***

    
    

    
  
Cleaning up the depends list.  Bug 91804 was duped onto us (so I'm removing it
from the list) and Bug 62460 was marked a dup of bug 156493 (so I'm changing it
in the list).
Depends on: 156493
No longer depends on: 62460, 91804

    
    

    
  
adding r48 plugin version to summary since it still hasn't fixed the problem.
Summary: Flash plugin (presumably 5.0 r47 only) hangs mozilla, when it's trying to play audio, while audio device is active. (Reproducible with some [e.g. es1371] sound driver) → Flash plugin (presumably 5.0 r47 and r48 only) hangs mozilla, when it's trying to play audio, while audio device is active. (Reproducible with some [e.g. es1371] sound driver)

    
    

    
  
*** Bug 156403 has been marked as a duplicate of this bug. ***

    
  
Severity: major → normal

    
    

    
  
*** Bug 157948 has been marked as a duplicate of this bug. ***

    
  
Whiteboard: [NOT A MOZILLA ISSUE THIS IS A FLASH 5 ISSUE -- NOT MOZILLA][SEE COMMENT #210 for fixes, attached patch for workaround or use Flash 4 in other attachment][SEE COMMENT #168 from Macromedia] → [PL2:Vendor][NOT A MOZILLA ISSUE THIS IS A FLASH 5 ISSUE -- NOT MOZILLA][SEE COMMENT #210 for fixes, attached patch for workaround or use Flash 4 in other attachment][SEE COMMENT #168 from Macromedia]

    
    

    
  
if we all email
flashplayerunix@macromedia.com

with this bug, maybe they will finally pay attention, here is a sample...

using mozilla (1.0) with the flash plugin (5.0-r48) when the sound
device is busy under gnu/linux (2.4.18) freezes mozilla...
please see
http://bugzilla.mozilla.org/show_bug.cgi?id=58339
for mozilla's full log of this bug

    
    

    
  
reassign to me
Assignee: serge → beppe
Status: REOPENED → NEW

    
    

    
  
Here's my letter to Macromedia I've sent a while ago:

There's a critical bug in libflashplayer, for which there is a known fix.
Implementing that fix would probably take a Macromedia programmer 5 minutes
(most of this time being the formal paperwork for getting approvement for patch
check-in). Fixing this bug will make life better for thousands of users of
Mozilla/Netscape browser worldwide, who visit 2 million flash-enabled web pages
that exist today.

The bug is because certain latest versions or flash player library released by
Macromedia forget to open audio device with in a non-blocking mode (O_NONBLOCK),
as it was with older versions. This causes the Flash plugin and the whole
browser to hang, if any other application (like a sound player) uses the audio
device.

The simple solution which should eliminate the problem would be to add
O_NONBLOCK flag to the open() system call which opens the audio device for sound
playback. It is located in libflashplayer, and instead of:
  open("/dev/dsp", O_WRONLY), Flash player library should do this:
  open("/dev/dsp", O_WRONLY|O_NONBLOCK)

The general description of the bug can be found here:
http://bugzilla.mozilla.org/show_bug.cgi?id=58339
Possible fix suggestion
http://bugzilla.mozilla.org/show_bug.cgi?id=58339#c210:

Proof that this has broken with latest versions of libflashplayer:
http://bugzilla.mozilla.org/show_bug.cgi?id=58339#c135

Interesting comment outlining the importance of fixing this bug as soon as possible:
http://bugzilla.mozilla.org/show_bug.cgi?id=58339#c218

This issue really hurts lots of users and hurts the popularity of swf format itself!


    
    

    
  
BTW, I've got no response :-(

    
    

    
  
Heh, you were expecting one? ;)
I'm wondering -- that seems like a very simple fix.  Would it be possible for
someone to make a binary patch?


    
    

    
  
Vendors respecting their customers _do_ respond.

And Macromedia stresses that Flash plugin is almost ubiquitous, see their Flash
Player Adoption statistics:
http://www.macromedia.com/software/player_census/flashplayer/

You'd think they would be interested in this issue...

Anyway, for those of you that are interested in drawing Macromedia's attention
to this issue, I'd recommend visiting those pages:
http://www.macromedia.com/software/flash/contact/wishlist/
http://www.macromedia.com/support/forums/team_macromedia/
http://www.macromedia.com/support/players/
http://www.macromedia.com/support/flash/ts/documents/linux_player_readme.htm
http://webforums.macromedia.com/macromediaplayers/categories.cfm?catid=184
http://www.macromedia.com/shockwave/download/index.cgi?P1_Prod_Version=ShockwaveFlash&P2_Platform=Linux&P3_Browser_Version=Netscape4

This starts to look like an evangelism case.

    
    

    
  
If someone sends me the affected binary, I can try to create a binary patch.

    
    

    
  
I wonder if it is possible for Mozilla to do something like turning off Flash
animations like the gifs, or make the flash stuff stop wasting CPU cycles
without disabling the plugin.  Trashy sites that use 3 or 4 animations at the
same time are just begging people to stay away (hello blockbusters branches).

    
    

    
  
*** Bug 160841 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 161873 has been marked as a duplicate of this bug. ***

    
    

    
  
Same problem as everyone, had to close down xmms before starting to surf the
web. I was using the OSS driver. I moved to the Esound output plugin
(libesdout.so) and had xmms play some tune. Then opened Mozilla (1.1b) and
visited www.joecartoon.com  - hey... it didn't freeze this time! Then I moved my
mouse over the links and miraculem est... I hear the website sounds right
through the xmms tune. This workaround works for me, maybe also for others?

    
    

    
  
Yeah, but guess what? No one wants to use esound or artsd or any of those lame
peices of software. Sound multiplexing belongs in the driver, not in userland
applications. And this still doesnt matter. Flash has a bug which makes it
clueless. It needs to be fixed. If Macromedia doesnt fix it, then no one will
use Flash because, really, it sucks unless you are on a major platform. And
guess what? Linux/fbsd/other-platforms-effected-by-this-bug collectivly _are_ a
major platform now. So either they fix it, or they loose customers.

    
    

    
  
> Yeah, but guess what? No one wants to use esound or artsd or any of those lame
> peices of software. Sound multiplexing belongs in the driver, not in userland

I agree.

> guess what? Linux/fbsd/other-platforms-effected-by-this-bug collectivly _are_ 
> major platform now. So either they fix it, or they loose customers.

Instead we would use...what..? There are _no_ allternative to flash, either we
(the OpenSource society) creates one, that gets accepted as widely as Flash
_is_, or we do work arounds. The esound solution seems to work. And even thou it
is not the best solution, its a solution.

By the way, this "thread" seems to decade into something non constructive.
Please let this be the last oftopic note.


    
    

    
  
A: argue in email
B: There is an open-source flash player
C: I've said it before and I'll say it again: STOP SPAMMING THIS BUG WITH "use a
multiplexer" !!!

    
    

    
  
Macromedia just released 5.0r50 plugin for Linux. It doesn't fix the problem.
updating summary.
Summary: Flash plugin (presumably 5.0 r47 and r48 only) hangs mozilla, when it's trying to play audio, while audio device is active. (Reproducible with some [e.g. es1371] sound driver) → Flash plugin (presumably 5.0 r47 - r50 only) hangs mozilla, when it's trying to play audio, while audio device is active. (Reproducible with some [e.g. es1371] sound driver)

    
    

    
  
Confirming that 5.0r50 shows the same (bad) behaviour for me.

    
    

    
  
*** Bug 158119 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 163368 has been marked as a duplicate of this bug. ***

    
  
Blocks: majorbugs

    
    

    
  
The problem went away for me when using Flash 5.0r50, but came back after
upgrading to Mozilla1.1.  So mozilla1.0 + flash5.0r50 OK; Mozilla1.1 +
flash5.0r50 hangs again

    
    

    
  
No the problem still persists for 5.0.50 with any mozilla, netscape or any other
browser that uses the same plugin.

    
    

    
  
I also see an error message : I wonder if it is related too.

LoadPlugin: failed to initialize shared library
/usr/lib/mozilla-1.1/plugins/libflashplayer.so
[/usr/lib/mozilla-1.1/plugins/libflashplayer.so: undefined symbol:
overrideShellClassRec] 

    
    

    
  
*** Bug 168765 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 169361 has been marked as a duplicate of this bug. ***

    
    

    
  
I just controlled my system and I noticed, that Mozilla will stop hanging up
loading a flashfile if in arts (KDE3) the Fullduplex-feature is disabled.

Seems that this is the best workaround. I disabled it on Gnome also and it worked.

    
    

    
  
Is this really a problem of Macromedia only?
I just crosschecked this fullduplex-resolution with the project mplayer and this
resolves the problem to open /dev/dsp also.
Ok, maybe they have a bug also but it is possible that this is a general problem
of arts, isn't it?


    
    

    
  
Is this really a problem of Macromedia only?
I just crosschecked this fullduplex-resolution with the project mplayer and this
resolves the problem to open /dev/dsp also.
Ok, maybe they have a bug also but it is possible that this is a general problem
of arts, isn't it?


    
    

    
  
No.  It's not arts specific.
A lot of people tracking this bug don't use arts, myself included.
The problem occurs when multiple applications are trying to access the sound
device (I think, correct me if I'm wrong.)
What arts does is act as an intermediary, mixing sound from various signals, and
then sending it to the sound device.  So arts is actually the only application
accessing the sound device.  I guess that's not the case when arts is in full
duplex mode.

    
    

    
  
Just for the record: Flashplayer 5.0.51 was released on 2002-09-11.
It doesn't fix the problem.

Do those people at Macromedia ever listen to the users? 
Why do they provide a link to their technical issues forum at
http://www.macromedia.com/support/forums
if no one listens?

This bug has been unfixed for so long. In my opinion it will cause a more
serious Public Relations problem than the cost of having a programmer sit down
and fix it a thousand times. Lots of people getting mad at Macromedia for this
are still there and numbers are rising.

And developers too. I planned to develop a UI for a web app using Flash; now I'm
definately not doing it. I'll probably  develop it in Jav. Flash? - never.

    
    

    
  
Macromedia will have a public beta available mid-Oct 2002 that resolves this 
issue.  We currently have this issue resolved and is confirmed fixed by our 
beta testers.

More to come soon - and Flash Player 6 too.

Regards
troy evans
flash player product manager

    
    

    
  
oh...an oficial statement from macromedia! Wow ! Spetacular ! Can you tell us
why it is taking so much time to fix a "one line bug" ? This kind of bugs get
corrected in the same day even in the slowest open source project. Also, I bet
if this bug had appeared in windows it wouldn't take a couple of hours to be
fixed. We are talking about months !

Even if you tell us "there is only one person working on flash_linux, it would
still be ridiculous"

    
    

    
  
We will make sure this kind of situation doesn't happen again.  You will see 
this kind of commitment with our upcoming public beta for Flash Player 6 on 
Linux.  I will notify the list once it is available.

Regards
Troy

    
    

    
  
Thanks,Troy for the update! Guys, please stop spamming this bug now that we know 
that we are soon going to get a release with this bug  fixed. thanks all for 
your patience !!!

    
    

    
  
Hi there,

I have a small workaround after trying the following three types of Sound output:

- aRTSSound
- OSS
- eSound

my environment:

- RedHat 7.3 (stable)
- Kernel ... (I dunno but this is nothing to do with this)
- Sound driver Cirrus Logic CS4297A rev A (although this is not the true sound
card, but system probed this to me)
- Mozilla 1.1 with plugins: JRE 1.3.1_04, Shockwave Flash 4.0r12 (copied from
Netscape communicator preinstalled with RH7.3)

Scenario 1:

1. aRTsSound output (server loaded)
2. XMMS running using aRTsSound
3. Mozilla loaded about:plugins (no problem sometimes)
4. Surf any site, with or without flash.....crash.... :(

Scenario 2:

1. aRTsSound disabled
2. XMMS running using OSS
3. Mozilla loaded about:plugins (no problem)
4. Surf any site OK, flash site (flash running w/o sound)

Scenario 3:

1. aRTsSound disabled
2. XMMS running using eSound
3. There will be an error "Make sure you selected the correct sound output",
when loaded XMMS for the first time and pressed play button. Press play again,
XMMS played cool...(some mixer adjustment have to be made, XMMS uses its own
volume control, have to turn on Video volume, Speaker volume for flash to have
sound strong, otherwise flash running with weak sound)
4. Flash is running cool, but there will be a millisecond phase delay for flash
playing

Assumption:

- I think the best approach is to be on scenario #3 for my environment and for
my own workout.

Further action:

- I will try to use aRTsSound with XMMS again with the 4.0r12 as earlier in
scenario 1, I am still loaded with flash 5.0rXX.

Note: I think this is not a Mozilla bug but rather that flash for linux could
not cook with a different sound driver nicely (for OSS and eSound output) and
could not have process prioritization (aRTsSound) properly.

Thanks

Neo Gigs
gigsvoo@yahoo.com

    
    

    
  
*** Bug 172256 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 172603 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 174444 has been marked as a duplicate of this bug. ***

    
    

    
  
Flash Plugin for Browsers 6 beta (Flash 6)
http://freshmeat.net/releases/100769/

according to freshmeat a new beta for flash on linux was just released.

"A fix was made for Flash Player crashing on remote display or hanging the
browser when it's trying to play audio, while the audio device is active.
Support for ECMA scripting and Flash 6 was also added."

Test the flash plugin for fun and profit:
http://www.macromedia.com/software/flashplayer/special/beta/


Does this indeed solve the problem?

    
    

    
  
My mozilla v 1.2b crashes when going to www.flash.com.  I just installed the new
Flash plugin.

Wardini

    
    

    
  
The new beta Flash plugin solves the problem for me.

I'm running RH 8.0 with mozilla 1.2b/20021017

After running the install script as root, I needed to chmod the files for it to
work...I guess the script doesn't make sure this is correct:

chmod a+r libflashplayer.so flashplayer.xpt
chmod a+x libflashplayer.so

The new plugin seems to fix bug 63182 too. :)

    
    

    
  
I reported a deleted bug.
This bug seems resolved.

I tried www.ford.it that it didn't work but with Macromedia Flash Player 6 beta
it works fine.

    
    

    
  
Im not sure its the new plugins fault. But when (if) flash grabs sound, Mozilla
dies. No messages, just a clean crash. If the sound device is busy (playing MP3s
with OSS) flash seems to work, but without sound.

    
    

    
  
For me it crashes everytime visiting e.g. www.macromedia.com, wether the sound
device is in use or not. Tested with trunk build 2002101920 on Debian sid.

Talkback ID is TB12814930Z

    
    

    
  
New flash works for me. Mozilla 1.2alpha on gentoo linux.

    
    

    
  
*** Bug 175785 has been marked as a duplicate of this bug. ***

    
    

    
  
Be careful!  Mozilla has a warning about compiling against GCC 3.2, saying that
binary plugins may be incompatible unless rebuilt.



    
    

    
  
iirc dmose hacked things so that Flash in particular should continue working
(but if you want Java your SOL and should mail Sun and ask them to provide a
working gcc 3.2 java pretty please).

    
    

    
  
Flash 6 works fine on 2002102004, no probs to play CDs and browse flash web
pages anymore.

Keep in mind that there is a bug in the beta so that you have to have the
gtk-devel packs installed to get the plugin working at all.  If not the flash
apps will not move.

The bug is addressed by macromedia and will be corrected before the final release.

    
    

    
  
Just to remind you, playing a cd doesnt tie up /dev/dsp (well, of course unless
you are playing it using a digital cd reading app)

Try playing an mp3 or something at the same time.
And no, before you say something, using esound or artsd is _not_ a solution to
this, especially since esound and artsd suck. (Multiplex mixing shouldnt _ever_
be  done by a userland application, especially since it requires alteration of
applications to use the mixer, instead of using /dev/dsp _like they are suppost
to do)

    
    

    
  
Correction;  MP3 , realaudio and all other apps using /dev/dsp is working fine
with 2002102004. 

    
    

    
  
Correction!!!

www.ford.it is not working...soem improvement but when you go to car (any car,
Focus, Mondeo, etc,) configuration page, page freezes (is it an issue of Flash
or Ford??)

    
    

    
  
Macromedia explicitly acknowledges that this is/was their bug in their release
notes and claims to fix it (about 3/4 down the page):

http://www.macromedia.com/software/flashplayer/special/beta/release_notes/

Some Issues Addressed

    * Linux Players
          o bug #58937 (in Bugzilla): Flash Player crashes on remote display.
          o bug #58339 (in Bugzilla): Flash Player hangs Mozilla, when it's
            trying to play audio, while audio device is active.


    
    

    
  
*** Bug 175985 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 176247 has been marked as a duplicate of this bug. ***

    
    

    
  
The new beta works fine for me (build 2002102322 , SuSE 7.3) in ish.com (bug
#158119)

    
    

    
  
Also for me using Mzilla 1.1 the new beta is working till now fine!

    
    

    
  
*** Bug 177420 has been marked as a duplicate of this bug. ***

    
    

    
  
WFM with Flash 6 beta on Linux with latest nightly build.

    
    

    
  
Works for me: Mozilla 1.1 on Linux, Flash 6.0 r60 (beta)
md5sum libflashplayer.so: cb1b427f5ef6e4fdedf18d82cc9ee9e9

Should we keep spamming this list with "works for me"?

    
    

    
  
Why ask questions you know the answers to?
Whiteboard: [PL2:Vendor][NOT A MOZILLA ISSUE THIS IS A FLASH 5 ISSUE -- NOT MOZILLA][SEE COMMENT #210 for fixes, attached patch for workaround or use Flash 4 in other attachment][SEE COMMENT #168 from Macromedia] → GET THE FLASH 6 BETA -- IT'S FIXED THERE

    
    

    
  
stop the hurt.

    
    

    
  
since this is fixed with the new version of flash and has been verified as such,
marking as fixed
Status: NEW → RESOLVED
Closed: 23 years ago → 22 years ago
Resolution: --- → FIXED

    
    

    
  
http://mindscape.planetaclix.pt - does not work on mozilla 1.2beta on linux with
the flash6beta plugin.

I was told it works fine with mozilla 1.1 on windows. Any comments ?

    
    

    
  
Comment #336 should filed as a separate bug, unless it is related to audio. When
i go the site mentioned, nothing appears in the browser window, but Mozilla does
not hang as it did with the bug filed here.

    
    

    
  
*** Bug 180691 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 181369 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 182105 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 185071 has been marked as a duplicate of this bug. ***

    
    

    
  
really true - unbelivable, but: it's REALLY fixed!
Status: RESOLVED → VERIFIED

    
    

    
  
Im glad this works now. I wont have to shut off xmms if I want to browse
websites that have ads in Mozilla now. (A lot of companies use flasha ads now)

    
    

    
  
*** Bug 185393 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 195082 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 202130 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 215548 has been marked as a duplicate of this bug. ***

    
    

    
  
i have linux redhat 9, flash player version 6, mozilla v1.5a, kernel
2.4.20-20.9, when flash contains any audio it crashes after the first sound is
played, it doesn't matter weather i stop xmms or not, i have been able to play
it once. when i turned off xmms al together but even then i had the sound come
only in jerks so it was completly incoherant

    
    

    
  
o and by the way it doesn't hang it crashes, if i have xmms on at the same time
then that hangs until i kill mozilla but it doesn't work the other way around. 

    
    

    
  
Loki: try to trace system calls that the plugin/mozilla are making using this
command:

strace -p MOZILLA_PID

Where MOZILLA_PID is the PID of your mozilla process that is displaying that
page. Redirect output of strace to a file, and attach this file here.

Also try reporting this problem to Macromedia as it is more likely a bug in
their player.

    
    

    
  
*** Bug 226465 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 237375 has been marked as a duplicate of this bug. ***

    
    

    
  
*** Bug 216013 has been marked as a duplicate of this bug. ***

    
  
No longer depends on: 156493

    
  
No longer blocks: majorbugs

    
  
Component: Plug-ins → Flash (Adobe)
Product: Core → Plugins
QA Contact: shrir → adobe-flash
Target Milestone: Future → 2002
Version: Trunk → 5.x

    
    

    
  
Version and milestone values are being reset to defaults as part of product refactoring.
Target Milestone: 2002 → ---
Version: 5.x → unspecified

    
  
Product: External Software Affecting Firefox → External Software Affecting Firefox Graveyard

          You need to log in
          before you can comment on or make changes to this bug.
        






  

    
  







  

    

      
Attachment

      

      
      
      
      
    

    

      

        

          

            
General

            

              Creator: 



            

            
Created: 

            
Updated: 

            
Size: