bug reports have the best chance of being examined
if actual code accompanies the text description;
this is because it is very hard to attack a bug
that cannot be independently reproduced on another system;
otherwise there can be no independent verifictation
that a particular bug has been fixed
it also allows the bug to be categorized
e.g., solaris 11 only, little endian only,
the last release, all known instantaiations
for this example
what does "sub-shell" mean?
how is the suspend done within the script?
these questions would be answered by an annoted script
asynchronous/interactive actions can be commented inline
e.g., something like
( ... ) &
echo pid=$!
sleep 100
# at this point do "kill -TSTP $pid" from another window
# and then next statement is never executed
echo done
this particular case involves ksh93s+ and /bin/ksh,
so two scripts should be provided,
one using just ksh93 that illustrates the bug,
and the other using the /bin/ksh|ksh93s+ combination that does work
On Mon, 14 Apr 2008 13:10:30 -0600 (MDT) bugmail-sender at Sun.COM wrote:
> *Synopsis*: ksh93 causes parent shell to die when child shell is suspended
> === *Description* ============================================================
> I have a script that starts a sub-shell. When I suspend the subshell, to
> return
> to the login shell, often the login shell logs out. Sometimes the login shell
> hangs with a message saying I have stopped jobs, and then I can't type into
> the
> window. (In both cases I'm using rlogin or ssh to login.)
> /bin/ksh doesn't have this problem.
> If I use /bin/ksh for the login shell, and /bin/ksh93 for the sub-shell, the
> problem doesn't occur.
> This seems like another job control / tty modes race condition.
