Hi!
On Fri, Nov 18, 2022 at 12:11 AM Chet Ramey <chet.ramey@case.edu> wrote:
>
> On 11/17/22 5:43 AM, Tadeus Prastowo wrote:
[...]
> >>> Is it true that the intention of the statement in the man page is that
> >>> at most one coprocess exists for every concurrently running Bash
> >>> process where $$ expands to the same PID?
> >>
> >> No. Coprocs are process-based.
> >
> > In that case, why does Bash 4.3.48 when executing the following lines
> > issues the coproc-exists warning despite the fact that if I understand
> > it correctly, the (...)-construct creates a new process?
> >
> > coproc sleep 1
> > echo $$, $BASHPID, $COPROC_PID >&2
> > (coproc sleep 1; echo $$, $BASHPID, $COPROC_PID >&2; wait)
> > wait
>
> A (subshell) is an exact copy of the parent shell, so it inherits the
> coproc. It just won't be able to write or read to the coproc, and can't
> wait for a coproc it inherits, since the subshell is not the coproc
> process's parent.
Given that coprocs are process-based and a subshell is a new process
despite being an exact copy of the parent process, I think logically
it follows that it is not a bug if the subshell creates a coproc.
Therefore, Bash back in the version 4.3.48 should have not issued the
coproc-exists warning.
Is my reasoning correct?
Thank you.
--
Best regards,
Eus