Re: RFS: libvigraimpex (updated package)

[Jakub Wilk <ubanus@users.sf.net>]

* George Danchev <danchev@spnet.net>, 2009-09-04, 18:06:
> http://mentors.debian.net/debian/pool/main/l/libvigraimpex/libvigraimpex_1.6.0-2.dsc
> > I would be glad if someone uploaded this package for me.
>
> I'm not that optimistic with the C++ code and symbols files.
>
> First, yours appear to be architecture dependent:
> http://qa.debian.org/cgi-bin/mole/seedsymbols?pkgname=libvigraimpex2ldbl

Yes, I'm aware of that:

$ grep arch= debian/libvigraimpex2ldbl.symbols | c++filt
  (arch=!alpha !amd64 !ia64 !s390)vigra::void_vector_base::resize(unsigned int)@Base 1.5.0
  (arch=alpha amd64 ia64 s390)vigra::void_vector_base::resize(unsigned long)@Base 1.5.0

> Second, there are problems: see #521569

Yes, using -c 2 or higher is a bit risky. But I don't do such things.

> and the whole thread at:
> http://lists.debian.org/debian-devel/2009/04/msg00297.html

Right, but since #521551 is fixed, providing symbols files for C++ 
libraries is not such a big issue anymore. (That reminds me that I am 
missing versioned build dependency on dpkg-dev.)

> Third, as I can see from debian/rules, dh_makeshlibs won't pass any check
> option to dpkg-gensymbols and the default -c1 is used which is a request for
> FTBFS on certain architectures in that case.
>
> So, I suggest we remove the symbols part or simply disable it (pass -c0 to
> dpkg-gensymbols, as I did with one of my library packages bobcat) until a
> better solution is found or GCC people accept the solution provided at:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36022

I can pass -c0 to dpkg-gensymbols if you insist, but I don't think this is
really necessary.

--
Jakub Wilk