Re: "echo | ps -j $(:) | cat | cat | cat" runs components in different process groups

[Joey Pabalinas <joeypabalinas@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2514 bytes --]

On Fri, Mar 23, 2018 at 10:19:59PM -0700, Bart Schaefer wrote:
> It may be this easy?
> 
> It fixes the specific example in the subject by causing the zsh parent
> shell to re-assume job leader for all subsequent processes in the pipeline
> if the current leader is dead.  However, I'm concerned that if the first
> leader is reaped after a few processes in a long pipeline are created,
> there may still be a new process group asserted for remaining processes.
> 
> On the other hand I haven't found an example where that occurs, instead
> I get jobs where the process group ID is the PID of the now-dead leader.
> It could be I just haven't constructed the right test.

It's kind of an *incredibly* unlikely edge-case but FWIW:

> $ echo | ps -j $(: $(cat)) | cat | cat | cat

The shell _seems_ to completely lock up after applying your patch; on
my zsh 5.4.2 release version there are no problems (aside from the
original process group leader issues).

This is the shell output after I `pkill -9 cat`:

> $ echo | ps -j $(: $(cat)) | cat | cat | cat
> ^C^C^C^C^Z^Z^Z^Z
> PID  PGID   SID TTY          TIME CMD
> 16007 16007 18994 pts/5    00:00:03 zsh
> 17582 17582 18994 pts/5    00:00:01 zsh
> 18994 18994 18994 pts/5    00:00:26 zsh
> 20626 17582 18994 pts/5    00:00:00 ps
> 20627 17582 18994 pts/5    00:00:00 cat
> 20629 17582 18994 pts/5    00:00:00 cat

The shell _appears_ to deadlock and accordingly ignores all user
input such as ^C or ^D (but you can, however, still externally
kill the parent zsh process group leader with any fatal signal
or `kill -9 cat` without issue). It seems it's mostly an issue
stemming from the parent zsh process still being the process
group leader; this causes cat to wait indefinitely for input
while being unable to actually receive any.

In my zsh release version 5.4.2 ^C works as expected, and to
be honest I don't actually know if this is anything worth
being concerned about.

Output from the release version (no need to ^C or ^D or anything):

> $ echo | ps -j $(: $(cat)) | cat | cat | cat
> cat: -: Input/output error
> PID  PGID   SID TTY          TIME CMD
> 21366 21366 27814 pts/12   00:00:00 ps
> 21367 21367 27814 pts/12   00:00:00 cat
> 21368 21368 27814 pts/12   00:00:00 cat
> 27814 27814 27814 pts/12   00:00:03 zsh

The "cat: -: Input/output error" and differing number of entries is kind
interesting, although I would be lying if I said that I understood why.

-- 
Cheers,
Joey Pabalinas

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]