[lkml]   [2000]   [Nov]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] list-style makefile boilerplate without reordering

[Christoph Hellwig <>]
> +
> +# include a local makefile, if present
> +-include Makefile.local


> +%.i: %.c
> + $(CPP) $(CFLAGS) $(EXTRA_CFLAGS) $(CFLAGS_$@) $< > $@

Why? Well, I know, it's historical, but does anyone think we actually
need this?

> +%.s: %.S
> + $(CPP) $(AFLAGS) $(EXTRA_AFLAGS) $(AFLAGS_$@) $< > $@


> +# Rule to compile a set of .o files into one .o file
> +$(O_TARGET): $(obj-y)
> + rm -f $@
> + ifneq "$(strip $(obj-y))" ""
> + $(LD) $(EXTRA_LDFLAGS) -r -o $@ $(filter $(obj-y), $^)
> + else
> + $(AR) rcs $@ $(filter $(obj-y), $^)
We have just determined that this is empty. I think it would be more
readable to say

$(AR) rcs $@

The whole line is a clever hack to produce something in the event of an
empty list, and this makes it more obvious what is going on.

> +# this make dependencies (not that ...) quickly
> +fastdep: dummy
> + $(TOPDIR)/scripts/mkdep $(wildcard *.[chS] local.h.master) > .depend
> +ifdef all-subdirs
> + $(MAKE) $(patsubst %,_sfdep_%,$(all-subdirs)) _FASTDEP_ALL_SUB_DIRS="$(all-subdirs)"
> +endif
> +
> +$(patsubst %,_sfdep_%,$(_FASTDEP_ALL_SUB_DIRS)):
> + $(MAKE) -C $(patsubst _sfdep_%,%,$@) fastdep
> +endif

This needs rethinking. I know this is how Rules.make does it, but
wouldn't it be faster to eliminate one level of recursion?

fastdep: dummy $(patsubst %,_sfdep_%,$(all-subdirs))
$(TOPDIR)/scripts/mkdep $(wildcard *.[chS] local.h.master) > .depend
$(patsubst %,_sfdep_%,$(all-subdirs)):
$(MAKE) -C $(patsubst _sfdep_%,%,$@) fastdep

Untested, but is there some reason that wouldn't work? Order of
execution shouldn't matter to mkdep, right?

> +# a rule to make subdirectories
> +sub_dirs: dummy $(patsubst %,_subdir_%,$(subdir-y))
> +
> +ifdef subdir-y
ifneq "$(strip $(subdir-y))" ""

> +PDWN=$(shell $(CONFIG_SHELL) $(TOPDIR)/scripts/

I think invoking is probably unnecessary, see below.

> +.PHONY: $(patsubst %,_modinst_%,$(subdir-m))
> +$(patsubst %,_modinst_%,$(subdir-m)) : dummy
> + $(MAKE) -C $(patsubst _modinst_%,%,$@) modules_install
> +endif

This is the only case where $(PDWN) is needed, so IMHO we should just
supply it internally -- faster than a shell process:

$(patsubst %,_modinst_%,$(subdir-m)): dummy
$(MAKE) -C $(patsubst _modinst_%,%,%@ modules_install \
PDWN=$(PDWN)$(patsubst _modinst_%,%,%@)/

Or, if that looks funny, we can initialize PDWN to '.' in
$(TOPDIR)/Makefile and use
...PDWN=$(PDWN)/$(patsubst _modinst_%,%,%@)

> +ifdef CONFIG_SMP

Eric Raymond won't like this. He at one point was pushing for phasing
out and eliminating 'ifdef CONFIG_*' from makefiles so he could define
CONFIG_* to be 'n' as distinguished from ''. I think it's a good idea.

ifneq "$(CONFIG_SMP)" ""

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 12:45    [W:0.031 / U:3.236 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site