On Tue, 2008-12-30 at 13:24 +0000, Colin Watson wrote:
> On Mon, Dec 29, 2008 at 11:51:26AM -0700, Tim Gardner wrote:
> > Scott James Remnant wrote:
> > > On Mon, 2008-12-29 at 11:05 -0700, Tim Gardner wrote:
> > >> It appears that changing the kernel config thusly requires
> > >> 'pty.legacy_count=N' as a kernel boot parameter for alternate and server
> > >> installs.
> > >
> > > Why do servers need that at all?!
> "Alternate and server installs", i.e. d-i. There's nothing
> server-specific here.
Right, I figured d-i and was wondering what wasn't using /dev/pts
because I was sure it did <g>
> > > Do you have a bug reference, or has somebody misunderstood the options
> > > we're changing here?
> > I've not filed a bug report on it yet since I just noticed the problem.
> > About the time /sbin/init is launched, the screen clears and
> > continuously prints "can't get a pty: no such device or file exists" (or
> > something close to that).
> This message comes from bterm, the UTF-8 framebuffer terminal used in
> d-i. bterm does use Unix98 ptys if available (I patched it to do this
> way back in 2004). However, /dev/pts mistakenly wasn't getting mounted,
> and therefore it had to fall back to legacy ptys. It looks like it's
> been doing this silently at least back to hardy, so I guess we should be
> grateful for this change making it obvious! :-)
Great! So we can leave legacy ptys at zero! o/ -- if
anything else comes up, we can fix those too
kernel-team mailing list