From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <3c88b8bfb5216d40c70392c084358983@coraid.com>
From: erik quanstrom <quanstro@coraid.com>
Date: Tue, 19 Dec 2006 22:55:09 -0500
To: rsc@swtch.com, quanstro@coraid.com, 9fans@cse.psu.edu
Subject: Re: [9fans] wait hang
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Cc: 
Topicbox-Message-UUID: f6b80206-ead1-11e9-9d60-3106f5b1d025

there's a version of rc on sources (/n/sources/contrib/quanstro/rc-wait.tbz)
that keeps careful track of waits so "wait $pid" should be dependable,
at least in the basic approach, i did minimal minimal testing of the details.
there is somewhat verbose information printed along these lines:

	; sleep & sleep & sleep & sleep
	waitqadd(772665)
	waitqadd(772666)
	waitqadd(772667)
	waitqadd(772668)
	waitfor 772668
	  w->pid 772668
	waitqrm(772668)
	Xreturn i=0; p 205256, i=1
	broken! wait 772665
	waitfor 772665
	  w->pid 772667
	  w->pid 772666
	  w->pid 772665
	waitqrm(772665)
	Xreturn i=0; p 205256, i=1
	broken! wait 772666
	waitfor 772666
	Xreturn i=0; p 205256, i=1
	broken! 

this output can be supressed by redefining the wdebug macro in exec.h.

the previous case of
	; sleep $longtime &
	; echo&echo&echo&
	; wait $apid

should illustrate the difference.

- erik