The samples/ subdirectory contains only modules.

But the only make run done there is in commands for vmlinux.

I can't see why this was ever done in this nonstandard fashion.

As things stand, the modules don't get built by 'make modules'.
I didn't make the addition of the directory use core-$(CONFIG_SAMPLES)

because there is no other conditional like that in the top-level Makefile

and samples/Makefile already uses obj-$(CONFIG_SAMPLES) as if it expects

always to be included.
Signed-off-by: Roland McGrath <roland@redhat.com>

---

 Makefile |    5 +----

 1 files changed, 1 insertions(+), 4 deletions(-)
diff --git a/Makefile b/Makefile

index c162370..9e9ce33 100644

--- a/Makefile

+++ b/Makefile

@@ -602,7 +602,7 @@ export mod_strip_cmd
 ifeq ($(KBUILD_EXTMOD),)

-core-y		+= kernel/ mm/ fs/ ipc/ security/ crypto/ block/

+core-y		+= kernel/ mm/ fs/ ipc/ security/ crypto/ block/ samples/
 vmlinux-dirs	:= $(patsubst %/,%,$(filter %/, $(init-y) $(init-m) \

 		     $(core-y) $(core-m) $(drivers-y) $(drivers-m) \

@@ -802,9 +802,6 @@ vmlinux: $(vmlinux-lds) $(vmlinux-init) $(vmlinux-main) vmlinux.o $(kallsyms.o)

 ifdef CONFIG_HEADERS_CHECK

 	$(Q)$(MAKE) -f $(srctree)/Makefile headers_check

 endif

-ifdef CONFIG_SAMPLES

-	$(Q)$(MAKE) $(build)=samples

-endif

 	$(call vmlinux-modpost)

 	$(call if_changed_rule,vmlinux__)

 	$(Q)rm -f .old_version

--

Sam, is this ok with the samples intent ? I think as long as we do not

include them with the kernel image and have a "make samples" to build

them, it's ok. Having them built upon make modules seems like a good

idea to me.
--

Mathieu Desnoyers

Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal

OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68

--

The samples code are supposed to be what the name says 'samples'.

This is not code that are supposed to be part of the built-in kernel.

This is not modules that are supposed to be installes when

installing modules.
Adding it to core-y as Roland does in the patch below is plain

wrong as it enabled both points above.

The fact that the present code in samples/ does not do this is

in this respect irellevant.
Do we have problems when to build the sampls - then lets

address this issue but not by trying to upgrade the samples

to first class citizen in the kernel - they are not that

and should not be handled like that.
	Sam

--

Then is there some other way to have the samples built upon "make

modules" that would not install them with other modules ?
--

Mathieu Desnoyers

Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal

OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68

--

The simple solution is like attached.

I do not care if they are build as part of vmlinux or modules build.

So if it is generally preferred we could move it like this.
Comments?
	Sam
diff --git a/Makefile b/Makefile

index c162370..c72afa9 100644

--- a/Makefile

+++ b/Makefile

@@ -802,9 +802,6 @@ vmlinux: $(vmlinux-lds) $(vmlinux-init) $(vmlinux-main) vmlinux.o $(kallsyms.o)

 ifdef CONFIG_HEADERS_CHECK

 	$(Q)$(MAKE) -f $(srctree)/Makefile headers_check

 endif

-ifdef CONFIG_SAMPLES

-	$(Q)$(MAKE) $(build)=samples

-endif

 	$(call vmlinux-modpost)

 	$(call if_changed_rule,vmlinux__)

 	$(Q)rm -f .old_version

@@ -1036,6 +1033,9 @@ all: modules

 PHONY += modules

 modules: $(vmlinux-dirs) $(if $(KBUILD_BUILTIN),vmlinux)

 	$(Q)$(AWK) '!x[$$0]++' $(vmlinux-dirs:%=$(objtree)/%/modules.order) > $(objtree)/modules.order

+ifdef CONFIG_SAMPLES

+	$(Q)$(MAKE) $(build)=samples

+endif

 	@echo '  Building modules, stage 2.';

 	$(Q)$(MAKE) -f $(srctree)/scripts/Makefile.modpost
--

I don't have any axe to grind here, I just could not figure out how to get

the samples/ modules built (without resorting to make M=.../samples/foo and

clobbering my source directory).  If there is another plan, point me to its

documentation.  I can't see the point of the Kconfig option for samples at all.

What does it do?
Because they are only ever modules, the samples are never built in.  I

honestly don't see why the modules shouldn't be installed like any others.

People who are really going to install the kernel for production are not

going to set this Kconfig option.
Thanks,

Roland

--

I just double checked and the usual:

make O=output samples/
It works as expected and build all the samples.

It is documented by make help like this:

$ make help

  ...

  dir/            - Build all files in dir and below
The above is generic functionality and I do not see any

The samples were added so we avoided them to bit-rot.

We do not want to build samples unconditionally - that should be pretty

obvious.

There is _nothing_ preventing us from adding stuff tomorrow that has

be build as built-in. Some sample code documenting early cpu init

stuff for instance. And we do not want this code to clash with

other kernel code if global symbols are duplicated.
This thinking that samples are solely modules are wrong. It is correct that

most of it is expected to be modules but not all of it.
I only found one little glitch.

$ make modules

does not build the samples.
	Sam

--

There used to be a make target called "samples_check" (I could have

misspelled it).  Where did it go?  Sam?
---

~Randy

--

I"m not sure what you mean by "with the kernel image".

AFAICT, everything under samples/ can only ever be built as modules.

In no event would it affect the vmlinux link.
Thanks,

Roland

--