Re: [Help-bash] get pid of exited process

[Stas Sergeev]

04.01.2016 17:42, Greg Wooledge пишет:
> On Mon, Jan 04, 2016 at 05:47:27PM +0300, Stas Sergeev wrote:
> > 04.01.2016 17:22, Greg Wooledge ??????????:
> > > On Mon, Jan 04, 2016 at 05:15:03PM +0300, Stas Sergeev wrote:
> > > > As I explained in the same mail, I need it for cleanups:
> > > > if the program crashes, the script should remove the
> > > > stalled sockets/fifos that usually embed the pid in their
> > > > names.
> > > Eh?  If the program exits, it should clean up its own sockets.  Can't you
> > > fix the program, instead of trying to write this workaround wrapper?
> > Well, fixing program is certainly good, but even if I could,
> > please don't forget about "kill -9", in which case the stalled
> > resources are unavoidable.
> If someone is killing the program with kill -9, then your answer is to
> STOP THAT PERSON FROM DOING THAT.  This is not a technological problem.
> It's a human resources problem.
>
> Never, ever use kill -9 unless you are prepared to clean up the mess
> personally.
Unfortunately, in real world, users eventually do "kill -9",
and/or the programs eventually crash. I will ask the user to
not do that, and I will fix the program, but as things are,
you need to take into account both of these troubles.
I believe most of any program that uses the lock files,
have some code to check if the lock is stale, for example
(we can see minicom). I don't believe this can simply be
ignored, but I certainly don't want this to cost me more than
a couple of lines in a shell script. Because if it costs just that,
the question changes from "why do we care" to "why not".