[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

re: last chance to do something about discover1 etc: I give up



David Nusinow wrote:
>    I didn't realize that this was a pressing issue.
> Since it worked for sarge, I figured that the kludge
> could remain until we can get a proper solution in
> place.

There have been large changes to the installer since sarge that broke the
old solution.

>    My plan for some time now has been to fix this in
> the future. Not in the far future though, it's my
> major release goal for etch+1. A great deal of the
> necessary work is in development upstream to make this
> happen. Daniel Stone is working on input device
> hotplugging and Intel and Redhat both are working on
> monitor autoconfiguration. I plan to write the
> necessary code to remove the dependency on discover if
> it's not done before I get to it. I also plan to move
> the xresprobe logic in to the server as necessary.
> These changes are in the works and will be ready for
> etch+1. I simply can not get them to you for etch
> though.

Monitor autoconfiguration, device hotplugging, World Peace, and what
have you are not actually necessary to fix this. All it takes is a
refactoring of X's code, moving the hardware detection and setup code
from the config script to the postinst.

For example, a brutal and yet valid way to do it is to change the
postinst:

#!/bin/sh
. /usr/share/debconf/confmodule
. /usr/share/X/what-was-my-config-script || true
...

This *will* work, because that's what debconf degrades to doing if
preconfiguration doesn't happen: the postinst first runs the config script.

There are of course cleaner approaches, like integrating the config
script into the postinst script, that would be a bit more work, but work
on the order of hours or days, not years.

-- 
see shy jo

Attachment: signature.asc
Description: Digital signature


Reply to: