Re: PIPESTATUS differs from $? for compound command

[Greg Wooledge]

On Tue, Dec 10, 2024 at 09:28:31 +0100, Ulrich Müller wrote:
> Then what about these?
> 
> $ true; case a in esac
> $ echo "ret = $?, status = ${PIPESTATUS[*]}"
> ret = 0, status = 0
> 
> $ false; case a in esac
> $ echo "ret = $?, status = ${PIPESTATUS[*]}"
> ret = 0, status = 1
> 
> "case a in esac" is a one-command pipeline, and certainly it is executed
> after "false". So it should qualify as "the most-recently-executed
> foreground pipeline (which may contain only a single command)".

My own testing:

hobbit:~$ if true; then (exit 11) | cat; fi
hobbit:~$ echo "$? ${PIPESTATUS[*]}"
0 11 0

hobbit:~$ if true; then (exit 11) | cat; fi | false
hobbit:~$ echo "$? ${PIPESTATUS[*]}"
1 0 1

hobbit:~$ while (exit 12); do :; done
hobbit:~$ echo "$? ${PIPESTATUS[*]}"
0 12

hobbit:~$ true | while (exit 12); do :; done
hobbit:~$ echo "$? ${PIPESTATUS[*]}"
0 0 0


Given all that we've tried here, all I can determine is:

1) If an "if", "case" or "while" (and probably "until" too) is executed
   by itself as the sole command in a pipeline, the compound command sets
   the value of $? but not the value of PIPESTATUS.  It doesn't matter
   whether the "body" is entered.

2) If the same command appears as part of a pipeline with other commands
   in it, then the compound command's exit status does become part of
   PIPESTATUS.

It's hard to see how this is anything other than a bug.  But it's an
incredibly long-standing behavior -- all the way back to 2.05b or
earlier.  Given that PIPESTATUS is a bash extension, not part of POSIX,
it's up to Chet to decide what to do about it.