Pid or not pid

Pid or not pid

[Peter William Stephenson]

Some of the code in execcmd() is getting executed in the parent rather
than the child, presumably since the last rearrangement... in
particular, nice(5) is getting called if fork() returned a non-zero
pid, i.e. in the parent (line 1350 of exec.c).  This means each time a
background process is forked, the parent shell has a lower priority.
This may be a good way to deal with process hogs, but I don't think
it's what's intended.

% ps -l -p$$
 F S   UID   PID  PPID  C PRI NI  P    SZ:RSS      WCHAN TTY      TIME COMD
30 S  2082 18544 21831  2  39 20  *   775:366   801bf9c0 pts/54   0:01 zsh
% sleep 3 &
[1] 19247
+4:52% ps -l -p$$
 F S   UID   PID  PPID  C PRI NI  P    SZ:RSS      WCHAN TTY      TIME COMD
30 S  2082 18544 21831  1  39 25  *   775:361   801bf9c0 pts/54   0:01 zsh
% 

-- 
Peter Stephenson <pws@ifh.de>       Tel: +49 33762 77366
WWW:  http://www.ifh.de/~pws/       Fax: +49 33762 77330
Deutches Electronen-Synchrotron --- Institut fuer Hochenergiephysik Zeuthen
DESY-IfH, 15735 Zeuthen, Germany.

Re: Pid or not pid

[Richard Coleman]

> Some of the code in execcmd() is getting executed in the parent rather
> than the child, presumably since the last rearrangement... in
> particular, nice(5) is getting called if fork() returned a non-zero
> pid, i.e. in the parent (line 1350 of exec.c).  This means each time a
> background process is forked, the parent shell has a lower priority.
> This may be a good way to deal with process hogs, but I don't think
> it's what's intended.

err...  No I didn't intend that.  But maybe we should leave it
as an option :-)

rc

Re: Pid or not pid

[Zefram]

>> Some of the code in execcmd() is getting executed in the parent rather
>> than the child, presumably since the last rearrangement... in
>> particular, nice(5) is getting called if fork() returned a non-zero
>> pid, i.e. in the parent (line 1350 of exec.c).  This means each time a
>> background process is forked, the parent shell has a lower priority.
>> This may be a good way to deal with process hogs, but I don't think
>> it's what's intended.
>
>err...  No I didn't intend that.  But maybe we should leave it
>as an option :-)

One has to wonder, though, whether this FG_NICE option should be on by
default.  Obviously, the maximally backward compatible choice would be
for it to be on.  (-:

-zefram

Re: Pid or not pid

[Richard Coleman]

> Some of the code in execcmd() is getting executed in the parent rather
> than the child, presumably since the last rearrangement... in
> particular, nice(5) is getting called if fork() returned a non-zero
> pid, i.e. in the parent (line 1350 of exec.c).  This means each time a
> background process is forked, the parent shell has a lower priority.
> This may be a good way to deal with process hogs, but I don't think
> it's what's intended.

This is for beta12-test2.  This bug is not present in beta10 or beta11.

rc


*** exec.c	1995/11/16 03:31:23	1.89
--- exec.c	1995/11/16 22:05:14
***************
*** 1345,1353 ****
  	    read(synch[0], &dummy, 1);
  	    close(synch[0]);
  	    if (how & Z_ASYNC) {
- 		/* Check if we should run this job at a lower priority */
- 		if (isset(BGNICE))
- 		    nice(5);
  		lastpid = pid;
  	    } else if (!jobtab[thisjob].stty_in_env && nonempty(cmd->vars)) {
  		/* search for STTY=... */
--- 1345,1350 ----
***************
*** 1367,1372 ****
--- 1364,1373 ----
  	forked = 1;
  	if (sigtrapped[SIGINT] == 2)
  	    holdintr();
+ 	/* Check if we should run background jobs at a lower priority. */
+ 	if ((how & Z_ASYNC) && isset(BGNICE))
+ 	    nice(5);
+ 
      } else if (is_cursh) {
  	/* This is a current shell procedure that didn't need to fork.    *
  	 * This includes current shell procedures that are being exec'ed, *