neumoDVB

[RyLe]

Hello friends ;
I use 3 programs in ubuntu Tvheadend, Openpli PC-TV and Kodi (I keep getting the same errors while trying to install neumodvb, what should I do to fix these errors.)

Thank you in advance for your help.

Hi

From the report, it seems that you have a 20.04 Ubuntu

-- xxx=/focal/

and Clang 10.0:

-- The C compiler identification is Clang 10.0.0 -- The CXX compiler identification is Clang 10.0.0

Neumodvb needs Clang 14:

-- Ancient Unbuntu detected - unstall clang-14 first!!!

So you should at least update Clang to the required version, and you should really follow all the steps on this page.

 

[ptergrand]

I'm facing the same errors

/home/star/neumodvb/src/util/util.cc:124:3: error: expected expression
else
^
1 error generated.
make[2]: *** [src/util/CMakeFiles/neumoutil.dir/build.make:63: src/util/CMakeFiles/neumoutil.dir/util.cc.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:1753: src/util/CMakeFiles/neumoutil.dir/all] Error 2
make: *** [Makefile:130: all] Error 2

 

[deeptho]

So if I understand correctly, outside of HTS use (for internet for example ?) the 6909-X is better suited for feedhunting than the 6903-X ?
The absence of a "big" heatsink doesn't make the 6909-X too hot when parallel blindscanning ?

For very low symbol rates I think I saw receptions of small TP for radios using 6903-X /6909-X so it should be quite OK on this side (even if I think you may be talking about cards like a 6983 that I already got). I also got a 6904SE for the PLS auto discovery if that's what you meant.


For the price question, in Europe/France the 6909-X is ~35€ cheaper than the 6903-X using tbs-technology.de and about 30€ cheaper using amazon.com, shipping and taxes included.

I will probably buy the 6909-X if that's the better choice for purely feedhunting, but I will need your help with the current bug (?) that makes neumodvb crash on my machine when I try to view a channel.

If it is cheaper , then definitely buy 6909x. There is no danger of overheating and the 6909x has a heatsink as well.
Low symbol rate is a matter of some luck with these cards, but it is possible to receive some muxes with
between 200 and 1000 Ks/s.

 

[deeptho]

I'm facing the same errors

/home/star/neumodvb/src/util/util.cc:124:3: error: expected expression
else
^
1 error generated.
make[2]: *** [src/util/CMakeFiles/neumoutil.dir/build.make:63: src/util/CMakeFiles/neumoutil.dir/util.cc.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:1753: src/util/CMakeFiles/neumoutil.dir/all] Error 2
make: *** [Makefile:130: all] Error 2

Same problem => same solution.

Why not upgrade your ubuntu?

 

[RyLe]

If it is cheaper , then definitely buy 6909x. There is no danger of overheating and the 6909x has a heatsink as well.
Low symbol rate is a matter of some luck with these cards, but it is possible to receive some muxes with
between 200 and 1000 Ks/s.

Thanks for the confirmation, I just placed an order for a 6909-X.

 

[RyLe]

Today I received my 6909-X.
I had to recompile the tbsecp3 driver, changing i2c_speed back to 9 (from 4) as I was getting i2c errors in dmesg (no frontend could be attached).
I successfully did some parallel blindscans, they are indeed pretty fast

I still need to resolve the crash I get when I try to view a channel through neumodvb.

 

[deeptho]

Today I received my 6909-X.
I had to recompile the tbsecp3 driver, changing i2c_speed back to 9 (from 4) as I was getting i2c errors in dmesg (no frontend could be attached).
I successfully did some parallel blindscans, they are indeed pretty fast

You can try other values between 4 and 9 and then report the lowest value that works.

I still need to resolve the crash I get when I try to view a channel through neumodvb.

 

[RyLe]

You can try other values between 4 and 9 and then report the lowest value that works.

Hi

It starts working again with a value of 7.

I don't know if this value will be of any help for others as I have quite a convoluted setup:

*Proxmox 8.X =>Fedora 38 VM with pciex passthrough for the tuner.
*Intel NUC9i9QNX=>Thunderbolt 3 cable=>Razer Core X Thunderbolt 3 case=>TBS 6909-X

(I couldn't use one of the NUC's internal pciex ports as the other is taken by a 10Gb NIC and the two ports share the same pciex root.
I can't do passthrough on one of them without the other also getting kicked out of Proxmox).

The segmentation fault I get in neumodvb looks like this in the /tmp/neumo.log:

2023-07-22 14:47:01,923 DEBUG: [tuner ] tune add ch action src/receiver/tune.cc:256 2023-07-22 14:47:01,936 DEBUG: [tuner ] TUNE ADD_CHANNEL STARTED BBC One Lon HD er/active_adapter.cc:291 2023-07-22 14:47:01,936 DEBUG: [receiver RECEIVER-CMD SUB[-1] [00000] 10817.303V - BBC One Lon HD] SUBSCRIBE - returning to caller receiver/receiver.cc:1258 2023-07-22 14:47:01,936 DEBUG: [CH[4:10363]BBC CH[4:10363]BBC One Lon HD] OPEN DEMUX_FD=49 ver/active_stream.cc:113 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] PLAY SUBSCRIPTION (service): subscribed=0 c/viewer/neumompv.cc:695 2023-07-22 14:47:01,936 DEBUG: [CH[4:10363]BBC CH[4:10363]BBC One Lon HD] Adding pid=0 ver/active_stream.cc:118 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] Register audio_changed_cb subscription_id=0 s=0 iver/playback_mpm.cc:844 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] PLAY SUBSCRIPTION (service): mpm init done c/viewer/neumompv.cc:704 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] Starting move_to_time iver/playback_mpm.cc:336 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] Opening /home/ryle/neumo/live/A04_ts02048_sid10363_20230722_14:47:01/00_20230722_144701.ts iver/playback_mpm.cc:522 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] currently_playing_file.fileno=0 fd=50 iver/playback_mpm.cc:539 2023-07-22 14:47:01,936 ERROR: [mpv/opener ] Could not obtain last marker iver/playback_mpm.cc:906 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] Requested start_play_time is beyond last logged packet iver/playback_mpm.cc:396 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] MMAP 0 127827968 ceiver/filemapper.cc:139 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] PLAY SUBSCRIPTION (service): mpm move_to_live done c/viewer/neumompv.cc:711 2023-07-22 14:47:01,936 DEBUG: [mpv/opener ] /home/ryle/neumodvb/src/viewer/neumompv.cc_376 TIME: 565 vb/src/util/logger.h:158

 

[deeptho]

for a segmentation violation the logs are useless. Instead you should run neumodvb from a debugger and get a stacktrace.
Somethign like
cd ~/neumodvb/gui
gdb
-> type the following in gdb
set breakpoint pending on
exec-file /usr/bin/python3
set args neumodvb.py
set debuginfod enabled off
set pagination off

First recompile ig needed with better options for debugging: in the toplevel CMakeLists.txt
make sure it is like below:

add_compile_options (
# $<$<CONFIGEBUG>:-Og>
-fPIC -fsized-deallocation -DHAS_UNCAUGHT_EXCEPTIONS=1 -D DATA_DIR=${CMAKE_SOURCE_DIR}/data)
#add_link_options( -lasan)


Are you sure any video can be played at all? Does the mpv command play videos
?

 

[RyLe]

for a segmentation violation the logs are useless. Instead you should run neumodvb from a debugger and get a stacktrace.
Somethign like
cd ~/neumodvb/gui
gdb
-> type the following in gdb
set breakpoint pending on
exec-file /usr/bin/python3
set args neumodvb.py
set debuginfod enabled off
set pagination off

First recompile ig needed with better options for debugging: in the toplevel CMakeLists.txt
make sure it is like below:

add_compile_options (
# $<$<CONFIGEBUG>:-Og>
-fPIC -fsized-deallocation -DHAS_UNCAUGHT_EXCEPTIONS=1 -D DATA_DIR=${CMAKE_SOURCE_DIR}/data)
#add_link_options( -lasan)


Are you sure any video can be played at all? Does the mpv command play videos
?

On this setup, mpv can play videos, even if it's clearly not optimal without a true hardware acceleration due to the remote connection ( with SPICE (qxl), VirtIO-GPU or VirGL-GPU as the GPU for the VM).

Just adding # in front of $<$<CONFIG:DEBUG>:-Og> and recompiling neumodvb made the channel viewing with CTRL+Enter usable without any crash.
Now it only crashes if I try to add a second channel in PIP using CTRL+SHIFT+Enter.
It probably has to do with the remote display thing and low video resources I think.
If I remove the #and recompile, the crashes are back when trying to view a single channel.

I don't really want to use the channel viewing of neumodvb because of the remote access side of things but at least I wanted it to work.

I will mainly use your neumodvb set of tools in a setup like this:
-parallel blindscan with stid135-blindscan
-look for feed type transponders in /tmp/blindscan_rf0.dat
-test those transponders using the spectrum tool in neumodvb (locking, level of reception, is it a GSE or a TS one ?, etc)
-stream the data via network to VLC, TSReader Lite or other tools using dvbsnoop/dvbstream and netcat (or socat)

Thank you so much for all your work, making the 6903-X and 6909-X usable for quick blindscan

 

[satesco]

...

Just adding # in front of $<$<CONFIG:DEBUG>:-Og> and recompiling neumodvb made the channel viewing with CTRL+Enter usable without any crash.
Now it only crashes if I try to add a second channel in PIP using CTRL+SHIFT+Enter.
It probably has to do with the remote display thing and low video resources I think.
If I remove the #and recompile, the crashes are back when trying to view a single channel.

I don't really want to use the channel viewing of neumodvb because of the remote access side of things but at least I wanted it to work.

...

I have the same problem with neumoDVB crashing when I want to watch programs simultaneously. I'm surprised that you work on Fedora. So far, Deeptho and Llew have shown us that there are no problems with neumoDVB in this respect, because the former has developed the application on Fedora, and the latter has demonstrated that you can work more programs on neumoDVB.
In my case, on Ubuntu, the crash is due to a bug in the Mesa library, which can only be fixed with a patch. And here, Deeptho is right, at least for Ubuntu, but in the case of Fedora, you'll probably fix it yourself. You seem to be an expert in Linux, or at least you're no stranger to how this OS is handled.
I will follow your intervention to see how you solve this problem. I hope it will be the same for Ubuntu 22.04.1.

 

[Llew]

I have the same problem with neumoDVB crashing when I want to watch programs simultaneously. I'm surprised that you work on Fedora. So far, Deeptho and Llew have shown us that there are no problems with neumoDVB in this respect, because the former has developed the application on Fedora, and the latter has demonstrated that you can work more programs on neumoDVB.
In my case, on Ubuntu, the crash is due to a bug in the Mesa library, which can only be fixed with a patch. And here, Deeptho is right, at least for Ubuntu, but in the case of Fedora, you'll probably fix it yourself. You seem to be an expert in Linux, or at least you're no stranger to how this OS is handled.
I will follow your intervention to see how you solve this problem. I hope it will be the same for Ubuntu 22.04.1.

Problems with a Samsung SSD meant I had to reinstall Fedora on another disk.

I now have exactly the problem you and RyLe have since compiling, i.e crashing when choosing another channel (Tune-Add).

Something changed or added in a recent update?

 

[RyLe]

I have the same problem with neumoDVB crashing when I want to watch programs simultaneously. I'm surprised that you work on Fedora. So far, Deeptho and Llew have shown us that there are no problems with neumoDVB in this respect, because the former has developed the application on Fedora, and the latter has demonstrated that you can work more programs on neumoDVB.
In my case, on Ubuntu, the crash is due to a bug in the Mesa library, which can only be fixed with a patch. And here, Deeptho is right, at least for Ubuntu, but in the case of Fedora, you'll probably fix it yourself. You seem to be an expert in Linux, or at least you're no stranger to how this OS is handled.
I will follow your intervention to see how you solve this problem. I hope it will be the same for Ubuntu 22.04.1.

You may be talking about this bug ?

I did more tests:
-If I run neumodvb normally and try to bring the overlay with CTRL+O, I get a segfault.
-If I run neumodvb under gdb with the commands listed by deeptho, the overlay is working, so I can't even know where the crash comes from.

On another subject:
-If I boot my Fedora VM but I don't log on gdm and instead I connect to this machine via SSH from my Windows machine, export DISPLAY to my Windows machine IP which is running the VcXsrv X Window server and try to run neumodvb, it displays the interface for a second or two, then it crashes with this:
AssertX: 0 failed at /home/ryle/neumodvb/src/receiver/devmanager.cc:311 in function void dvbdev_monitor_t::on_new_frontend(adapter_no_t, frontend_no_t)
-I can't access to my tuner either via SSH using neum-otune if I'm not logged on gdm:
[ryle@mir ~]$ neumo-tune -a4 -S 27500 -m QPSK --delsys DVBS -p H -f 11425000 -U 12 adapter=4 rf_in=-1 frontend=0 freq=11425000 pol=1 pls_codes[5]={ 4202496, 2048, 98139136, 134216704, 80015360, } diseqc=UC: U=12 C=-1 Blindscan drivers found open_frontend failed: Permission denied

-If I do the same thing when I'm logged on gdm before trying to run neumodvb or neumo-tune via SSH, then it works correctly on my Windows machine for both of them.

 

[satesco]

Indeed, this is what Deeptho describes in your link. Probably in your case, the bug is even more, who knows, insidious on Fedora. I hope it's not that complicated.
I have to wait, because I have depended on Deeptho since the beginning and I hope that eventually there will be a patch for Ubuntu.

 

[deeptho]

On this setup, mpv can play videos, even if it's clearly not optimal without a true hardware acceleration due to the remote connection ( with SPICE (qxl), VirtIO-GPU or VirGL-GPU as the GPU for the VM).

Just adding # in front of $<$<CONFIG:DEBUG>:-Og> and recompiling neumodvb made the channel viewing with CTRL+Enter usable without any crash.
Now it only crashes if I try to add a second channel in PIP using CTRL+SHIFT+Enter.

That is a problem in mesa (a library) and there is even a non
practical solution. Non practical because it is beyond the reach of
end users.
See

kwin patch causes program using two gl windows to crash (#9125) · Issues · Mesa / mesa · GitLab

System information System: Host: streacom.mynet...

gitlab.freedesktop.org

There seems no progress in fixing it, so it might help if someone reports
the problem once more.

It probably has to do with the remote display thing and low video resources I think.
If I remove the #and recompile, the crashes are back when trying to view a single channel.

No.

I don't really want to use the channel viewing of neumodvb because of the remote access side of things but at least I wanted it to work.

You can try accessing the vm with xpra. It is easy to use and even gives decent video playback speed. Setting up sound is a bit counter intuitive but
it also works.

I will mainly use your neumodvb set of tools in a setup like this:
-parallel blindscan with stid135-blindscan
-look for feed type transponders in /tmp/blindscan_rf0.dat
-test those transponders using the spectrum tool in neumodvb (locking, level of reception, is it a GSE or a TS one ?, etc)
-stream the data via network to VLC, TSReader Lite or other tools using dvbsnoop/dvbstream and netcat (or socat)

Thank you so much for all your work, making the 6903-X and 6909-X usable for quick blindscan

 

[deeptho]

I have the same problem with neumoDVB crashing when I want to watch programs simultaneously. I'm surprised that you work on Fedora. So far, Deeptho and Llew have shown us that there are no problems with neumoDVB in this respect, because the former has developed the application on Fedora, and the latter has demonstrated that you can work more programs on neumoDVB.
In my case, on Ubuntu, the crash is due to a bug in the Mesa library, which can only be fixed with a patch. And here, Deeptho is right, at least for Ubuntu, but in the case of Fedora, you'll probably fix it yourself. You seem to be an expert in Linux, or at least you're no stranger to how this OS is handled.
I will follow your intervention to see how you solve this problem. I hope it will be the same for Ubuntu 22.04.1.

That mesa bug is everywhere and will affect all distributions.

 

[RyLe]

Hi

With each kernel update from Fedora, I have to recompile the blindscan drivers.

Going from 6.2.X to 6.3.X, I had to modify
blindscan_kernel/media_build/v4l/Makefile
to add
EXTRA_CFLAGS += -Wno-error=incompatible-pointer-types
otherwise there were blocking errors.
It went ok for a few kernel updates.

Now Fedora pushes the 6.4.4-200 and I have a new error when I try to compile the drivers (a matching kernel-devel is installed) :

/home/ryle/blindscan_kernel/media_build/v4l/dvbdev.c: In function 'init_dvbdev': ./include/linux/export.h:27:22: warning: passing argument 1 of 'class_create' from incompatible pointer type [-Wincompatible-pointer-types] 27 | #define THIS_MODULE (&__this_module) | ~^~~~~~~~~~~~~~~ | | | struct module * /home/ryle/blindscan_kernel/media_build/v4l/dvbdev.c:1103:34: note: in expansion of macro 'THIS_MODULE' 1103 | dvb_class = class_create(THIS_MODULE, "dvb"); | ^~~~~~~~~~~ In file included from ./include/linux/device.h:31, from ./include/linux/input.h:19, from /home/ryle/blindscan_kernel/media_build/v4l/compat.h:47: ./include/linux/device/class.h:230:54: note: expected 'const char *' but argument is of type 'struct module *' 230 | struct class * __must_check class_create(const char *name); | ~~~~~~~~~~~~^~~~ /home/ryle/blindscan_kernel/media_build/v4l/dvbdev.c:1103:21: error: too many arguments to function 'class_create' 1103 | dvb_class = class_create(THIS_MODULE, "dvb"); | ^~~~~~~~~~~~ ./include/linux/device/class.h:230:29: note: declared here 230 | struct class * __must_check class_create(const char *name); | ^~~~~~~~~~~~ make[3]: *** [scripts/Makefile.build:252: /home/ryle/blindscan_kernel/media_build/v4l/dvbdev.o] Error 1


If I simply try to patch it using this technique, I also have to modify

/blindscan_kernel/media/drivers/media/pci/ddbridge/ddbridge-core.c because of
.owner = THIS_MODULE, that doesn't exist in the class anymore

and then

blindscan_kernel/media/drivers/media/pci/tbsmod/tbsmod.c because of the same error listed for blindscan_kernel/media_build/v4l/dvbdev.c

It compiles, it installs, but it crashes when it loads as the modifications I made are probably stupid without reading what the rest of the code does.

I could use some help

I'll revert to booting the previous kernel in the mean time.

 

[RyLe]

Ok, so, a few 6.4.X FC38 kernel updates went by and I was tired of having to go back to the old 6.3.X kernel each time, so I looked at this one more time.
The "technique" link I mentioned above is wrong as it gives the wrong parameter to class_create, I should have looked into the updated kernel sources sooner.
I can now build the modified drivers and have a functionnal setup using a 6.4.X kernel

Here are my modifications to be able to compile and use blindscan_kernel on a 6.4.X kernel:

in
~/blindscan_kernel/media_build/v4l/Makefile :
add
EXTRA_CFLAGS += -Wno-error=incompatible-pointer-types
near to other EXTRA_CFLAGS

in
~/blindscan_kernel/media/drivers/media/dvb-core/dvbdec.c :
change
dvb_class = class_create(THIS_MODULE, "dvb");
to
#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 4, 0) dvb_class = class_create(THIS_MODULE, "dvb"); #else dvb_class = class_create("dvb"); #endif

in
~/blindscan_kernel/media/drivers/media/pci/ddbridge/ddbridge-core.c
near class ddb_class, change
.owner = THIS_MODULE,
to
#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 4, 0) .owner = THIS_MODULE, #else // #endif
/!\warning/!\, as I don't have a Digital Devices card, I don't know if this modification is working.


in
~/blindscan_kernel/media/drivers/media/pci/tbsmod/tbsmod.c
change
mod_cdev_class = class_create(THIS_MODULE, "tbsmod");
to
#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 4, 0) mod_cdev_class = class_create(THIS_MODULE, "tbsmod"); #else mod_cdev_class = class_create("tbsmod"); #endif

And that's all

 

[deeptho]

Ok, so, a few 6.4.X FC38 kernel updates went by and I was tired of having to go back to the old 6.3.X kernel each time, so I looked at this one more time.
The "technique" link I mentioned above is wrong as it gives the wrong parameter to class_create, I should have looked into the updated kernel sources sooner.
I can now build the modified drivers and have a functionnal setup using a 6.4.X kernel

Here are my modifications to be able to compile and use blindscan_kernel on a 6.4.X kernel:

in
~/blindscan_kernel/media_build/v4l/Makefile :
add
EXTRA_CFLAGS += -Wno-error=incompatible-pointer-types
near to other EXTRA_CFLAGS

in
~/blindscan_kernel/media/drivers/media/dvb-core/dvbdec.c :
change
dvb_class = class_create(THIS_MODULE, "dvb");
to
#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 4, 0) dvb_class = class_create(THIS_MODULE, "dvb"); #else dvb_class = class_create("dvb"); #endif

in
~/blindscan_kernel/media/drivers/media/pci/ddbridge/ddbridge-core.c
near class ddb_class, change
.owner = THIS_MODULE,
to
#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 4, 0) .owner = THIS_MODULE, #else // #endif
/!\warning/!\, as I don't have a Digital Devices card, I don't know if this modification is working.


in
~/blindscan_kernel/media/drivers/media/pci/tbsmod/tbsmod.c
change
mod_cdev_class = class_create(THIS_MODULE, "tbsmod");
to
#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 4, 0) mod_cdev_class = class_create(THIS_MODULE, "tbsmod"); #else mod_cdev_class = class_create("tbsmod"); #endif

And that's all

It is best to not upgrade your kernel, otherwise you will have problems like this frequently.

 

[ptergrand]

Hello friends, I could not figure out how to set this program.