Segmentation fault with latest version

[MartenR]

Thanks, but you misunderstood me. I wanted log files with the new code.
Even if it looks fine too you, it is possible, that vomp uses some fall back code, since it is likely that no CEA modes are available.
In this case I might add a second hdmi power on command before querying the cea modes, that is for getting the best availiable mode.

Marten

[clausmuus]

Hi,
I took the log files with the latest vompclient version (the version of the 08.11.2012).
After every reboot, the vompclient starts the first one without problems. Mostly after a restart of the vompclient, the TV lost the connection and shows "no signal",...

Claus

[MartenR]

Mostly after a restart of the vompclient

What does a restart mean, really restating the executable or using the power key to turn off the display?
Or did you kill it? (Then anything can happen)

Mostly after a restart of the vompclient, the TV lost the connection and shows "no signal",...

This is because, the firmware says no hdmi cable connected so we are falling back to analog.

Please try to add in the beginning of  VideoOMX::selectVideoMode to add

Code Select Expand

vc_tv_hdmi_power_on_preferred();
This will first power on hdmi, before doing anything else.
May be you have to add a

Code Select Expand

MILLISLEEP(100);
after it. (Play with the 100).

Marten

[clausmuus]

Hi,

with restart I mean to kill and start the vompclient.
I will try your suggestions tomorrow.

Does you power off the hdmi if I kill the vompclient? It seems so, because if I kill it, the TV switched to an other source. If I start the vompclient, the TV switch to the vompclinet hdmi source, but if it fails (about every second try), it switched back to an other source after one second.
On all this tests I only start the vompclient, and do not select a TV channel. Also I have no seg fault with this vompclinet version.

Claus

[MartenR]

with restart I mean to kill and start the vompclient.

What kind of kill with signal 9 or so that it can exit gently.

On clean exit the frame buffer device is restored, may be this causes some problems with your hdmi set.
Also libcec is torn down, probably also powering off your tv??

Marten

[clausmuus]

Yes, I made a "kill -9" since a SIGTERM do not kill the vompclient in most cases. OK, I try the SIGTERM some weeks ago at last. I will check if a SIGTERM will be better.

Claus

[MartenR]

Well, if you use -9, then it clear that it can not work, because the routines for restoring the raspberry state are not called.

SIGTERM should always work, if not it is a bug, which has to be fixed.

Marten

[clausmuus]

Hi,

OK, now I have made some tests. The kill SIGTERM works. But it makes no different to a kill -9.
I have add your changes and detect, that it helps with a millisleep of 3000 (2000 is to less). If I take a look on the "tvservice -m CEA" output until I start the vompclient, I see, that for about 0.2 Seconds no modes avaliable. In the case, that the start get well, there are two times 0.2 Seconds (with a distance of 0.5 Seconds) no modes available. So I think, it is necessary, to test for some Seconds, if there will be modes available. Or better, if you can found the reason why this filed (now exact) every second try, to fix it if it's possible.

Also I have a other with. If I stop the vompclient, my TV switched to a other input source. If I than switch the TV beck to the hdmi source, I can see the terminal. Is it possible, to let the hdmi source activated on vompclient stop, so that the TV don't switch to an other source?

Claus

[MartenR]

I have add your changes and detect, that it helps with a millisleep of 3000 (2000 is to less).

Ok, then forget about the changes, 2-3 seconds is too long.
I have another idea, go to remotelinux.cc and comment the line:

Code Select Expand

                cec_adap->SetInactiveView();
Tell me what is happening then, maybe changing the cec stuff will help in the end. (This is not a solution but something, that helps me to get anidea waht is happening.)


Marten

[clausmuus]

OK, this helps. Now the vompclient starts every time correct.

Also the TV do not switch to an other source on vompclinet stop.

Claus

[MartenR]

@clausmuus
I did not have time to lock into this, this weekend.
Maybe next, I had a very difficult bug to find, connected with signal off and on.
Try the git maybe it is solved by accident, but I do not believe so.

Marten

[clausmuus]

Hi Marten,

that's no problem, since I give back my borrowed RPI today and not get already my own. I think I will get my own one this or next week. But until next week I will be in vacation for two week, so now you have three weeks to find a solution

Claus