On Mon, Dec 19, 2011 at 05:32:21PM -0500, Kyle McMartin wrote:
On Mon, Dec 19, 2011 at 05:25:31PM -0500, Dave Jones wrote:
> Turns out we were getting lucky anyway, as it depended on CONFIG_ATALK, which
> was modular.
>
Ah, that's a relief. I have an idea for how we can monitor this sort of
thing and try to prevent it from occuring in the future... I'll see what
I can do in January.
We can almost do it automatically.
scripts/sort-config has a mode to check for options changing state
from what we set them. But it gets confused by the way we do overrides and
thus spews a bunch of false positives.
so if we did a pass through the generated file post merge.pl, but pre oldconfig,
to weed out the duplicates (just keep the last version of every symbol),
I think we could just pass the result to sort-config -m to find out what's
getting changed.
it still won't catch the 'bool became tristate' in all cases, but if
there's
a dependant symbol as there was in this case, it should get flagged.
Dave