Hello.
On 08/16/2018 03:29 PM, Michael Matz wrote:
> Hi,
>
> On Thu, 16 Aug 2018, Martin Liška wrote:
>
>> SHLIB_SOVERSION = 1
>> SHLIB_SONAME = @shlib_base_name@.so.$(SHLIB_SOVERSION)
>>
>> which would require a substitution for soversion, show I generalize even
>> so much?
>
> I wouldn't bother. libgcov.so.1 sounds just fine, and due to symbol
> versioning we never need to change the SO version anymore (like with
> libgcc_s).
Good.
>
>> About the location of libgcov.{a,so}, I believe right place would be now:
>>
>> /home/marxin/bin/gcc/lib64/libgcov.so.1
>> /home/marxin/bin/gcc/lib64/libgcov.so
>> /home/marxin/bin/gcc/lib64/libgcov.a
>
> No. With version dependend dirs the .so symlink and the .a file have to
> be placed into the version dependend dirs. Look at where libgcc_s.so and
> libgcc.a are placed on your system.
>
>> Because of LD_LIBRARY_PATH location I believe libgcov.so should not be in:
>> /home/marxin/bin/gcc/lib64/gcc/x86_64-pc-linux-gnu/9.0.0/libgcov.so
>
> LD_LIBRARY_PATH doesn't affect the .so symlink (which is used for link
> editing), but where the full SONAME library (i.e. libgcov.so.1) is found.
> _That_ one should indeed be placed into a normal system directory like
> /usr/lib, again look at the distribution of files from your system.
Ok, so proper layout would be:
/home/marxin/bin/gcc/lib64/libgcov.so.1
/home/marxin/bin/gcc/lib64/gcc/x86_64-pc-linux-gnu/9.0.0/libgcov.so
/home/marxin/bin/gcc/lib64/gcc/x86_64-pc-linux-gnu/9.0.0/libgcov.a
Where /home/marxin/bin/gcc/lib64/gcc/x86_64-pc-linux-gnu/9.0.0/libgcov.so
points to
/home/marxin/bin/gcc/lib64/libgcov.so.1 ?
>
>> One another question has popped up:
>>
>> libgcov.a archive consists of many one-function-per-object-file objects,
>> this is guarded with e.g. #ifdef L_gcov_reset macro. For shared library I
>> need
>> something like:
>> +#if defined(L_gcov_reset) || defined(L_gcov_shared)
>>
>> question is whether it worth separating the functions into separate objects?
>> I know
>> during linking linker can strip the unused ones. But still.
>
> No, that wouldn't make sense because the shared library needs to contain
> all the functions that are potentially callable (i.e. all of them). There
> are no unused ones.
Yes, but we build objects for libgcov.a separately, each object with one
L_gcov_* macro defined.
So still it works in case of the static library.
Martin
>
>
> Ciao,
> Michael.
>
