background processes

background processes

[Tore Morkemo]

Hi !

If I start a backgroud process in rc, it seems to loose it's information
about the controlling tty. The result of this is that the background
process doesn't appear when I do a simple ps command, since ps by default
shows processes associated with the current terminal.

I'm kinda used to doing a ps to check whether my make's have finished yet.
Now I have to do ps -u .... It's not a big problem but is this a bug or
was it made like this ?

Why ?



I've compiled rc with NOEXECVE and EXECVE=execve.o, and
running SCO (I know....don't blame me)


Tore.

+----------------------------------------------------------------------+
!   Tore Morkemo, Bibliotek-Systemer A/S, N-3250 Larvik, Norway        !
!   uucp: ...!bibsyst.no!tore        Tel: +47 34 82 202                !
!            tore@bibsyst.no         Fax: +47 34 85 185                !
+----------------------------------------------------------------------+

Re: background processes

[Byron Rakitzis]

Trapping SIGCHLD is the right thing to do.

I'll get around to this eventually, but I would like to incorporate
it with whatever other changes to the signal handler that need to
be done.

background processes

[Donn Cave]

I have noticed that if rc only executes background commands (i.e. with
a trailing ampersand), it seems to accumulate large numbers of "defunct"
children.

Apparently the wait() call is only issued when waiting for a specific
process to complete - so if rc executes a trivial command in the foreground,
all the zombies finally get to die.

Would it be sensible to catch SIGCHLD for this - or would that invite
trouble with the signal-handling problems discussed last week?  (I confess
I didn't really understand that thread.)  Would it be better to just
wait() once for each prompt?

	Donn Cave, University Computing Services, University of Washington
	donn@cac.washington.edu