Moving the applet directories.

[Bernhard Fischer]

On Wed, Feb 01, 2006 at 07:21:17PM +0100, Peter S. Mazinger wrote:
>On Wed, 1 Feb 2006, Rob Landley wrote:
>
>> On Tuesday 31 January 2006 02:54, Peter S. Mazinger wrote:
>> > On Mon, 30 Jan 2006, Rob Landley wrote:
>> >
>> > [...]
>> >
>> > > --- Rules.mak   (revision 13743)
>> > > +++ Rules.mak   (working copy)
>> > > @@ -222,7 +223,7 @@
>> > >      STRIPCMD:=$(STRIP) -s --remove-section=.note
>> > > --remove-section=.comment endif
>> > >  ifeq ($(strip $(CONFIG_STATIC)),y)
>> > > -    LDFLAGS += $(call check_ld,--static,)
>> > > +    LDFLAGS += $(call check_ld,-static,)
>> > >  #else
>> > >  #    LIBRARIES += -ldl
>> >
>> > This seems to me an overkill, who can't use -static?
>> 
>> It seems like overkill to me too, but I'm not a makefile expert.
>> 
>> I now have the opportunity to intermittently test on cygwin, and if it works 
>> there that's about as evil and noncompliant an environment as we're ever 
>> likely to see.
>> 
>> > > Something about the way check_ld unnecessarily mangles static into a -W,l
>> > > option instead of a normal gcc option is confusing it...
>> > >
>> > > Rob
>> >
>> > an --as-needed -lgcc_s --no-as-needed could do that too, either it is bb
>> > or in your gcc specs.
>> 
>> Ok, so somehow:
>> 
>>   -W,l,--static --as-needed -lgcc_s --no-as-needed
>
>that is completely wrong, if you want static build, then it's sure you 
>don't want to depend on libgcc_s.so (there is no libgcc_s.a)
>
>gcc -static is preferred (not ld -static), that will take care of proper 

I did that sometimes earlier this week; see svn.

>crt*.o files and will link against libc.a/libgcc.a/libgcc_eh.a if needed.
>
>--as-needed construct is only interesting for non-static builds

We will have to revamp our whole make rules after 1.1.1 is out (I, at
least won't have time to look at this until in approximately 10 days or
so anyway, for what that's worth):

We need -fPIC when building a .so, but this would need touching about
any Makefile.in to add it to the respective cflags used to build the
prerequisites of the .a (resp. that what would be the prerequisites for
.a if not in IMA mode).
I refuse to do this, it's unesthetic.
Better provide generic Rules e.g. like Peter does for uClibc.

We only need whole-archive if not compiling in IMA mode; That is merely
cosmetic and not too kludgy to do, but it would be better suited for a
generic rule.