waitpid() with WUNTRACED flag? (was Re: Hang on ctrl+Z after the MPSAFE tsleep/wakeup commit)

[YONETANI Tomokazu]

On Fri, Dec 02, 2005 at 10:08:59AM -0800, Matthew Dillon wrote:
> :Hello.
> :After the following commit, stopping a user process with SIGSTOP
> :does not give control to the shell:
> :http://leaf.dragonflybsd.org/mailarchive/commits/2005-11/msg00109.html
> :It only happens with a program which does fork() inside it.  Here's
> :a simple program to demonstrate this.
> :
> :$ gcc -g -Wall a.c && ./a.out cat
> :(press ctrl+Z here, and it accepts no other signals until you send
> :SIGCONT to the child process from another screen)
> :
> :`ps' command shows that the stuck processes are marked as "TL+"
> :(meaning that they didn't give up even after having been stopped?).
> 
>     Ah ha!  I have reproduced it.  I was running './a.out cat' separately,
>     from csh, while you were running it from /bin/sh.
> 
>     When I changed over to /bin/sh the problem reproduced.
> 
>     I'll get it fixed today, hopefully soon.

While this has been fixed, I realized that I overlooked another
similar problem; try running vipw, suspend it with ctrl-Z and
continue with `fg'; the shell immediately reports that vipw has been
suspended(processes are marked as `TL' by ps command); another `fg'
seems to continue it, but vi(or whatever editor invoked by vipw)
doesn't handle the screen properly, even though it's responding to
the keystrokes.
Attached is a small code fragment that demonstrates the problem.
BTW, it behaves differently on -CURRENT and 1.2.6-RELEASE;
on -CURRENT, waitpid() returns after `fg' command followed by a ctrl-Z,
while on 1.2.6-RELEASE it never returns when suspended or continued.
Without WUNTRACED flag, the code seems to work the same way on both
versions of the OS.
Too bad I couldn't find it before the new release.