make allyesconfig: logical failure?

[Rob Landley]

On Tuesday 06 September 2005 02:51, Vladimir N. Oleynik wrote:
> Rob,
>
> > On Monday 05 September 2005 05:36, Dirk Clemens wrote:
> >>make allyesconfig will select networking/"Support using httpd only from
> >>inetd".
> >>But this is a negatived (corret word?) choice:
> >>  if selected it will disable the standalone httpd.
>
> And
> CONFIG_FEATURE_TELNETD_INETD

Should be pretty easy to add symbols to the sed list, especially the 
consolidated version.  Just |NEW_SYMBOL before the closing parentheses.

> > In theory "allyesconfig" is mostly for testing, and also for switching on
> > everything as a base config it's then easier to switch stuff off from.
>
> I think, after allyesconfig result minimal summary size, but switch on all
> features.

Possibly we need a "make everything".  There's a difference between "exercise 
all possible code paths" and "give me all the standard features that don't 
require special setup".

Or possibly just have a "make defconfig" that's what we want "make 
allyesconfig" to be?

> > + sed -i -r -e
> > "s/^(CONFIG_DEBUG|CONFIG_USING_CROSS_COMPILER|CONFIG_STATIC|
> > CONFIG_SELINUX|CONFIG_FEATURE_HTTPD_USAGE_FROM_INETD_ONLY).*/# \1 is not
>
> But why switch off CONFIG_STATIC?

Same reason to switch off CONFIG_DEBUG, I suppose.  It doesn't actually 
exercise any additional code path in our build.

Now switching off CONFIG_SELINUX is the one that I find curious...

I dunno, sounds like what we're looking for is a "make allfeatures"?  Right 
now "make allnoconfig" does exactly what you suspect, but "make allyesconfig" 
has at least one corner case...

And "make defconfig" seems almost like "make random", really.  Who decides tar 
needs -j but not -X?  Isn't this policy?

Rob