Re: [kbuild-devel] Re: [patch] config language dep_* enhancements

[Greg Banks <gnb@alphalink.com.au>]

Peter Samuelson wrote:
> 
> >
> >    [...]CONFIG_BLK_DEV_IDESCSI[...]
> >
> That one example is more than enough to convince me *not* to try
> changing the "ignore empty deps" rule for 2.4.  2.5 is a different
> matter, though..

Ah, good.

> This is like the Stanford checker stuff.  These are bugs.  You have
> the means to find them automatically, but not the time or desire to
> fix them.  

Actually I have got the desire to fix them, what I lack is the ability
to get patches into the kernel that are too non-trivial to go through Rusty.

> Post a list and perhaps others will pitch in.  Something
> like
> 
> drivers/ide/Config.in:17: In dep_tristate CONFIG_BLK_DEV_IDESCSI:
> drivers/ide/Config.in:17: CONFIG_SCSI not defined for i386, ppc, ...

Ok, but perhaps it's not clear how many problems there are.  The full
log file from a gcml2 run on 2.5.29 is 573 KiB.  Here's a summary:

977    missing-experimental-tag
113    spurious-experimental-tag
145    variant-experimental-tag
30     inconsistent-experimental-tag
13     missing-obsolete-tag
41     spurious-obsolete-tag
25     variant-obsolete-tag
5      spurious-dependencies
187    nonliteral-define
28     unset-statement		(ignore these)
25     different-banner
61     different-parent
36     overlapping-definitions
1      primitive-in-root
310    undeclared-symbol
73     forward-compared-to-n
287    forward-reference
1093   forward-dependancy
1      symbol-like-literal
103    constant-symbol-dependency
8      different-compound-type
3562   total

These numbers are aggregates over 17 arches, so you need to divide by
a number between 1 and 17.  Also some of these have been fixed in 2.5.30.
I can post the full list if people want, but it would be a bit overwhelming.

> In this case the fix would be to move CONFIG_BLK_DEV_IDESCSI to the
> SCSI subsystem, where in my opinion it belonged anyway.

Ok then.

> This would break down if there are any actual cycles - things which
> can't logically be moved to a place after the definitions of the
> facilities on which they depend.

I'm not able to detect anything like that.

> That we have to worry about this at all is an artifact of using a
> procedural langauge, rather than a declarative language like Prolog or
> CML2. 

Actually config language isn't really procedural, pseudo-procedural would
be more like it.

> I *really* don't want to go *there*, though. (:

Yep.

> > And "" != "n" in other contexts, like if [];then statements.
> 
> That is true.  But that should not affect the dep_* logic, should it?

Correct, I'm just pointing out that using orthogonality arguments with
the config language is not going to get you anywhere useful.

> The point here is that people who aren't hip-deep in config language
> code don't think about them being separate.  Ergo bugs.

Agreed.

> I've been thinking hard about a new sort of 'if' statement that
> doesn't look like a test command.  

Interesting, I'd like to hear more.

My favourite idea is a different form of choice which worked more
like a menu, so you could intersperse conditionals with the choice
items.  This would help with the several places in CML1 were the
same choice is mostly-duplicated in different conditionals, e.g.
'Kernel page size' in arch/ia64/config.in.  ESR worked out a sensible
set of semantics for how this should work.

> (Yes, it's my mission to eventually
> get rid of the '$'s in config language.  I think it can be done,
> piecemeal, over a long period of time.  

Sounds good to me.

> This is if we don't end up
> adopting a whole new language like CML2 or Roman's stuff.)

Or a new parser & frontends for the existing language.

I'm not convinced that a complete new language will ever succeed after
the CML2 debacle, machine translated or not.  This is why I gave up the
idea of automatically converting CML1 to CML2.

Greg.
-- 
the price of civilisation today is a courageous willingness to prevail,
with force, if necessary, against whatever vicious and uncomprehending
enemies try to strike it down.     - Roger Sandall, The Age, 28Sep2001.