Closed ncp1113 closed 6 years ago
mikebrady
commented
7 years ago Thanks for the report. I'd like to try to replicate it, to see if I can reproduce the crash. Does it happen reliably? If so, would you be kind enough to post details such as machine and OS version?
ncp1113
commented
7 years ago It happens most of the time but not all of the time. Shairport-sync is running on a Raspberry Pi. At first I thought it was just when used with an iPad Mini with iOS 9.3.5 but then it did happen with my MacBook Pro running Sierra version 10.12.3. I would be happy to help with debugging. I am rusty using gdb so I need instructions on how to run the daemon with gdb.
mikebrady
commented
7 years ago Thanks. Let me see if I can get it to happen here. Is the raspberry pi connected over Ethernet or WiFi?
ncp1113
commented
7 years ago It is running on Wifi.
mikebrady
commented
7 years ago Thanks.
mikebrady
commented
7 years ago One last question, for now, please. What kind of Raspberry Pi is it? (Original, Pi 2, Pi 3?)
ncp1113
commented
7 years ago Raspberry Pi 3 (RPi3) Model B Quad-Core 1.2 GHz 1 GB RAM
mikebrady
commented
7 years ago Hi there. Trying without too much success to reproduce this. I notice you have soxr interpolation enabled. If you turn it back to basic, can you get the crash to recur? I realise this is a bit of a pain to ask, but the only crashing I can get – in highly unusual circumstances – is a memory allocation bug associated with the soxr decoding (if it's real, I'm sure it's my fault!).
mikebrady
commented
7 years ago Actually, don't trouble yourself. On reflection, it's probably an artefact of a race condition at termination, something going on between some of the threads. I'll keep digging. I was using valgrind, BTW, on an x86 processor.
ncp1113
commented
7 years ago I used to program in C. Let me know if there is anything I can do.
I will turn off soxr because I am curious. If the crashes stop I’ll let you know.
On Mar 4, 2017, at 3:50 PM, Mike Brady notifications@github.com wrote:
Actually, don't trouble yourself. On reflection, it's probably an artefact of a race condition at termination, between some of the threads. I'll keep digging.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-284179885, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tSo_-I-UVvEggfwnfWxOKJ1BYgwzks5ric6CgaJpZM4MMgFX.
mikebrady
commented
7 years ago Thanks. Looking through the code, it looks like a new player thread can come into existence at the start of a new play session before the thread associated with old one has gone away, and the code is not fully re-entrant. So, the old thread is deallocating resources that the new thread needs, or something like that... It may take a little while to disentangle this stuff. This is one of the joys of open source!
ncp1113
commented
7 years ago I turned off soxr and so far it hasn’t crashed. We tried some turning off one device and turning on another and so far so good. I’ll let you know if it crashes again.
Thanks for all of your attention to this. Often with open source developers get bored and move on.
Nancy
On Mar 5, 2017, at 4:39 AM, Mike Brady notifications@github.com wrote:
Thanks. Looking through the code, it looks like a new player thread can come into existence at the start of a new play session before the thread associated with old one has gone away, and the code is not fully re-entrant. So, the old thread is deallocating resources that the new thread needs, or something like that... It may take a little while to disentangle this stuff. This is one of the joys of open source!
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-284216647, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tX6jBx8SH1lokXUUvXRpRVH58liCks5rioLvgaJpZM4MMgFX.
mikebrady
commented
7 years ago Hi there. I've been cleaning up code that I suspect, but can't prove, is possibly causing problems. It's now on the development branch. I'd be obliged if you tried it out, to see it it's any better.
ncp1113
commented
7 years ago Sure, Mike, I will. Just so you know, I haven’t had a problem since I turned of soxr. But I will be happy to update the software and let you know if I have a problem.
Nancy
On Mar 11, 2017, at 3:26 PM, Mike Brady notifications@github.com wrote:
Hi there. I've been cleaning up code that I suspect, but can't prove, is possibly causing problems. It's now on the development branch. I'd be obliged if you tried it out, to see it it's any better.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-285897593, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tZf3hGtZEN_GzRYFkFVEv2DH4M7aks5rkwOLgaJpZM4MMgFX.
mikebrady
commented
7 years ago Thanks – whenever is convenient. I've made the player thread fully (?) reentrant, so it should have tidied up the issues I think I identified. If you wanted to get a little more detail on what's happening, set the log_verbosity to 1 or 2 and some extra stuff will be recorded in the system log, accessible via the journalctl command, e.g. $sudo journalctl -n 100 -f...
ncp1113
commented
7 years ago Mike, I updated and restarted. Now I am only getting sound out of the left speaker. We are using hdmi. We checked to make sure it is not the receiver. I then recompiled from scratch. Still only one speaker. I rebooted. No luck. I put soxr back on. Still no luck.
I have verbosity set to 3. But the log is not informative. Here is my conf general = { name = "Den"; interpolation = "soxr"; log_verbosity = (3); udp_port_base=0 };
alsa = { output_device = "hw:0,1"; mixer_control_name = "PCM"; mixer_device = "hw:0"; };
sessioncontrol = { run_this_before_play_begins = "/home/pi/turnonreceiver"; run_this_after_play_ends = "/home/pi/turnoffreceiver"; wait_for_completion="yes"; allow_session_interruption="no"; };
I am not familiar with journalctl. Here is the syslog using cat pi@raspberrypi:~ $ cat /var/log/syslog | grep "shairport" Mar 12 16:41:35 raspberrypi shairport-sync[31507]: Successful Startup Mar 12 17:12:59 raspberrypi shairport-sync[31644]: Successful Startup Mar 12 17:19:00 raspberrypi shairport-sync[31740]: Successful Startup Mar 12 17:24:34 raspberrypi shairport-sync[3661]: Successful Startup Mar 12 17:24:49 raspberrypi shairport-sync[3725]: Successful Startup Mar 12 17:28:16 raspberrypi shairport-sync[782]: Successful Startup
On Mar 12, 2017, at 10:04 AM, Mike Brady notifications@github.com wrote:
sudo journalctl -n 100 -f...
mikebrady
commented
7 years ago Thanks very much for this. I'm really surprised about the left speaker only thing. A couple of small things. The verbosity should not have brackets around it and you probably don't need the udp_port_base. But none of these explain the left speaker only. I'll check as soon as I can.
ncp1113
commented
7 years ago I tried reverting to two different commits and I am still having the problem with only one speaker. So I am thinking that it has nothing to do with shairport-sync. It might have something to do with the apt-get upgrade that I did before upgrading shairport-sync.
On Mar 12, 2017, at 1:11 PM, Mike Brady notifications@github.com wrote:
Thanks very much for this. I'm really surprised about the left speaker only thing. A couple of small things. The verbosity should not have brackets around it and you probably don't need the udp_port_base. But none of these explain the left speaker only. I'll check as soon as I can.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-285958731, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tZo0IyRS8edbzjfePaMaoT3nY1f3ks5rlCdGgaJpZM4MMgFX.
mikebrady
commented
7 years ago I've just been experimenting with an original pi, going via hdmi to a TV, and am not getting the problem with the one speaker. What I have noticed though is that to fully reset the Pi after a problem on the HDMI port, it is necessary to actually power it down and then power it back up. Merely rebooting won't work.
ncp1113
commented
7 years ago Okay. This is embarrasing. Thanks for all of your help, Mike. It turned out to be a coincidence that I upgraded then there was no right channel. We plugged a device into the receiver and we still weren’t getting the right speaker. There was a wire unplugged.
I am going to bring shairport-sync back up to date and let you know how if we have a crash.
I’m so sorry for sending you down a rabbit hole.
Nancy
On Mar 12, 2017, at 3:37 PM, Mike Brady notifications@github.com wrote:
I've just been experimenting with an original pi, going via hdmi to a TV, and am not getting the problem with the one speaker. What I have noticed though is that to fully reset the Pi after a problem on the HDMI port, it is necessary to actually power it down and then power it back up. Merely rebooting won't work.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-285969568, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tX1jqgZBh6cLSO5fC_-TFdY9t-6cks5rlEl-gaJpZM4MMgFX.
mikebrady
commented
7 years ago I'm glad the speaker thing is resolved, Nancy. Please let us know how you do with the update.
ncp1113
commented
7 years ago I remembered a very old joke. How many programmers does it take to change a light bulb? None. That’s a hardware problem.
Thanks for your understanding.
On Mar 12, 2017, at 4:58 PM, Mike Brady notifications@github.com wrote:
I'm glad the speaker thing is resolved, Nancy. Please let us know how you do with the update.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-285976015, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tWRqucr_PiCXLYL2ty2IihvJKGytks5rlFyKgaJpZM4MMgFX.
ncp1113
commented
7 years ago Hi, Mike,
I just want to let you know that we haven’t had a single crash since I took off soxr interpolation. So your first instinct was right.
Thanks so much for all of your help.
Nancy
On Mar 12, 2017, at 4:58 PM, Mike Brady notifications@github.com wrote:
I'm glad the speaker thing is resolved, Nancy. Please let us know how you do with the update.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-285976015, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tWRqucr_PiCXLYL2ty2IihvJKGytks5rlFyKgaJpZM4MMgFX.
mikebrady
commented
7 years ago Thanks. Good joke too, by the way! 😊
ncp1113
commented
7 years ago If you live long enough you get to recycle old jokes.
On Apr 1, 2017, at 3:53 PM, Mike Brady notifications@github.com wrote:
Thanks. Good joke too, by the way! 😊
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mikebrady/shairport-sync/issues/482#issuecomment-290943209, or mute the thread https://github.com/notifications/unsubscribe-auth/ABl2tSOvjiH1ace2RZ4_HGAmoZsIXIsrks5rrqsigaJpZM4MMgFX.
uzlonewolf
commented
6 years ago I too am seeing an occasional 11/SEGV. Raspberry Pi 3, installed via "Super Simple Raspberry Pi Audio Receiver Install" https://github.com/BaReinhard/Super-Simple-Raspberry-Pi-Audio-Receiver-Install . I'm not sure what version they're installing, shairport-sync --version reports 3.2RC12-OpenSSL-Avahi-ALSA-stdout-pipe-soxr-metadata-sysconfdir:/etc and interpolation = "soxr" is set in the configuration file. As a workaround for now I have added "Restart=on-failure" to the systemd unit file.
FnasBas
commented
6 years ago I have this issue to. It's 100% reproducible for me by switching between soxr and basic. I'm using a raspberry pi 3 model b+.
I'm on the latest release, 3.2RC13.
The workaround using Restart=on-failure is sub-optimal in my case since I also use scripts to turn off the amplifier, and as this is in a car with finite battery power available, the turnoff-script not being called can be a very bad thing indeed.. ;) I know there's ExecStopPost= in systemd, but my shutdown script also has a delay in it so the amplifier doesn't switch on and off constantly as soon as anything interrupts the audio playback, so it doesn't play well together.
For now I'll use basic instead of soxr, but soxr sure would be nice.
Some more info:
root@volvon:~/shairport-sync# apt-cache show libsoxr-dev Package: libsoxr-dev Source: libsoxr Version: 0.1.2-2 Architecture: armhf
My configure-line is:
./configure --with-soxr --with-alsa --sysconfdir=/etc --with-avahi --with-ssl=mbedtls --with-metadata --with-systemd
Mike, this project is great and I thank you sincerely for all the work you put in to it. It's invaluable. I've used a different approach to the one laid out in CAR INSTALL.md, I'd like to offer it, how would I best go about doing it? sorry to go off topic..
mikebrady
commented
6 years ago Thanks for the report. I will investigate further but the bottom line is that the Pi isn’t really powerful enough for soxr interpolation, and unfortunately, that part of the code is single threaded.
If you have a different way of setting up a car installation, perhaps you’d be good enough to post it here, and then we can decide whether to add it as an alternative. Would that work for you?
FnasBas
commented
6 years ago I don't want to go against you on this, but on my pi 3 model b+, shairport uses about 30% of one core with soxr enabled, and as far as I can see there are no hidden kernel tasks or anything else loading the CPU.
Sure I can post my car approach here, just wanted to check first as not to fill this thread up with unrelated stuff.
FnasBas
commented
6 years ago Here's a starting point. I hope it is useful to someone :)
This is offered in the hopes it will be useful to someone. The background is that, besides all the advantages of wifi/airplay over bluetooth, when travelling my phone is my only internet connection, and I want to be able to share it with other devices. Like an iPad or laptop. I may also want to be able to send audio to the car shairport-sync service from my other devices while simulteneously having internet access on them, via the phone.
So the basic idea is just to turn the table on which device acts as the Access point for the wifi. Which works perfectly well. The prerequisite is that your carrier/contract allows teathering as otherwise you won't be able to have your phone act as a wifi access point. I also think this is a bit simpler to set up as you don't need to set up hostapd and dhcp server on the Pi. The drawback is that you need to do a gesture and tap a button on the phone to make it's network discoverable when entering your car.
Enable teathering on your iPhone with a secure passphrase.
Set up shairport according to your needs and instructions found elsewhere.
Configure your Raspberry pi with only one wifi network, through experimentation I found that if I was in the vicinity of my home network, the pi would not reliably choose the iPhone's wifi network even though it was prioritized over the home network. And it seems wpa-supplicant in general is a bit difficult to get to roam between different wifi networks once it's decided on one.
Example of /etc/wpa_supplicant/wpa_supplicant.conf
ctrl_interface=DIR=/var/run/wpa_supplicant GROUP=netdev
update_config=1
country=SE
network={
priority=3
ssid="My unique clever wifi network name"
psk="My very secure wifi key"
} By configuring your Pi to have a read only root filesystem, you need not ever worry about an improper shutdown of the OS causing the Pi to fail to boot. You will also be sure that the Pi always boots to a predetermined state. And you will eleminate wear on the SD card completely except for when modifying the system. This will result in a very reliable Airplay service.
These instructions are a bit old but a good starting point: (https://openenergymonitor.org/forum-archive/node/5331.html)
In addition to what's laid out there I've also:
In my case I do not use a car stereo but go straight Pi>Dac>Power amplifier. So I've also set up EQ with alsa, and a relay module to a GPIO pin and some scripts that run on session start/stop that turns the power amplifier on and off accordingly. But this could turn in to "How I set up my car audio" rather than offering an alternative to what's currently suggested for a Car installation. So I stop. ;)
Make your phone's network discoverable. In iOS 11 there's a shortcut for this, swipe up and hold down on the connectivity section. There's a button to turn on Internet sharing. In my experiments the Pi connects almost instantaneously to this network and a few moments later the airplay service offered by shairport-sync will become visible.
Make your phone network discoverable. The Pi connects. Connect your laptop to the same network. The iPhone does not appear to resolve host names from one tethered device to another so you need to ssh to the Pi's IP-address. There's a number of ways to find the Pi's IP-address:
ssh to the pi
rpi-rw
change stuff
rpi-ro Other apple devices can wake the wifi access point on the iPhone. My conclusion is that this must be accomplished via bluetooth, but Apple doesn't seem to be open spec on this subject as all I've been able to find is that certain devices, when paired with bluetooth, connected to icloud etc etc will offer this functionality. Being able to reverse engineer what bluetooth communication is actually done in order to turn on the wifi access point on the phone would mitigate the drawback of this method, as the raspberry pi could be constantly looking for the phone via BT and enabling Internet sharing that way.
mikebrady
commented
6 years ago Very useful, thanks. I'll get to it in a few days.
mikebrady
commented
6 years ago Hi there. I've just posted a new file in the development branch. Let me know what you think.
FnasBas
commented
6 years ago If it's good enough for you, it's good enough for me :) I'm happy to have contributed in some some way to this project.
It's written very much from my point of view, and if one would want, it could be reworked to look a bit more like proper documentation, but I guess this is such a fringe use case that it's good enough. If anyone is thinking of using shairport-sync i their cars, both car install files together should provide a good starting point.
On the line "Example of /etc/wpa_supplicant/wpa_supplicant.conf" there're backslashes in front of the , I'm not that experienced in markdown, bu I think I had to but a \ in front of the to escape them. So maybe double escapes snuck in there.
mikebrady
commented
6 years ago I've just fixed a bug that might be causing the segmentation error -- see #722. I'd be grateful if you would put the latest development version (3.2d67 or later) through its paces and let me know how it goes, please.
FnasBas
commented
6 years ago Great! I am on the road though so maybe a bit daring to go development? I'll give it a shot ;)
I found a huge disadvantage btw with my car setup approach. Something that never occurred to me. That is that if you're without cell phone coverage, the phone wont properly act as a WiFi access point.
But I still need that Internet sharing when I have it. So maybe some hybrid set up would be ideal. Need to figure out if and how one would do e.g. provide both a network from the Pi and have it probe for the phones network and whenever found, jump to that rather than being an AP.
If it's possible to get stable enough, maybe we should consider merging the two car install documents. Totally understand if this is no priority to you. I'll let you know how it goes.
Perhaps there should be a separate thread for Car setups.. if there's enough interest for it.
FnasBas
commented
6 years ago Ok, built and installed development. Turned on soxr. Tried connecting, disconnecting 4-5 times. Can see that my disconnect-script gets run. No errors reported!
Maybe you fixed it? Knock on wood, etc.. :)
Edit, add version info:
3.2d67-mbedTLS-Avahi-ALSA-soxr-metadata-sysconfdir:/etcGreat, thanks for trying it out. I've been trying to break it here without success, so I think it's a real fix. Let me think a bit more about the car installation thing -- I like the idea about a car setup thread.
FnasBas
commented
6 years ago It seems to still work fine! I think you fixed it!
Since there's no car thread, I continue to misuse this one.
However my problem with no cell reception became a real problem. It's nice to be forcibly disconnected from everything, but not without my music! I'm in a part of the country where basically only one carrier has coverage. So I had to fix it. And the only way of working on the pi was removing the sd-card, editing config files on my lap-top, re-inserting it and hoping for the best. So not the best of circumstances. Either a local console or an ethernet connection would have helped greatly. (thanks Apple for removing ethernet from the macbook pro..)
Anyways, the way it's set up now is: At boot, look for the iphones network If not available, start up hostapd and dnsmasq as dhcp-server. ta-da. It boots so quick and I have the power switch very handy so it's completely acceptable for the moment.
The limitation is obviously that it only happens at boot, and that it's a hack. But I needed to have music on my trip :) I'll probably work on getting it more stable when I get home.
The point of my alternate car approach was of course to have booth teathering and airplay available. And only audio when there's no coverage. Which never happens at home.. So the setup I'm going to gp for is to primarily use the Phone as an AP whenever possible and then with some degree of automation falling back to the pi providing it's own network.
It's getting complex, but fun as well :)
By the way. My phone (6S iOS 11.4) doesn't behave as you claim in the CAR SETUP.md file. Maybe I've missed something, but my dhcp server is not pushing any gateways or routes. Just assigns an IP on a /24 network. When the phone is connected to this wifi and has cell coverage, I can't get internet access. I've probably missed something. Don't care for the moment. Have music, Happy traveling. :)
mikebrady
commented
6 years ago Thanks. I've created a Car Installation Discussion Thread -- maybe you could move car installation discussions to there.
Might it now be possible to close this issue?
FnasBas
commented
6 years ago Sorry if you where waiting for confirmation. Yes as far as I can tell, this can be closed.
mikebrady
commented
6 years ago Great. Thanks for your help!
I am getting an error main process exited, code=killed, status=11/SEGV. It occurs when the device switches from Airplay to the local device sound. I am using scripts to turn on the receiver when shairport-sync starts and stop. turnonreceiver:
turnoff receiver:
My config file is:
Here is my log: