CONFIG_INPUT problems

CONFIG_INPUT problems

[Martin J. Bligh]

So we seem to have a problem with the way the config stuff deals
with CONFIG_INPUT, and I'd like some advice on the best way to
fix it.

The problem:

CONFIG_INPUT was commonly set to "n" in 2.4 kernels (and hence the
configs that upgrading people pick up). Even though it defaults to
"y" under the 2.5 subsystem, the old config overrides it, which I
guess is correct for most things, but it didn't govern keyboard
and (by cascade) CONFIG_VT_CONSOLE under 2.4 (I think).

The result:

Lots of people just get a really screwed up 2.5 kernel that does
nothing useful for them.

What to do ...

I thought about inverting the logic, and creating CONFIG_HEADLESS
which is !CONFIG_INPUT basically ... but changing all the stuff
depending on CONFIG_INPUT is rather invasive. So I was thinking of
something like

CONFIG_HEADLESS
	bool "headless console support
	default "n"

if HEADLESS = y
	define_bool CONFIG_INPUT = n
else
	define_bool CONFIG_INPUT = y
endif

or something vaguely along those lines ... except there doesn't
seem to be a way I can see to force a config option on from the
new config system? So that's actually a more general question, I guess ;-)

However, if someone has a better idea on how to not shoot the poor 
users over CONFIG_INPUT / CONFIG_VT_CONSOLE, I'd be very interested ...

M.

Re: CONFIG_INPUT problems

[Roman Zippel]

Hi,

On Wed, 9 Apr 2003, Martin J. Bligh wrote:

> What to do ...
> 
> I thought about inverting the logic, and creating CONFIG_HEADLESS
> which is !CONFIG_INPUT basically ... but changing all the stuff
> depending on CONFIG_INPUT is rather invasive. So I was thinking of
> something like
> 
> CONFIG_HEADLESS
> 	bool "headless console support
> 	default "n"
> 
> if HEADLESS = y
> 	define_bool CONFIG_INPUT = n
> else
> 	define_bool CONFIG_INPUT = y
> endif

config INPUT
	default y if !HEADLESS

> or something vaguely along those lines ... except there doesn't
> seem to be a way I can see to force a config option on from the
> new config system? So that's actually a more general question, I guess ;-)

I know, this has been requested a few times, I hope to have something 
soon. (It's more a time problem.)

bye, Roman

Re: CONFIG_INPUT problems

[Martin J. Bligh]

>> I thought about inverting the logic, and creating CONFIG_HEADLESS
>> which is !CONFIG_INPUT basically ... but changing all the stuff
>> depending on CONFIG_INPUT is rather invasive. So I was thinking of
>> something like
>> 
>> CONFIG_HEADLESS
>> 	bool "headless console support
>> 	default "n"
>> 
>> if HEADLESS = y
>> 	define_bool CONFIG_INPUT = n
>> else
>> 	define_bool CONFIG_INPUT = y
>> endif
> 
> config INPUT
> 	default y if !HEADLESS

I don't see how that'll work ... we already have it defaulting to y,
but there's a previous setting that's 'n' from the 2.4 config file
they're upgrading from ... and that overrides the default, right?

I think in general, that's the right behaviour (else you'd override
your previous settings every time). It's just that in this case, it
creates a serious user problem that we really need to fix. Personally,
I'd be happy to just force it to on all the time, but I suspect that'd
get me lynched ;-)

>> or something vaguely along those lines ... except there doesn't
>> seem to be a way I can see to force a config option on from the
>> new config system? So that's actually a more general question, I guess ;-)
> 
> I know, this has been requested a few times, I hope to have something 
> soon. (It's more a time problem.)

OK, thanks.

M.

Re: CONFIG_INPUT problems

[Martin J. Bligh]

>> > config INPUT
>> > 	default y if !HEADLESS
>> 
>> I don't see how that'll work ... we already have it defaulting to y,
>> but there's a previous setting that's 'n' from the 2.4 config file
>> they're upgrading from ... and that overrides the default, right?
> 
> A default without a visible prompt works like derived variable.
> If there is prompt, the .config value and the default value is used as 
> default input for the prompt.

Aha! that's perfect - thanks.

M.

Re: CONFIG_INPUT problems

[Roman Zippel]

Hi,

On Wed, 9 Apr 2003, Martin J. Bligh wrote:

> > config INPUT
> > 	default y if !HEADLESS
> 
> I don't see how that'll work ... we already have it defaulting to y,
> but there's a previous setting that's 'n' from the 2.4 config file
> they're upgrading from ... and that overrides the default, right?

A default without a visible prompt works like derived variable.
If there is prompt, the .config value and the default value is used as 
default input for the prompt.

bye, Roman