i2c-tools: lib/Module.mk: fixing LIB_LINKS dependency

LIB_LINKS should be added as a dependency only when
BUILD_DYNAMIC_LIB is enabled

Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
---
 lib/Module.mk | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

Hi Angelo,

On Mon, 11 Dec 2017 18:00:24 +0100, Angelo Compagnucci wrote:
> LIB_LINKS should be added as a dependency only when
> BUILD_DYNAMIC_LIB is enabled
> 
> Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
> ---
>  lib/Module.mk | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/lib/Module.mk b/lib/Module.mk
> index 8a58f5b..de189a2 100644
> --- a/lib/Module.mk
> +++ b/lib/Module.mk
> @@ -27,10 +27,9 @@ LIB_SHSONAME	:= $(LIB_SHBASENAME).$(LIB_MAINVER)
>  LIB_SHLIBNAME	:= $(LIB_SHBASENAME).$(LIB_VER)
>  LIB_STLIBNAME	:= libi2c.a
>  
> -LIB_LINKS	:= $(LIB_SHSONAME) $(LIB_SHBASENAME)
> -
>  LIB_TARGETS	:=
>  ifeq ($(BUILD_DYNAMIC_LIB),1)
> +LIB_LINKS	:= $(LIB_SHSONAME) $(LIB_SHBASENAME)
>  LIB_TARGETS	+= $(LIB_SHLIBNAME)
>  endif
>  ifeq ($(BUILD_STATIC_LIB),1)

This is correct, good catch. I'll commit it. In fact I thought about it
when reviewing your previous patch, but (wrongly) assumed it did not
matter. After adding all the missing dependencies, it clearly does.

I have noticed another problem. While "make BUILD_STATIC_LIB=0"
actually prevents the static flavor of the library from being built,
"make BUILD_DYNAMIC_LIB=0" does not prevent the dynamic flavor of the
library from being built. The reason is that USE_STATIC_LIB is not set,
so tools depend on the dynamic library, which in turn gets built as a
dependency.

I suppose that BUILD_DYNAMIC_LIB=0 should imply USE_STATIC_LIB=1.
Something like:

--- a/Makefile
+++ b/Makefile
@@ -43,6 +43,8 @@ endif
 ifeq ($(BUILD_DYNAMIC_LIB),0)
 ifeq ($(BUILD_STATIC_LIB),0)
 $(error BUILD_DYNAMIC_LIB and BUILD_STATIC_LIB cannot be disabled at the same time)
+else
+USE_STATIC_LIB := 1
 endif
 endif

What do you think?

Thanks,

DEar Jean,

On Thu, Dec 14, 2017 at 9:36 AM, Jean Delvare <jdelvare@suse.de> wrote:
> Hi Angelo,
>
> On Mon, 11 Dec 2017 18:00:24 +0100, Angelo Compagnucci wrote:
>> LIB_LINKS should be added as a dependency only when
>> BUILD_DYNAMIC_LIB is enabled
>>
>> Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
>> ---
>>  lib/Module.mk | 3 +--
>>  1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/lib/Module.mk b/lib/Module.mk
>> index 8a58f5b..de189a2 100644
>> --- a/lib/Module.mk
>> +++ b/lib/Module.mk
>> @@ -27,10 +27,9 @@ LIB_SHSONAME       := $(LIB_SHBASENAME).$(LIB_MAINVER)
>>  LIB_SHLIBNAME        := $(LIB_SHBASENAME).$(LIB_VER)
>>  LIB_STLIBNAME        := libi2c.a
>>
>> -LIB_LINKS    := $(LIB_SHSONAME) $(LIB_SHBASENAME)
>> -
>>  LIB_TARGETS  :=
>>  ifeq ($(BUILD_DYNAMIC_LIB),1)
>> +LIB_LINKS    := $(LIB_SHSONAME) $(LIB_SHBASENAME)
>>  LIB_TARGETS  += $(LIB_SHLIBNAME)
>>  endif
>>  ifeq ($(BUILD_STATIC_LIB),1)
>
> This is correct, good catch. I'll commit it. In fact I thought about it
> when reviewing your previous patch, but (wrongly) assumed it did not
> matter. After adding all the missing dependencies, it clearly does.
>
> I have noticed another problem. While "make BUILD_STATIC_LIB=0"
> actually prevents the static flavor of the library from being built,
> "make BUILD_DYNAMIC_LIB=0" does not prevent the dynamic flavor of the
> library from being built. The reason is that USE_STATIC_LIB is not set,
> so tools depend on the dynamic library, which in turn gets built as a
> dependency.
>
> I suppose that BUILD_DYNAMIC_LIB=0 should imply USE_STATIC_LIB=1.
> Something like:
>
> --- a/Makefile
> +++ b/Makefile
> @@ -43,6 +43,8 @@ endif
>  ifeq ($(BUILD_DYNAMIC_LIB),0)
>  ifeq ($(BUILD_STATIC_LIB),0)
>  $(error BUILD_DYNAMIC_LIB and BUILD_STATIC_LIB cannot be disabled at the same time)
> +else
> +USE_STATIC_LIB := 1
>  endif
>  endif
>
> What do you think?

I think you are totally right. If BUILD_DYNAMIC_LIB is 0 the static
variant should be built and used too.

Could you realease a new version of i2c-tools with these patches? I'm
in the phase of updating the package in buildroot and I do really
prefer to have a new stable release than an old release with a bunch
of patches on top.

Thanks!

>
> Thanks,
> --
> Jean Delvare
> SUSE L3 Support

On Thu, 14 Dec 2017 10:43:33 +0100, Angelo Compagnucci wrote:
> On Thu, Dec 14, 2017 at 9:36 AM, Jean Delvare <jdelvare@suse.de> wrote:
> > I suppose that BUILD_DYNAMIC_LIB=0 should imply USE_STATIC_LIB=1.
> > Something like:
> >
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -43,6 +43,8 @@ endif
> >  ifeq ($(BUILD_DYNAMIC_LIB),0)
> >  ifeq ($(BUILD_STATIC_LIB),0)
> >  $(error BUILD_DYNAMIC_LIB and BUILD_STATIC_LIB cannot be disabled at the same time)
> > +else
> > +USE_STATIC_LIB := 1
> >  endif
> >  endif
> >
> > What do you think?  
> 
> I think you are totally right. If BUILD_DYNAMIC_LIB is 0 the static
> variant should be built and used too.

Thanks. Both fixes committed.

> Could you realease a new version of i2c-tools with these patches? I'm
> in the phase of updating the package in buildroot and I do really
> prefer to have a new stable release than an old release with a bunch
> of patches on top.

I don't have such a plan, sorry. Version 4.0 is still relatively new, I
expect more problems to be found and fixed in the next weeks, I don't
think it makes sense to rush a new release. There are only 13 commits
since 4.0, it's not that many.

Backporting patches for distribution packages is the usual process, and
6 patches is not a bunch. If the patch count really bothers you, you can
always consolidate the fix-ups into the original patch and you'll be
down to 4, that seems easily manageable. Quilt is your friend ;-)