Re: [PATCH] [lambda] Extract lambda functions from semantics.c.

[Adam Butcher <adam at jessamine dot co dot uk>]

On 04.09.2013 03:41, Gabriel Dos Reis wrote:
> On Tue, Sep 3, 2013 at 9:33 PM, Mike Stump <mikestump@comcast.net> 
> wrote:
> > On Jul 12, 2013, at 11:18 PM, Adam Butcher <adam@jessamine.co.uk> 
> > wrote:
> > >       * gcc/cp/semantics.c (build_lambda_expr),
> > >       (build_lambda_object), (begin_lambda_type), 
> > > (lambda_return_type),
> > >       (lambda_function), (lambda_capture_field_type), 
> > > (is_capture_proxy),
> > >       (is_normal_capture_proxy), (insert_capture_proxy),
> > >       (insert_pending_capture_proxies), (lambda_proxy_type),
> > >       (build_capture_proxy), (vla_capture_type),
> > >       (register_capture_members), (add_default_capture),
> > >       (lambda_expr_this_capture), (maybe_resolve_dummy),
> > >       (nonlambda_method_basetype), (maybe_add_lambda_conv_op) and
> > >       (is_lambda_ignored_entity): Moved definitions into ...
> > >       * gcc/cp/lambda.c: ... this new file.
> >
> >
> > This can cause an incremental build failure because there are no 
> > dependencies:
> >
> > diff --git a/gcc/cp/Make-lang.in b/gcc/cp/Make-lang.in
> > index 2c1774f..65dfe08 100644
> > --- a/gcc/cp/Make-lang.in
> > +++ b/gcc/cp/Make-lang.in
> > @@ -351,6 +351,7 @@ cp/vtable-class-hierarchy.o: 
> > cp/vtable-class-hierarchy.c \
> >  cp/name-lookup.o: cp/name-lookup.c $(CONFIG_H) $(SYSTEM_H) 
> > coretypes.h \
> >         $(TM_H) $(CXX_TREE_H) $(TIMEVAR_H) gt-cp-name-lookup.h 
> > $(PARAMS_H) \
> >         $(DIAGNOSTIC_CORE_H) $(FLAGS_H) debug.h pointer-set.h
> > +cp/lambda.o: cp/lambda.c $(CXX_TREE_H) $(CGRAPH_H) $(VEC_H) 
> > $(SYSTEM_H) coretypes.h
> >
> >  cp/cxx-pretty-print.o: cp/cxx-pretty-print.c $(CXX_PRETTY_PRINT_H) 
> > \
> >    $(CONFIG_H) $(SYSTEM_H) $(TM_H) coretypes.h $(CXX_TREE_H) 
> > tree-pretty-print.h
> >
> > When tree codes are added or moved, the check is then against the 
> > wrong number, and this will kill the build.
> >
> > I'm still looking forward to the day when all the dependancies are 
> > unceremoniously ripped out, until then...
> >
> > Ok?
>
> OK.
Eek.  I didn't realize dependencies had to be manually specified.  
That's prompted me to update a more recent patchset I'm working on where 
I've introduced a new header.

Is anyone working on using some use, perhaps filtered, of -MD (or -MDD) 
to generate deps on the fly?  I haven't looked into the GCC makefile 
system in any detail but I assume dependency handling is more complex 
than the standard usage pattern for -MD or I guess someone would have 
done it already.  Or are maintainers worried that auto deps will slow 
the build down too much?  In our team's experience with using -MD the 
overhead is negligible; especially when weighed up against the effort 
required to manually maintain deps.  It just takes make a little longer 
to start actually building anything.