[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#936924: Moving libsvm to Debian Science team



Hi,

On Wed, Dec 25, 2019 at 06:26:49PM +0100, Christian Kastner wrote:
> 
> If you like gbp, then it's already a team win :-)

+1
  
> Yeah, the C++ symbols generated are the mangled versions. You need to
> unmangle them by piping the generated file through c++filt, and then add
> the (c++) prefix to tell dpkg-gensymbols about this. The
> UsingSymbolsFiles [1] page has a practical example on the bottom.
> 
> The c++filt(1) and dpkg-gensymbols(1) man pages have theoretical
> background on this, but TBH, I don't think it's needed.

As I tried to express:  I do not agree that the effort to maintain a
symbols file is really needed.  I just do so if upstream randomly
changes ABI.  Than dh_makeshlibs fails in the packaging step which is a
flag that we need to bump SOVERSION.  If you have other means to spot
the need of a SOVERSION bump that's fine.
 
> There are voices that say that symbols files for C++ libraries are
> overly hard to maintain, but in case of this specific upstream, I didn't
> have a negative experience, and it did help me detect breaking changes a
> few times.

Yes, exactly that's the point.   BTW, sometimes it helps to just have a
*.symbols.amd64 file which is sufficient to detect these changes while
there is no need to fiddle with other architectures which in some cases
might have different symbols and need extra care.

Kind regards and thanks for caring for these packages

      Andreas.
 
> [1] https://wiki.debian.org/UsingSymbolsFiles
> 

-- 
http://fam-tille.de


Reply to: