svn commit: r540710 - head/sysutils/arcconf

[Thomas Vogt]

> On 23 Feb 2021, at 13:42, Gerald Pfeifer <gerald at pfeifer.com> wrote:
> 
> On Wed, 3 Feb 2021, Thomas Vogt wrote:
>>>>> - Update Makefile to use current version of GCC as defined in
>>>>>   bsd.default-versions.mk. (USE_GCC=yes)
>>>> ...
>>>>> LIB_DEPENDS=	libstdc++.so.6:lang/gcc${GCC_DEFAULT}
>>>>> +USE_GCC=	yes
>>>> This looks very (very) odd.
>>>> 
>>>> USE_GCC already implies a run-time dependency on lang/gcc${GCC_DEFAULT}
>>>> and I do not recall seeing such a LIB_DEPENDS before.
>>>> 
>>>> Are you sure this is really, really needed with USE_GCC=yes?
>>> I believe, seeing NO_BUILD=yes, it's because this is some sort of binary-
>>> only program which links against libstdc++.so.6.  I can't verify because
>>> the distfile is ~73MB and that's a bit too much for me to download ATM.
>> arcconf links to libstdc++.so.6. An ancient libstdc++.so.6 version was 
>> provided with old freebsd versions (included with an old gcc). The 
>> binary failed to run with:  ld-elf.so.1: /usr/lib/libstdc++.so.6: 
>> version GLIBCXX_3.4.11 required ..
>> 
>> Newer gcc versions from the port provided libstdc++.so.6 with >= 
>> GLIBCXX_3.4.11. Thats the reason why gcc was added.
> 
> So I am wondering: Isn't USE_GCC=yes sufficient?
> 
> USE_GCC=yes implies 
>  RUN_DEPENDS+=   ${CC}:lang/gcc${GCC_DEFAULT}
> which pulls in lang/gcc${GCC_DEFAULT} and its run-time libraries at
> run time.
> 
> The LIB_DEPENDS statement above does not add to that.

Yes. Maybe RUN_DEPENDS is more obvious than USE_GCC.

Regards,
Tom