Re: PR/52117 CVS commit: src/tests/kernel

[Martin Husemann <martin%duskware.de@localhost>]

The following reply was made to PR kern/52117; it has been noted by GNATS.

From: Martin Husemann <martin%duskware.de@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: PR/52117 CVS commit: src/tests/kernel
Date: Wed, 29 Mar 2017 06:52:21 +0200

 On Tue, Mar 28, 2017 at 09:40:00PM +0000, Kamil Rytarowski wrote:
 >  For consistency I would add "ksi.ksi_signo = TRAP_CHLD;". My motivation
 >  is to attribute every SIGTRAP in the kernel with appropriate si_code,
 >  even if it works for some reasons in some scenarios without them.
 
 Yes, that certainly should be done.
 
 >  From user-space point of view we want a deterministic order of receiving
 >  signals by a debugger. Signal for "child" and next "child2".
 
 We can not guarantee this. The signals happen as soon as the process
 gets scheduled, it would be pretty hard to guarantee the forker running
 first (at least I wouldn't know how to do that).
 
 Is this realy a requirement from a debuggers perspective?
 
 >  I prefer the <sys/childreturn.h> version as it has bit less overhead...
 >  but it might be just a microoptimization not worth modifying the
 >  child_return code in all ports.
 
 Yes, the code is a bit clunky, but runtime performance and locking is actually
 better (I think).
 
 Martin