Open GoogleCodeExporter opened 9 years ago
Seems that the problem is solved for me too! No disconnection. Great job Régis.
Original comment by o...@netfusion.fr
on 20 Oct 2011 at 7:29
Has anyone tested successfully on Gbread 2.3.4? I am not able to register at
all using my Droid 3. Using my Xoom tablet works just fine.
Original comment by steve...@gmail.com
on 20 Oct 2011 at 11:16
I've installed r1058 today and soon got the problem. CSS run for several hours.
and I moved form one WiFi location to another. When I arrived to the last
location I've noticed that CSS is not registered (there is no its icon in top
bar). I waited for a long time - no luck. There is no attempts to register at
all. Checked with OSMonitor and have noticed, that there are two processes
running:
1. com.csipsimple. Start time: 15:53 (time when new version of CSS was launched)
2. com.sipsimple:sipStack. Start time: 18:30 (time when I arrived to new WiFi
location)
I suppose that both processes have to have the same start time. Right? If so,
why the second one indicates its start time approximately at the time, when I
arrived to the last WiFi location and discovered that it's not registered now?
Does that mean that sipStack was down (crashed) and then restarted?
And the second question. Do you have a procedure to restore inter-process
communication channel between those two processes in case when one is restarted
(sipStack)? If there is no such specific procedure (reuse the old connection
channel) - it could be the reason why CSS is not registered at that time.
Please verify that it's working - intentionally kill/crash sipStack and restart
it back again and see if CSS will register itself. If there is no such
procedure and you're not going to implement it for whatever reason - the only
solution is to kill main CSS process too and then restart them both... Keeping
CSS running while it's not even try to register itself is not a solution.
Original comment by yok...@gmail.com
on 21 Oct 2011 at 2:42
@yok : can you get the logs from the crash?
About restart stuff, actually in one case you're right, but most of the time,
that's not the way it works.
sipStack process is about the sip service only and the root process is about
the UI. I've separated the two process cause it could help with memory
management.
However, it does not changes the way things works : the sip service is
restarted when connection changes. (So it could explain why the start time is
different) and it is stopped when there is no connection available anymore. It
is automatically restarted when networks changes (independantly of the UI
thread) and it is also started when the UI thread explicetly ask for a
start/restart. (when you go in settings and go back and when you start the
dialer activity).
It was already the case in previous version but was not seen by user since
everything was in the same process.
The only case it can behaves the way you fear is when :
- the app is configured to run only for outgoing.
- the sip service crashes while the dialer UI is suspended (you quit it using home instead of back button or you didn't quit it at all).
-> at this point sip service can be not started when you resume the dialer UI. (I'll fix that by checking state on resume).
On all other case the service process ~should~ be independant of the other
process - and that's safer all the more so as, csipsimple can be used in plugin
mode by other apps without having to start csipsimple UI at all.
However, possible that it crashed in your case : reason why I'd like to get
logs cause else I can't guess what really happened.
I've already seen some crashed with my new timer implementation in pjsip. I'll
investigate that in next days, but if you have logs it can be very helpful for
me.
It is also possible that there is something completely different that make the
sipStack process not restarting correctly (something that block it somewhere).
For this reason too, logs can help me to understand what's happening.
But anyway the goal is not to link it to the UI process. It should be
independant.
@steve : weird that it does not work on your gingerbread device, I made most of
my tests on 2.3.x devices. Can you collect and send me some logs so that I can
see why it does not register.
Thx to all for your tests and ideas, it is very valuable :D !
Original comment by r3gis...@gmail.com
on 21 Oct 2011 at 7:05
For the sake of completeness, it works now for me, too (SGS2 w/stock 2.3.3).
And so far, it also re-registered when I changed networks. I configured it for
WiFi mode only, though, because SIP unfortunately doesn't work via 3G with the
telco provider I use.
Regards,
Uwe
Original comment by uwe.doer...@gmail.com
on 21 Oct 2011 at 7:35
Hello,
I would like to thank Regis for the great work. I would say that it works much
more better for me too. I still experience some issue with the registration but
I think it is due to something different. For a couple of times I see the usual
error 403 Forbidden which Regis fixed a long time ago. But this time it just
keeps showing the error and not registering (but still trying). I have to go to
settings and than go back and everything works fine from this point on.
(without changing anything). The other thing is that sometimes the icon
disappears but the sip is still connected because I can make and receive calls.
The problem is that I cannot get logs because I do not know in which cases this
happens. It may happen once a week or everyday. There isn't any pattern.
Original comment by amis...@gmail.com
on 21 Oct 2011 at 7:58
A couple of things:
1) Can someone tell me how to get the logs, as I want to be able to provide
this from my Droid 3.
2) Is there anything that CSS does to register that is different than say
Sipdroid?
The reason for question 2 is that I can register fine with Sipdroid on my Droid
3 but I cannot get CSS to register. I am wondering if I am missing something
that CSS is looking for.
Thanks,
Steve
Original comment by steve...@gmail.com
on 21 Oct 2011 at 4:11
@steve : To get logs : http://code.google.com/p/csipsimple/wiki/HowToCollectLogs
It does nothing really different from sipdroid. The sip stack is slightly
different but it's still SIP. *HOWEVER* I do not recommand to use two sip apps
at the same time on one phone for 3 reasons.
First, a sip application listen on TCP/UDP ports of the phone. If the two
application listen on same port, it will just break one of the two apps. In
CSipSimple, by default, it choose a random free port so should be fine. But if
you changed it to 5060 or are migrating from an older version it may overlap
with sipdroid that force 5060.
The second point, is that many sip providers only allow *ONE* sip client at a
time to be registered on one sip account. So if you are registered somewhere
else, you may break the other registration.
The last point is about audio resources, apps may lock audio resources and
prevent other app to get access to audio layer.
Else, I've just fixed a bug with timers that may explain crashes observed in
latests versions - espacially on network changes.
It will be produced automatically tonight. I'm still trying to debug a problem
with reset of settings. If you observe this behavior I'm interested by logs :).
-symptom is that you see the first setting screen when starting dialer while
you have already configured that once.
Original comment by r3gis...@gmail.com
on 21 Oct 2011 at 4:41
I just replicated the registration issue on my Droid 3 and sent you the logs
via email. If you need anything else just drop a note. And yes I only run one
app at a time...lol.
Original comment by steve...@gmail.com
on 21 Oct 2011 at 5:19
Regis, I've noticed the the bug you have just described, about easy
configuration screen coming up with default values, even after it had been
configured before. In one occasion, I roamed from Wi-Fi to mobile data, and I
noticed that CSS hadn't registered my accounts for many minutes. Then, I
checked easy configuration, and 'I'm allowed to use mobile' was unchecked. I'm
pretty sure I had checked that configuration before. It's the first thing I do
after installing CSS.
Original comment by casou...@gmail.com
on 21 Oct 2011 at 6:33
@casou : yes, I think you experimented the issue. It is probably also the root
of issue 1337 also.
I'm producing a build that at least should avoid some crash cases, maybe it
will help with this problem too. I noticed that after a crash so maybe it's
linked.
Maybe it will also help with Steven's problem, even if in his case sounds the
server does not reply at all.
@steven : are you sure that your mobile data connection allow you to do sip?
Sometimes some mobile carrier allow sip only under some antennas.
In fact, I never see reply from the sip server in logs, just like if it did
never receive the packets sent by the app. This occurs most of the time when
the carrier has a firewall or some equipement to avoid SIP over mobile data.
For example, here, in France 2 out of 3 carrier blocks SIP over mobile data.
Or, it could also be linked to what I fixed in my latest commit, but if so, I
think that it would be more random.
Original comment by r3gis...@gmail.com
on 21 Oct 2011 at 7:47
Both my Xoom and Droid 3 are on Verizon Wireless, so I definitely know this is
not carrier related. Second I can put Sipdroid on the same Droid 3 and it works
just fine. I have a couple more items I want to try on my Droid 3 and will
communicate results back.
Original comment by steve...@gmail.com
on 21 Oct 2011 at 7:52
Regis, r1061, just uploaded, contains the changes you pointed out above?
Original comment by casou...@gmail.com
on 21 Oct 2011 at 8:58
Yes. I launched a manual build with these changes.
I did tests over the afternoon on one device with this change and sounds better
regarding crashes at least. I'll try to do more tests tomorrow in order to make
sure my changes of the week does not break something else. Do not hesitate to
share if you notice something wrong :)
Original comment by r3gis...@gmail.com
on 21 Oct 2011 at 9:24
Already testing it too...
Original comment by casou...@gmail.com
on 21 Oct 2011 at 9:25
OK so my initial test results with 1061 are as follows:
1) I can register successfully on my Droid 3 with 1061....yea!
2) I selected Always Available in the initial config and hit save, but this is
not being saved truly. In going back into the Easy Config screen Available on
WiFi was selected. I had to re-select Always Available and hit save one more
time for it to be truly saved.
I will do more 3g to Wifi and back tests once I am home, but wanted to get this
out to you while it was fresh in my mind on the train.
Original comment by steve...@gmail.com
on 21 Oct 2011 at 9:53
Omg, I've got that odd condition overnight, where the icon vanishes from the
status bar, but all my accounts are shown green, but CSS is acting like a
ghost. I mean, it responds to taps, but when I tap on an account icon to
deactivate it, the bar beneath the icon turns grey, but its name is kept green
(registered). I had to kill CSS, to get rid of that situation. Well, its dock
is beside my bed, where both wifi and mobile data coverage are poor.
Original comment by casou...@gmail.com
on 22 Oct 2011 at 10:28
Wifi to 3G and back to Wifi yield no registration issues on 1061 so far. I have
left CSS up all night without issue. One thing I might be seeing is very high
battery usage on calls. I am going to keep an eye out for that.
Is there a document that details all the options and what they do?
Original comment by steve...@gmail.com
on 23 Oct 2011 at 5:20
I still experience issues but I do not know in which cases. But I think it is
when there is poor reception something crashes the sip service. The problem is
that in this moment when I want to start logging so I can send an actual log of
what is happening I guess the sip service restarts and registers. Is there
anyway I can collect a log so I can send it to you Regis?
Original comment by amis...@gmail.com
on 26 Oct 2011 at 11:53
I have the same problem Amis. What I do is tell Csip to start recording logs
before the problem occurs. Then when it fails to register, logs have built up
of the problem and even though it registers when I open csip, I can still send
the logs that have accumulated.
Registration is a whole lot better now. The problem has only happened once
since 1058.
Cheers.
Original comment by kro...@gmail.com
on 26 Oct 2011 at 7:44
I finally was able to collect logs so I sent them a minute ago.
Original comment by amis...@gmail.com
on 30 Oct 2011 at 3:09
Issue 1363 has been merged into this issue.
Original comment by r3gis...@gmail.com
on 5 Nov 2011 at 10:59
Issue 1362 has been merged into this issue.
Original comment by r3gis...@gmail.com
on 5 Nov 2011 at 11:06
r1090 was shut down by Android overnight. In the morning there was just a
dehidrated 2MB process running. I'm sticking to the last stable version for me,
r1075.
Original comment by casou...@gmail.com
on 8 Nov 2011 at 1:36
Yes, webrtc changes their code, which broke csipsimple builds. I'm doing a
manual build to replace this one. :)
Original comment by r3gis...@gmail.com
on 9 Nov 2011 at 10:12
r1094 seems to be pretty stable, regarding Android's memory management. It is
up for a couple of hours now, without being shut down by Android even once.
Good job, Regis!
Original comment by casou...@gmail.com
on 9 Nov 2011 at 8:44
Issue 304 has been merged into this issue.
Original comment by r3gis...@gmail.com
on 10 Nov 2011 at 10:46
r1094 seems to be very stable on my Motorola Droid3. Any major changes between
1094 and 1096. I noticed that the build size is 1 meg smaller in 1095 and 1096.
Original comment by steve...@gmail.com
on 12 Nov 2011 at 4:42
Mmmh, sounds the webRTC has gain broken nightly builds :/.
I'll fix that right now, thanks for reporting.
Original comment by r3gis...@gmail.com
on 12 Nov 2011 at 4:51
Actually it's half broken, it should work fine but does not contain optimized
armv7a lib :)
Original comment by r3gis...@gmail.com
on 12 Nov 2011 at 4:53
Fixed. Between 1094 and 1096, except webRTC changes (made by google), there is
just minor changes on wizards.
I think that I'll do the stable release on the market tomorrow. I still have to
make sure it does not bring regression on andorid 1.5 and 1.6 devices. But
that's on the good way :).
Original comment by r3gis...@gmail.com
on 12 Nov 2011 at 5:23
Running the latest build r1235 and got the old annoying problem - CSS doesn't
want to send any registration requests at all.
Scenario:
1) Connected to LAN at home - CSS is registered. 2) Went to another WiFi spot -
CSS is registered, I've made couple of test calls - everything is OK. 3)
Returned back to home LAN - CSS doesn't send any registration requests. I
started CSS's log, waited for 1h - nothing. I've turned WiFi off on the phone
and then turned it back on - nothing again. Stopped the log and I don't see any
messages, associated with PJSUA.
Is PJSUA module somehow stuck? And for how long? Waited for the nest 2h and
wrote this post... I guess it's stuck indefinitely. "Account" shows yellow bar
and message "Inactive" with account, that should be registered by now.
Why PJSUA doesn't react on system notifications like this one (from the log):
D/ConnectivityService( 84): handleConnect() networkinfo=NetworkInfo: type:
WIFI[], state: CONNECTED/CONNECTED
My guess - the problem is with PJSUA, that can't survive frequent CONNECTED /
DISCONNECTED notifications. May be it has sent some request to SIP server and
just waits for a reply, which, as it may wrongly expect, will always come back.
And it doesn't... In any case, it should _always_ reset all its status, when it
gets DISCONNECTED / CONNECTED message(s) from the OS. That's the only way to
make it work more or less reliable.
Original comment by yok...@gmail.com
on 23 Jan 2012 at 3:54
I've got the similar problem today too. But now it happened in a remote
hotspot. CSS was registered for a while. But eventually, when I tried to make a
test call - there was no registration anymore. Any attempts to restore
registration (by toggling account status) failed. Since that account always
stayed in "Inactive" state. Tried to exit CSS (using Menu |Disconnect action)
and it did not helped at all. My guess is - the stuck module is still in memory
(despite my requests to exit the program several times). There is only one
solution - to kill the program with a third party tool :(
Original comment by yok...@gmail.com
on 24 Jan 2012 at 4:26
Same here. May be stun related...
Original comment by kro...@gmail.com
on 24 Jan 2012 at 7:13
Today got similar story again. On remote hotspot CSS was registered. I've
started log, made successful test call. But when I returned home - there is no
registration. Tried to "Disconnect" option - it did not work. I had to kill
"com.csipsimple:sipStack" process with third party tool. Only after that CSS
got its registration back. Log shows no line mentioned "pjsip" at all...
How it's supposed to debug the issue, if there is no data in the log???
Talking about Android's log - the guys who developed it have absolutely no idea
what are they doing and why. There is no time stamps, almost no meaningful
messages, just a dump of a garbage, repeating the same meaningless lines again
and again... Need an example? Here is just one line (repeated 104 times per the
log):
"I/global ( 525): Default buffer size used in BufferedReader constructor. It
would be better to be explicit if an 8k-char buffer is required."
Sorry for the frustration about the Android's log, but how to find out why the
process "com.csipsimple:sipStack" is getting stuck and how to make it start
running again without applying a third party tools?
Original comment by yok...@gmail.com
on 25 Jan 2012 at 3:49
If the goal of r1241 change ("Fixes force close when the app is disconnected")
was to make the "Disconnect" work, then I may report, that I don't see any
change at all. Running the latest r1242 I click on "Disconnect" and two
processes still run in memory... And CSS still stuck in its weird state, when
it doesn't want to register to SIP servers anymore. Accounts are still in
"Inactive" state indefinitely. I see it every day. The only solution is to kill
one of the two processes running with a third party tool:
1. com.csipsimple - killing this process doesn't help
2. com.csipsimple:sipStack - killing this process always helps
Why do not kill the 2nd process when I ask CSS to "Disconnect"? It will
certainly reset the program. And the next time I start it - it will register my
accounts without any problems.
Original comment by yok...@gmail.com
on 2 Feb 2012 at 4:22
During the past week I'm running Sipdroid v2.4, instead of CSS. And you know
what? In the same environment, under the very same conditions, it always works!
As soon as I see "Connected" icon in notification bar, I start to see the big
green dot, indicating that Sipdroid is registered and ready to receive / make
calls.
Looking at the development of this important issue, I'm not confident, that
using PJSIP library was a right decision. It is not reliable and it looks like
no one cares. Sipdroid is a way much better from the reliability perspective.
Why not to use in CSS SIP library used by Sipdroind?
I think until this major problem with CSS will get fixed, I better run Sipdroid
as my SIP phone and will use CSS only at occasions, just as a test tool... :(
Original comment by ozo....@gmail.com
on 7 Feb 2012 at 5:13
@ozo :
1st point : I really care about this problem.
2nd point : I develop CSipSimple on my free time night and week ends. It's a
project I did for myself first and released to share with any kind people that
want to play and contribute with it. Maybe it sounds to you something
professional, but it's not. It's made of feedback and help of users. Sipdroid
is developed by the company behind pbxes.org, and is very linked to their
service. (and there is even bug they maintain to keep compatibility with their
server). There is team working all day long on their product.
3rd point : Nightly builds available here
(http://nightlies.csipsimple.com/trunk/) are as their name suggest nightly
builds that reflects the current development status. As the rest of the project
it's opened to everyone so that users can play with it. It sometimes helps to
solve issues because it contains bug fixes, but there is also regressions
sometimes.
4th point : You seems to have "good" technical advice about which sip stack is
the best... Did you think about contributing the opensource project? Any
contribution is welcome, and one that would fix this issue will be very welcome
;).
5th point : I'm not sure your advise is very good about sip stack. I have
considered integrating imsdroid stack and linphone stack but never sipdroid's
one (actually it's not a stack since many things are inside the android part).
You should have a look to imsdroid and linphone which are very good community
driven opensource projects. I don't force anybody to use CSipSimple. It can
becomes better if community raise. I hope so because many sip provider has
forks based on csipsimple so maybe one of them will contribute some fix for
this.
6th point : This issue is absolutely not linked to pjsip but to my own code.
Every stack needs a glue to be ported to a platform. And this problem is linked
to my glue. Other professional apps such as Bria use pjsip too. If I had a full
day to work on csipsimple probably it would be much more better, but I have
only 1 or 2 hours a day dispatched on my free time.
To conclude, of course if you are looking for some "professional" app,
csipsimple is absolutely not the right project. You should ignore it instead of
making me depressed seeing guys spit on my work. If you are looking for a
project community driven and where contribution are welcome and made for all
users -not only the one of one sip provider- by users it's the right place
however ;).
I don't mean to be rude, I just want to made clear the "goal" of the project
and try to make understand that in this kind of project only constructive
feedback is welcome. Those who spit on developers that just wants to contribute
something to users are not welcome. If it was a professional app, I would find
less chocking to see such feedback because there is a whole support team,
validation team and so on that could reply, investigate and try to report to
developer valuable information to improve the software. But here, all time I
loose trying to explain goal of the project and giving feedback on my
development progress is time that I do not spend improving the software. The
only thing you can do is make me think that I'm spending my time for guys that
do not understand the goal of the project. And trust me, it does not really
encourage to spend this time.
Also, don't worry I read comments even if I don't always reply, so all very
valuable feedback raised -thanks @yok and @krolaw that gives very valuable
reproduction scenario and logs- is taken into account. I don't always reply
because of lack of time but I always investigate logs.
Original comment by r3gis...@gmail.com
on 7 Feb 2012 at 8:39
I'm happy to report (while I keep fingers crossed) that the last two days
(since I've updated to r1252) I'm running CSS without the problem. :)
I don't know how important was the change in r1252 (fix for issue 1564,
considering Ethernet as WiFi), but the result so far is positive. CSS gets
registered as soon as phone gets connected to WiFi... Before that CSS may stuck
on event of connecting to a new WiFi and sit there until I kill the process.
And so far it survived several re-connections without that big problem :)
Now we may start talking about other issues with CSS ... ;)
Original comment by yok...@gmail.com
on 11 Feb 2012 at 4:15
The changes that probably impact this issue is r1249 and it becomes usable with
r1258.
In fact, I previously used the method of the stock android sip application to
stay registered. They used some hysteresis method to wait 2 sec in order to be
sure the wifi network is up. However they've changed that in android 4.0. As
consequence in r1249, I commited what I had under testing integrating their new
method -- which is basically what we had in csipsimple before integration the
stock sip app way minus the change to flag the ongoing notification as a
background running app to avoid kills from android.
So glad to hear that it helps. My fear is that it's not backward compatible
with older android versions. That's why I have to do some basic regression test
before to tell it's fixed. But if some have time here to do some tests with
latest binary it will be welcome.
Also, another point is that this method may consume more data than previous one
that did the job only if IP changed (but was also potentially one root cause of
the problem, and more with android 4.0 that does not announce IP the same way).
The r1258 should be there to at least reduce the double start when starting the
service first, but I think that anyway the app will try to restart sip more
often all the more so as 3G network is allowed.
So I'm also interested in changes in data consumption and battery consumption
you get with latest version.
Original comment by r3gis...@gmail.com
on 11 Feb 2012 at 11:06
Yay it works :-)
Original comment by kro...@gmail.com
on 13 Feb 2012 at 7:32
Found this issue after noticing the same issue on my HTC Thunderbolt with
CSIPSimple and a derivative of CM7 (Liquid Smooth 3.2 based on Gingerbread
2.3.7).
Moved to the nightlies after reading this issue thread and installed r1277 and
so far so good :)
Thanks for your work. Now if Just gotta get it worked out on my Triumph.
Original comment by thecooln...@gmail.com
on 26 Feb 2012 at 5:28
Sorry, spoke too soon. I got a notification on my phone and went to pick it up
and check it and sure enough, there was no CSIPSimple in the status bar.
Instead, my pbxes account was sitting as Inactive instead of "Unauthorized".
Original comment by thecooln...@gmail.com
on 26 Feb 2012 at 5:44
Beginning with 1279 and now 1281 I have the returned problem beck. :(
CSS refused to register. Trying to set account off / on shows:
"Error while registering - Service Unavailable"
Turning off/on WiFi doesn't help. Calling "Disconnect" menu - doesn't help
either. The only solution is to kill CSS processes. It has already happened
several times, including 2 times today...
Original comment by yok...@gmail.com
on 29 Feb 2012 at 3:09
Repeatable test case. Move to edge of wifi range, so connectivity is poor, yet
phone does not switch to wifi. Wait until sip registration times out. (I
think accounts will change status to inactive.) Now move closer to the access
point so packets may flow reliably. Csip should eventually retry and register,
but it seems locked on in active and cannot make it reregister without force
closing. Leaving wifi for
Original comment by kro...@gmail.com
on 2 Mar 2012 at 7:52
3g and back again works fine :-) suspect similar behaviour in poor reception
3g/2g areas.
Original comment by kro...@gmail.com
on 2 Mar 2012 at 7:54
Is this working as of now?
Original comment by nblanch...@movitas.com
on 5 Mar 2012 at 1:56
Issue 1272 has been merged into this issue.
Original comment by r3gis...@gmail.com
on 5 Mar 2012 at 4:15
I'm happy to announce my situation.
== environment ==
GingerBread 2.3.4
Kernel 2.6.35.7
CSipSimple r1326
1. when 3G on & Wifi off,
run CSipSimple(CSS) and it works.
2. when Wifi on, (automatically 3G is off)
CSS connection is closed and opened (good).
3. when Wifi off, (automatically 3G is on)
CSS connection is closed and NOT opened (bad).
4-1. when manually 3G off and 3G on,
CSS connection is opened (good)
or
4-2. when run CSS icon again
CSS connection is opened (good)
This case is repeatable...
I hope this can give any clue to this problem..
Thanks.
Original comment by hansg...@gmail.com
on 14 Mar 2012 at 6:57
one more thing...
5. after some time, when 3G goes off and on (unstable 3G)
CSS connection is closed and NOT opened(bad)
From these situations, I think,
some events(3 & 5) are missing ...
so re-registration code cannot be called.
the common part between 3 & 5 is that 3G becomes on automatically...
as the 4-1 shows, manual 3G off & on works fine.
one more test case,
6. when 3G off and Wifi on,
I turned off and on the Wifi AP, so wifi signal goes off and on.
CSS connection is closed and opened (good)
again, I hope this can help...
Thanks.
Original comment by hansg...@gmail.com
on 14 Mar 2012 at 7:44
Original issue reported on code.google.com by
yok...@gmail.com
on 9 Jul 2011 at 8:56