Re: Who's at fault here?

[Nicolas Williams <Nicolas.Williams%sun.com@localhost>]

On Thu, Dec 01, 2005 at 10:27:36AM -0500, der Mouse wrote:
> >>> Wait... did you say "window-change" ?
> >> I can't attribute that to anything but a mistake by Nicolas.  [...]
> > There are in cases I've seen that are similar to what you described,
> > only involving window-change instead of window adjusts
> 
> Aha.  My apologies for thinking it was a mistake you made, then.  Given
> what I saw of the OpenSSH code, though, it's not surprising, since it
> doesn't seem to consider session channels "real" until the
> shell/exec/etc request has come in - I would expect a lot of things to
> fail when applied to such channels.  (I didn't dig into it enough to
> know exactly what things.)

Well, recent versions of the OpenSSH server do the right thing, IIRC,
older versions didn't.  The client still sends window-change requests
too soon for broken servers, but the client could be smarter about this.

Hmmm, delaying a window-change request until after the shell/exec/
subsystem request might lead to the app starting to draw with the wrong
window size -- not pretty, but the app will get the size change
notification eventually.  Not delaying it can cause broken servers to
disconnect -- worse than not pretty.

A job for the compat db?  Sigh.

> > (oof, confusing terms).
> 
> Quite.  Too many "window"s floating around.

Yes.  Wish that had been "terminal-size-change" -- but it is what it is.

Nico
--