Bug#941774: lintian: False positive for symbols-file-contains-current-version-with-debian-revision
Hi Felix,
I forgot all about this, thanks for following up.
On Sat, Dec 14, 2019 at 02:15:41PM -0800, Felix Lechner wrote:
> $ fgrep -- -5 dir2/symbols
> _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE12_M_constructIPcEEvT_S7_St20forward_iterator_tag@Base
> 1.21.1-5+b1
>
> The symbol shows a Debian revision. Lintian is right.
Aha - I was searching the symbols file from the source package. Now I see that
the binary package's symbols file is not necessarily the same. Sorry for the
mistake.
> As a side note, I was surprised to find 188 additional Debian revisions:
These are intentional (and have overrides). See "Note on eolian-generated
symbols" in README.source if you're curious.
> Finally, I am not sure why some symbols were decoded properly using
> the appropriate pattern [1], while the offender is raw 'c++'. Did you
> mix C and C++ symbols in the same shared library?
Yes, by accident - libephysics uses a c++ library, and was leaking this c++
symbol. Looks like gcc8 -O2 inlined this function, while gcc9 doesn't until
-O3.
Ross
Reply to: