[Lsb-infrastructure] [Bug 2059] New: Insufficient inheritance data for some libstdc++ classes

bugzilla-daemon at linux-foundation.org bugzilla-daemon at linux-foundation.org
Tue Apr 22 06:10:08 PDT 2008


http://bugs.linuxbase.org/show_bug.cgi?id=2059

           Summary: Insufficient inheritance data for some libstdc++ classes
           Product: Infrastructure
           Version: unspecified
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Database
        AssignedTo: mats at freestandards.org
        ReportedBy: silakov at ispras.ru
         QAContact: lsb-infrastructure at freestandards.org


Created an attachment (id=904)
 --> (http://bugs.linuxbase.org/attachment.cgi?id=904)
SQL to add necessary data

In the recent update we've set some fields in the ClassInfo table that were
equal to zero (such as CIrtti, CIbase, etc.) for libstdc++ classes. This change
brought to light lack of data for some classes which leads to libchk failures
at the moment (the thing is that libchk doesn't test anything for classes which
have zeroes in almost all ClassInfo fields, even if these classes are
included).

The SQL attached will add data necessary to satisfy libchk.

Another tricky problem that should be solved by the patch is ambiguity of two
classes included in the LSB - std::basic_istream<char, std::char_traits<char> >
and std::basic_ostream<char, std::char_traits<char> >. The thing is that each
of them can have two different mangled names - 'normal' name, constructed by
general gcc mangling rules, and the 'short' one - _ZSi and _ZSo.

LSB includes both normal and short names in these cases. However, this causes
problems with RTTI information check - libchk tries to check RTTI for
_ZSt13basic_ostreamIcSt11char_traitsIcEE and finds '_ZSo' name which really
confuses it. The SQL will set CIrtti for problematic entries back to zero,
forcing libchk to omit the checks (though I think it future it would be better
to patch libchk to compare unmangled names; maybe this will solve the problem).

The SQL will also set for 'normal' entries the same Vtable information as for
'short' entries. Note that this probably will affect the specification text.


-- 
Configure bugmail: http://bugs.linuxbase.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.


More information about the lsb-infrastructure mailing list