Bug or feature?

Bug or feature?

[Ricky Beam]

In my dangerous mode of "if it's presented to me, compile it", I've noticed
alot of junk being presented as a buildable option under the wrong arch.

For example, the Sparc64 and ARM MTD Map drivers are selectable from an
x86 build.  Some might say this is a shell bug, but at any rate, dep_*
don't function correctly (read: as one would expect) in the cases where
the deps are not defined.

For:
  dep_tristate '  Sun Microsystems userflash support' CONFIG_MTD_SUN_UFLASH $CONFIG_SPARC64
$CONFIG_SPARC64 is null and this doesn't appear to the shell function as an
arg.  Thus, it's presented as a selectable (tho' not compilable) option.

The same is visable for CONFIG_MTD_SA1100 and CONFIG_MTD_DC21285 (ARM).

Options:
 1) Don't select things that aren't in your machine/arch.
    (Translation: "Live with it.")
 2) Quote all the options.
    (ewww.)
 3) Fix the function(s).

--Ricky

Re: Bug or feature?

[Alan Cox]

>   dep_tristate '  Sun Microsystems userflash support' CONFIG_MTD_SUN_UFLASH $CONFIG_SPARC64
> $CONFIG_SPARC64 is null and this doesn't appear to the shell function as an
> arg.  Thus, it's presented as a selectable (tho' not compilable) option.

In -ac these should have been fixed for a while. I'm waiting for David
Woodhouse to send me his official versions of these fixes

Re: Bug or feature?

[Keith Owens]

On Mon, 27 Aug 2001 18:10:05 -0400 (EDT), 
Ricky Beam <jfbeam@bluetopia.net> wrote:
>  dep_tristate '  Sun Microsystems userflash support' CONFIG_MTD_SUN_UFLASH $CONFIG_SPARC64
>$CONFIG_SPARC64 is null and this doesn't appear to the shell function as an
>arg.  Thus, it's presented as a selectable (tho' not compilable) option.
>
>The same is visable for CONFIG_MTD_SA1100 and CONFIG_MTD_DC21285 (ARM).
>
>Options:
> 1) Don't select things that aren't in your machine/arch.
>    (Translation: "Live with it.")
> 2) Quote all the options.
>    (ewww.)
> 3) Fix the function(s).

None of the above.  Null config variables are treated as "don't care"
when the user really meant "only allow if this option is 'y'".  It is a
generic config problem with arch variables and I have a fix somewhere
in my backlog.  I will try to dig it out and send to AC.