C++ Compiler 5.11 with dwarf produces twice as big binarries as with -xdebugformat=stabs -xs

Question

I am evaluating migrating to Studio 12u2 from Studio 11. First thing I noticed is that due to new debug info format (dwarf) binaries became twice bigger than they were when compiled with Studio 11 with -xs option (which puts debug info into binaries). Here are the sizes of shared library built with different Studios
1. Studio 11 with -g0 -xs:                                               411M
2. Studio 12u2 with -g0 (dwarf):                                      900M
3. Studio 12u2 with -g0, -xs, -xdebugformat=stabs:          481M
I understand that DWARF uses more space, but doubling size of binaries is not acceptable for us.
From other posts (Huge debugging binaries after switching from C++ Sun Studio 11 to 12u1)   I grasped that DWARF should produces similar size as stabs with -xs optionm which is clearly not the case.
The question is that if it is possible to reduce size of dwarf debug info with some compiler option or by reworking structure of the shared library.
PS. Another problem with 12u2, is that when I tried to workaround with -xdebugformat=stabs, it produced binary of acceptable size (20% bigger than Studio 11), but for some reason the shared library was missing some symbols in its table, while there were present in object files (worked correctly with 12u3 though).

Answer 1

Hello,
> I am interested in any symbols that occur more times than the number of object files that went into your shared libraries
There are actually quite a lot of such symbols: 2174 to be exact (out of 53152 unique symbols, 1983140 non-unique symbols, with 154 object files went to the lib). I will post top 30. I can provide the rest if you tell me where to drop it.
1401 std::basic_string<char,std::char_traits<char>,std::allocator<char> >::replace
1377 std::basic_string<wchar_t,std::char_traits<wchar_t>,std::allocator<wchar_t> >::replace
1360 VariantValue::VariantValue
1125 STATUS_TYPE::STATUS_TYPE
1125 SimpleHasher::hash
1125 SimpleHasher::equals
1102 InitField 
1016 CustomField<std::basic_string<char,std::char_traits<char>,std::allocator<char> > >::Write
1016 CustomField<std::basic_string<char,std::char_traits<char>,std::allocator<char> > >::Read
1016 CustomField<long long>::Write
1016 CustomField<long long>::Read
1016 CustomField<int>::Write
1016 CustomField<int>::Read
928 CustomField<DateTime>::Write
928 CustomField<DateTime>::Read
924 std::basic_string<char,std::char_traits<char>,std::allocator<char> >::insert
918 std::basic_string<wchar_t,std::char_traits<wchar_t>,std::allocator<wchar_t> >::insert
770 std::fpos<__mbstate_t>::operator>=
770 std::fpos<__mbstate_t>::operator>
770 std::fpos<__mbstate_t>::operator==
770 std::fpos<__mbstate_t>::operator<=
770 std::fpos<__mbstate_t>::operator<
770 std::fpos<__mbstate_t>::operator+=
770 std::fpos<__mbstate_t>::operator+
770 std::fpos<__mbstate_t>::operator!=
770 std::fpos<__mbstate_t>::operator-=
770 std::fpos<__mbstate_t>::operator-
762 std::list<std::basic_string<char,std::char_traits<char>,std::allocator<char> >,std::allocator<std::basic_string<char,std::char_traits<char>,std::allocator<char> > > >::list
756 std::map<int,OCIDefine*,std::less<int>,std::allocator<std::pair<const int,OCIDefine*> > >::insert
756 std::map<int,OCIBind*,std::less<int>,std::allocator<std::pair<const int,OCIBind*> > >::insert
> I am also interested in symbols whose counts are higher than you can find references to in your own source files.
I will have to work on this for some time.
PS I noticed that many of symbols are quite long (e.g. 3000 chars long, 11K non-unique symbols with length > 1K), which makes it on average 147 chars per non-unique symbol. Is that normal?
Thank you.