c++/6174: sparc-sun-solaris2.7 gcc-3.2 (trunk) C++, libstdc++-v3 & java testsuites have hundreds of execute failures

[ghazi at caip dot rutgers dot edu]

>Number:         6174
>Category:       c++
>Synopsis:       sparc-sun-solaris2.7 gcc-3.2 (trunk) C++, libstdc++-v3 & java testsuites have hundreds of execute failures
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          wrong-code
>Submitter-Id:   net
>Arrival-Date:   Thu Apr 04 08:26:00 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     Kaveh Ghazi
>Release:        gcc version 3.2 20020403 (experimental)
>Organization:
>Environment:
sparc-sun-solaris2.7 native as/ld
>Description:
I'm getting hundreds upon hundreds of execute failures from the C++, libstdc++-v3 & java testsuites w/solaris2.7.  See:
http://gcc.gnu.org/ml/gcc-testresults/2002-04/msg00134.html

It's not just me, independent confimation is here:
http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00465.html

It started as far back as 3/16
http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00382.html

The last relatively good run I had was dated 3/12:
http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00318.html

If I run the testsuite with RUNTESTFLAGS=--verbose, it seems that all of the tests get a SEGV but provide no other useful output.  I picked one test at random (g++.dg/eh/dtor1.C) and ran it under gdb, here's what I got:

Program received signal SIGSEGV, Segmentation fault.
0xff339ec4 in _Z41__static_initialization_and_destruction_0ii (__initialize_p=1, __priority=65535)
    at ../../../../egcc-CVS20020403/libstdc++-v3/src/locale.cc:522
522       const double __num_base::_S_scale_oct = log(10.0)/log(8.0);
Current language:  auto; currently c++
(gdb) bt
#0  0xff339ec4 in _Z41__static_initialization_and_destruction_0ii (__initialize_p=1, __priority=65535)
    at ../../../../egcc-CVS20020403/libstdc++-v3/src/locale.cc:522
#1  0xff330428 in __do_global_ctors_aux () at ../../../../egcc-CVS20020403/libstdc++-v3/libsupc++/tinfo2.cc:61
#2  0xff32f2d8 in _init ()
   from /teal/caip5/ghazi/gcc-testing/build/sparc-sun-solaris2.7/libstdc++-v3/src/.libs//libstdc++.so.4
#3  0xff3bad04 in ?? ()
#4  0xff3ba990 in ?? ()
#5  0xff3c4900 in ?? ()
#6  0xff3b2940 in ?? ()

I tried a couple of others, and they looked similar, so I'd have to guess that there is some fundamental overarching problem with constructors/destructors.

>How-To-Repeat:
Bootstrap the 3.2 trunk and run the testsuite (or just "make check-c++")
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted: