Subject: Re: CVS commit: src/sys/arch/i386/include
To: Christos Zoulas <christos@zoulas.com>
From: Matthias Drochner <M.Drochner@fz-juelich.de>
List: source-changes
Date: 09/10/2003 20:42:40
This is a multipart MIME message.

--==_Exmh_17012332204800
Content-Type: text/plain; charset=us-ascii


christos@zoulas.com said:
> No, si_code is reserved for MI stuff so we cannot use this. Read the
> new manpage for siginfo... 

So wouldn't the the appended patch make signals to sa programs work better?
(This is related to a test case I've sent to Nathan a while ago, where traps
with code=0 did confuse libpthread's signal delivery code.)

best regards
Matthias



--==_Exmh_17012332204800
Content-Type: text/plain ; name="sig.txt"; charset=us-ascii
Content-Description: sig.txt
Content-Disposition: attachment; filename="sig.txt"

Index: kern/kern_sig.c
===================================================================
RCS file: /cvsroot/src/sys/kern/kern_sig.c,v
retrieving revision 1.149
diff -u -p -r1.149 kern_sig.c
--- kern/kern_sig.c	2003/09/10 16:41:26	1.149
+++ kern/kern_sig.c	2003/09/10 18:28:18
@@ -1168,7 +1168,7 @@ kpsendsig(struct lwp *l, ksiginfo_t *ksi
 		si = pool_get(&siginfo_pool, PR_WAITOK);
 		si->_info = *ksi;
 		le = li = NULL;
-		if (ksi->ksi_trap)
+		if (ksi->ksi_code > 0)
 			le = l;
 		else
 			li = l;

--==_Exmh_17012332204800--