On Mon, Jan 18, 2010 at 3:33 PM, chandra reddy <creddym@xxxxxxxxx> wrote:
> Hi Eljay,
>
> My requirement is not generate two executable outputs. I want do the
> following under a new command line option to gcc (new option is needed
> because it is unusual producing two object files in single compilation)
>
> I want to create libabc.a and libabc.so in the following way:
>
> $gcc <cflags> -c <new_option> -fpic a.c
> $gcc <cflags> -c <new_option> -fpic b.c
> $gcc <cflags> -c <new_option> -fpic c.c
>
> The output would be: a.o a.So b.o b.So c.o c.So
> where a.o b.o c.o will be generated with -fpic disabled
> a.So b.So c.So will be generated with -fpic enabled
> Now,
>
> I will create libabc.a and libabc.so like this.
>
> $ ar cq libabc a.o b.o c.o
> $ld -o libabc.so a.So b.So c.So
>
> I want this feature only to create object files not for executables. We have
> to create thousands of static and dynamic libraries with and without -fpic
> option.
> It helps us if the above feature can be achieved in the gcc compiler.
>
> I can see that until reaching the function call "expand_gimple_basic_block"
> in gcc compiler, the code flow is same with/without -fpic. After this
> function call tree would differ.
It already differs before (see targetm.binds_local).
Richard.
> Thanks
> -Chandra
>
>
>
>
>
> On Mon, Jan 18, 2010 at 7:31 PM, John (Eljay) Love-Jensen <eljay@xxxxxxxxx>
> wrote:
>>
>> Hi Chandra (and Richard),
>>
>> > Would it be possible to the compiler to generate two object
>> > files:.<file>.o (general obj file) and <file>.So (obj file with -fpic) in
>> > single compilation?
>>
>> No.
>>
>> Consider, for example:
>>
>> int main()
>> {
>> #ifdef TWENTY
>> return 10;
>> #else
>> return 20;
>> #endif
>> }
>>
>> It is not possible to have two executable outputs. Something like...
>>
>> gcc -o ten.exe -DTWENTY -o twenty.exe test.cpp
>>
>> I could write a gcc-bash front end, which would parse the parameters and
>> run gcc twice for me...
>>
>> gcc -o ten.exe
>> gcc -DTWENTY -o twenty.exe test.cpp
>>
>> ...but that does not reduce any time for building anything.
>>
>> Even if GCC had the capability to do what I envision my gcc-bash front
>> end, I cannot see how that convenience would reduce any build time, since
>> none of the effort of the compilation process would be reusable for the
>> subsequent passes.
>>
>> Sincerely,
>> --Eljay
>
>
>
> --
> Thanks,
> cr();
> --------------------------------------------------------------------------------------------------------------------------
> "Remote debugging a buggy debugger with a cross buggy debugger is a funny
> thing"
> --------------------------------------------------------------------------------------------------------------------------
>
>
