Hi James,Thanks for taking a look at the code.I have tried building without linking to boost:date_time.
In debug mode,
linker errs with asking for 'boost::gregorian::greg_month::as_short_string() const'. However, in release mode, it is ok.
I tried with 'BOOST_DATE_TIME_NO_LIB' defined but still has this strange behavior.
On Sun, Nov 20, 2016 at 2:12 PM, James Clarke <jrtc27@jrtc27.com> wrote:Hi Olzhas,
> On 19 Nov 2016, at 04:14, Olzhas Rakhimov <ol.rakhimov@gmail.com> wrote:
>
> Hi James,
>
>> On Fri, Nov 18, 2016 at 10:01 AM, James Clarke <jrtc27@jrtc27.com> wrote:
>>> On 18 Nov 2016, at 02:30, Olzhas Rakhimov <ol.rakhimov@gmail.com> wrote:
>>>
>>> Hello James,
>>> Thanks again! I cleaned up things and reuploaded.
>>>
>>> Actually, I am not sure how to reproduce 'dpkg-shlibdeps' warnings,
>>> so I couldn't test the fix.
>>
>> I was building with `gbp buildpackage --git-pbuilder` in an up-to-date
>> sid chroot. I’d rather see the upstream build system stop linking
>> against the libraries rather than using --as-needed, since there’s no
>> reason why it needs to.
>
> Unfortunately, there's no standard 'FindLibXML++' module in CMake. I am using one of popular scripts found on many other projects.
> From that cmake module, I only use a command to link against libxml++2.6, but somehow the scripts seem to drag all libxml++2.6 dependencies with it.
> I am a bit reluctant to mess with those custom scripts.
> I believe those scripts are also setup for static linking, so I guess that's the reason it drags all its dependencies along.
> Currently, I don't have a better solution and couldn't find how other projects dealing with it.
> Can anyone point me into some examples dealing with libxml++ cleanly?
My guess is that the you only need the headers for those libraries, and
they have all the implementation you need, rather than needing to link
against the library itself. For example, it seems that the only reason
you use boost datetime is for the posix_time namespace to calculate
pt::to_iso_extended_string(pt::second_clock::universal_time( )).
Presumably these are simple enough that they are self-contained in the
header files. Given this explanation, I’m less concerned about including
--as-needed, since dropping the libraries could cause build errors in
future if you need functions that aren’t implemented in the headers.
At this point, I'll really be happy if '--as-needed' solution is accepted.Thanks,Olzhas