Date: Sun, 5 Jan 2003 15:18:57 -0800 (PST) From: Nate Lawson <nate@root.org> To: Bruce Evans <bde@zeta.org.au> Cc: Jake Burkholder <jake@FreeBSD.org>, cvs-all@FreeBSD.org, cvs-committers@FreeBSD.org Subject: Re: cvs commit: src/sys/dev/sab sab.c Message-ID: <Pine.BSF.4.21.0301051514140.8587-100000@root.org> In-Reply-To: <20030106075621.M616-100000@gamplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 6 Jan 2003, Bruce Evans wrote: > On Sun, 5 Jan 2003, Nate Lawson wrote: > > I've always wondered if you were supposed to deliver the char before > > calling breakpoint(). Looks like you do this but it seemed like sio > > didn't. What's the desired behavior? > > The escape sequence for the breakpoint shouldn't be delivered at all. > sio seems to get this right for the final character in the sequence > (using "goto cont;") but not for the previous characters (it sets the > state variables for them but falls through to handle them normally). > If the final character is delivered, then whether it is delivered > before or after calling breakpoint() doesn't matter much. Only the > internals of the interrupt handler and the state machine could > normally tell the difference. So we should buffer the chars if they are matching state and deliver them if the state machine fails to detect a match (probably by looping calls to the consumer func)? This makes a bigger difference with the gdb frame detector (up to about 32 chars max, in practice only about 8 chars). It also probably makes sense to provide the chars in the gdb case since they are important to gdb. Right now, my patch depends on gdb retransmitting the initial command. -Nate To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0301051514140.8587-100000>