[Bug c++/37452] New: -fPIC and inline seems to generate incorrect code with nested class templates

["notquitezeus at gmail dot com" <gcc-bugzilla@gcc.gnu.org>]

uname -a:
Linux lit3n4014e 2.6.22.18-0.2-bigsmp #1 SMP 2008-06-09 13:53:20 +0200 i686
i686 i386 GNU/Linux

OS is SuSE 10.3 IA32 on a Core2 Duo machine. 

See attachments for g++ -dumpspecs, pre-processed source that demonstrates the
problem, assembly generated by g++ and (on the chance that it's useful)
assembly generated by the latest version of Intel's compiler. 

There's a segfault on what gdb thinks is the first "real" line of
PerformanceStatistics<enabled_domain>::Scope<enabled_domain::function_c>::~Scope().
The segfault also occurs in the same place if:

1) I replace lCycles = mCounter.GetElapsedTime() with lCycles = 0
2) I leave the line lCycles = mCounter.GetElapsedTime() untouched and instead
comment out the rest of the if statement

The problem only occurs when -fPIC and -finline are both enabled. I can compile
with -O3 -fno-inline and the problem goes away or I can compile with -O0
-finline and the problem appears.


-- 
           Summary: -fPIC and inline seems to generate incorrect code with
                    nested class templates
           Product: gcc
           Version: 4.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: notquitezeus at gmail dot com
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37452