[LTP] [PATCH 2/2] tst_test: using SIGTERM to terminate process

[Joerg Vehlow]

Hi Cyril,

On 5/31/2021 2:54 PM, Cyril Hrubis wrote:
>>>> But one more strange thing here.
>>>> I wonder why this even works. $pid is used in _tst_kill_test and defined
>>>> in _tst_setup_timer as a local variable.
>>>> It looks like it is inherited through the call chain and since it is
>>>> copied to the background process, it cannot be manipulated by the tests.
>>>> Still I would vote for changing this at some point, because it is highly
>>>> confusing.
>>> That's actually a correct and well defined behavior, if you call a
>>> function g from function f the function g has access to the variables
>>> local to f.
>>>
>>> And yes it's confusing, but the alternative is having another global
>>> variable which I do not think is much better than this.
>> Hm shell code has more strange behavior than I would have ever expected...
> Strongly agree here, to be honest if there was as better scripting
> language that would be widely avaiable I would have switched long time
> ago...
>
>> But at least Li and myself did not know that and even while you know
>> about this "feature", you think it is strange.
>> I would rather like to be explicit and use a global variable (what harm
>> is it really?) instead of confusing the next one looking at this piece
>> of code....
> If you really think global variable would be better, then go ahead and
> send a patch, I do not have a strong feelings about this particular
> detail.
I think there is another solution, to make this at least more explicit 
without using a new global variable.
Passing the pid as a parameter to _tst_timeout_process and _tst_kill_test.
Although it is not required, it makes it a lot less implicit and a lot 
better simpler to read.

Jörg