From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <cc6860a6e57192e3c7adc96089b7f540@gmx.de>
To: 9fans@9fans.net
Date: Sun,  5 Oct 2008 03:56:07 +0200
From: cinap_lenrek@gmx.de
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: [9fans] APE, noted(), NSAVE/NRSTR
Topicbox-Message-UUID: 1718ea0e-ead4-11e9-9d60-3106f5b1d025

To emulate unix signals (that can nest) with notes (that cant nest),
ape uses the special noted() parameters NSAVE and NRSTR.

NCONT/NRSTR set up->ureg to the "old ureg":
/sys/src/9/pc/trap:^noted
		up->ureg = (Ureg*)(*(ulong*)(oureg-BY2WD));

NSAVE accepts the note as handled, but unlike NCONT/NRSTR keeps
the up->ureg in place (to be later restored by NRSTR).

but...

/sys/src/9/pc/trap.c:^notify

[1]	up->ureg = (void*)sp;
	memmove((Ureg*)sp, ureg, sizeof(Ureg));
[2]	*(Ureg**)(sp-BY2WD) = up->ureg;	/* word under Ureg is old up->ureg */
[3]	up->ureg = (void*)sp;
	sp -= BY2WD+ERRMAX;
	memmove((char*)sp, up->note[0].msg, ERRMAX);

Notify always overrides up->ureg in [1]. So what effect should
(not) restoring of up->ureg in noted() have? Why is up->ureg set
again in [3]?

What happens when a note is posted between a noted(NSAVE) /
noted(NRSTR)?

--
cinap