Re: for when dragonflyBSD next release?

[Chris Pressey <cpressey@xxxxxxxxxxxxxxx>]

On Fri, 18 Mar 2005 13:22:32 -0800 (PST)
Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx> wrote:

> :Followup: here is a pre-rolled ISO containing this installer:
> :
> :http://cvs.bsdinstaller.org/iso-images/dfly-2005.0317.iso.bz2
> :
> :made with 'make installer_release' from yesterday's sources plus the
> :patch in my previous message.
> :
> :-Chris
> 
>     Ok, here's some feedback, logging in as 'installer':
> 
>     * ESC doesn't work as expected.  When I hit ESC, nothing happens
>       until I hit another key (like return), and then the installer
>       executes both actions.  This leads to multiple actions being
>       unexpectedly executed.

OK, see below.

>     * ESC[+some_other_key] exits the top level menu.   It definitely
>     should not do that.

OK.

>     * A large number of directories, including /root, were created
>       with user id 1000 instead of 0.  That may have been a side
>       effect of the way you built the ISO because they are uid 1000 on
>       the ISO too?

Hmm, interesting - I did run 'make installer_release' as root, which I
assume should do everything (build+install world+kernel) as root. 
Possibly I didn't clean up enough before running it...

>     Other then that, a basic fresh installion appears to work.
> 
>     If you are trying to detect cursor escape sequences, which are
>     typically ESC+'['+<blah>, the only way to detect a lone 'ESC' is
>     to have a fairly fast timeout looking for the additional ANSI esp
>     sequence characters after you have received the ESC character
>     before you give up and just assume a single ESC key was hit.

I'm using what [n]curses provides for detecting an ESC keypress, which
should work like you describe.

But ncurses' timeout may be longer than you were expecting.  Have you
tried pressing 'ESC' then waiting about a full second?  It works for me.

There may be a way to tell ncurses to use a shorter timeout, but I don't
know what it is - I'll look into it.

(Beyond that - I mentioned this briefly to Sascha - there's no hard and
fast need that I can see for the "terminal" that is being emulated by
the system console to resemble a VT100 - i.e. it could use some
character like NUL to signal the start of an escape sequence, and the
termcap entry for syscons could be adjusted accordingly.  But that's
definately experimental future work, if it's even possible.)

Thanks,
-Chris