[mythtv-users] CTRL-ALT-BACKSPACE -> multiple instances of mythfrontend

[Stephen Worthington]

On Mon, 11 Mar 2019 09:41:08 -0700, you wrote:

>On Sun, Mar 10, 2019 at 7:31 PM Stephen Worthington <
>stephen_agent at jsw.gen.nz> wrote:
>
>> On Sun, 10 Mar 2019 14:59:19 -0700, you wrote:
>>
>> >I added a 10 second delay and the problem is gone, no duplicate instances
>> >of myth. Maybe 3 seconds would work but definitely a race condition.
>> Other
>> >than the delay, the code is the same as original.
>> >
>> >In fairness to the code, my test is not real life.  Typically if I were to
>> >do a CTRL-ALT-BS it would be because myth was frozen or missing. I don't
>> >know that I ever did a CTRL-ALT-BS just as a test with Myth running. The
>> >timing might be different and the service might be killed faster if it is
>> >actually frozen. Or maybe I am wrong and this really is an issue.  Either
>> >way, it should not keep piling on instances of mythfrontend.  But at least
>> >I think we know what is going on now.
>> >
>> >Allen
>>
>> Using a delay to fix a race condition is usually a bad idea -
>> something happens which changes the timing, such as a bit of swapping
>> going on, and then the problem is back again.
>>
>> Unfortunately, mythfrontend does die quite often for me.  The usual
>> cause is playing a downloaded video file.  I have quite a few that it
>> can not handle, and either it dies, or locks up and I have to kill it.
>> Some of the deaths leave behind "assert failed" messages.  But if you
>> do not normally play such files, then mythfrontend can be very
>> reliable and not need the wrapper script.
>>
>> It looks like the wrapper script needs to be able to detect that the
>> script itself is still running and to exit if it can see another copy
>> of itself.  This should work to do that:
>>
>> pid=$(ps -ef | grep '/bin/sh /usr/bin/mythfrontend --service' | grep
>> -v grep | awk '{print $2}')
>> if [ ${#pid} -ne 0 ]; then
>>     exit 0
>> fi
>> _______________________________________________
>>
>> I do not understand why this does not work but I am no script expert so I
>am not sure what exactly is going on with ${#pid} and how that relates to
>the initial pid. Perhaps there is a $ missing or misplaced on the first
>line?  I did run some tests.
>
>*dad at NewMyth:~$ ps -ef | grep '/bin/sh /usr/bin/mythfrontend --service' |
>grep -v grep*
>*dad       2310     1  0 09:17 ?        00:00:00 /bin/sh
>/usr/bin/mythfrontend --service*
>*dad       2605     1  0 09:17 ?        00:00:00 /bin/sh
>/usr/bin/mythfrontend --service*
>*dad       2961  2939  0 09:17 ?        00:00:00 /bin/sh
>/usr/bin/mythfrontend --service*
>
>And
>
>*dad at NewMyth:~$ ps -ef | grep '/bin/sh /usr/bin/mythfrontend --service' |
>grep -v grep | awk '{print $2}'*
>*2310*
>*2605*
>*2961*
>
>For now I will just run mythfrontend without the --service option.  That
>should be fine for my needs. I am still willing to troubleshoot this if you
>think that would be useful so let me know.
>
>I wish to thank both Stephen and Ian for your help.
>
>Allen

$(#pid} is the number of characters in the $pid string.  If there are
any copies of mythfrontend running, then the $pid string should be a
list of the PIDs for those copies (actually, a list of the PIDs of the
bash tasks running those copies of the scripts).  So if there are no
other copies of mythfrontend running, $pid should be an empty string,
and ${#pid} should be 0.  I tested it using this script (t1.sh):

#!/bin/bash
pid=$(ps -ef | grep '/bin/sh /usr/bin/mythfrontend --service' | grep
-v grep | awk '{print $2}')
echo $pid
echo ${#pid}
if [ ${#pid} -ne 0 ]; then
    echo "pid != 0"
fi

On my MythTV box, it displays the "pid != 0" string whenever it finds
a copy of the mythfrontend script running.  If I add an extraneous
character to the grep string (eg an extra d on the end of
"mythfrontend", it gets no matches and continues on to the end of the
script without doing the echo.