gcc-3.0 20010607 problems

[Kurt Garloff <kurt at garloff dot de>]

Hi,

doing more extensive tests with gcc-3.0 20010607 on my machine here
(Linux-2.4.x, glibc-2.1.3), I ran into a number of problems.
All compilations are done with -O3 -ffast-math -felide-constructors
-fschedule-insns2 -funroll-loops -freduce-all-givs -frerun-loop-opt
-fstrict-aliasing  -mcpu=pentiumpro -Wall -DSMP -D_REENTRANT -D_THREAD_SAFE  
-DEXCEPT=1

(1) Linking problems: __default_alloc_template<true, 0>

    g++ -o opencoax opencoax.o -L../../lina/bin-ix86 -L../bin-ix86 -lFD -ltbcicplxdouble -ltbcidouble -lpthread
 ../bin-ix86/libFD.so: undefined reference to std::__default_alloc_template<true, 0>::_S_start_free'
 ../bin-ix86/libFD.so: undefined reference to std::__default_alloc_template<true, 0>::_S_free_list'
 ../bin-ix86/libFD.so: undefined reference to std::__default_alloc_template<true, 0>::_S_chunk_alloc(unsigned, int&)'
 ../bin-ix86/libFD.so: undefined reference to std::__default_alloc_template<true, 0>::_S_heap_size'
 ../bin-ix86/libFD.so: undefined reference to std::__default_alloc_template<true, 0>::_S_end_free'

 But my libstdc++ contains only:
 
 00085d04 V std::__default_alloc_template<false, 0>::_S_end_free
 0005e820 W std::__default_alloc_template<false, 0>::_S_round_up(unsigned) 
 00085cc0 V std::__default_alloc_template<false, 0>::_S_free_list
 00085d08 V std::__default_alloc_template<false, 0>::_S_heap_size
 00085d00 V std::__default_alloc_template<false, 0>::_S_start_free
 0005e8e0 W std::__default_alloc_template<false, 0>::_S_chunk_alloc(unsigned, int&)

(2) Segmentation fault when compiling a large C++ code.
 
 Compiling without support for exceptions (they are ifdefed) and with
 -fno-rtti -fno-exceptions does solve the problem. 
 (Compilation time is more than an hour of pIII-700 CPU time and mem usage
  goes up to 1GB, but the machine can handle this!)
 
 The segmentation fault is reproducible and does not occur when the system
 is going low on memory, but happens much earlier.
 
 ../lina/include/band_matrix.h: In member function TBCI::BdMatrix<T>& 
    TBCI::BdMatrix<T>::reconfig(const TBCI::BVector<unsigned int>&) [with T = TBCI::cplx<double>]':
       ../lina/include/band_matrix.h:939:   instantiated from
   TBCI::BdMatrix<T>& TBCI::BdMatrix<T>::expand(unsigned int) [with T = TBCI::cplx<double>]'
 ../lina/include/band_matrix_gd.h:61:   instantiated from int
   TBCI::lu_decomp(TBCI::BdMatrix<T>&) [with T = TBCI::cplx<double>]'
 ../lina/include/solver/bd_lu_solver.h:162:   instantiated from
   TBCI::TVector<T> TBCI::lu_solve(TBCI::BdMatrix<T>&, const TBCI::Vector<T>&)
   [with T = TBCI::cplx<double>]'
 source/hphi_wave.cc:130:   instantiated from here
 ../lina/include/band_matrix.h:742: Internal error: Segmentation fault
 Please submit a full bug report,
 [...]
 
 Another one:
 
 include/dops_2d.h: In member function Dop<T, 2> Dops2d<T>::get_all_pts(const 
    Coordinate<2>&) const [with T = double]':
 /raid/gcc300/include/g++-v3/bits/codecvt.h:228:   instantiated from here
    include/dops_2d.h:92: Internal error: Segmentation fault
 Please submit a full bug report,
 [...]   

(3) ICE in allocate_dynamic_stack_space
 (This happens compiling the second example of (2) without exceptions, i.e.
  -fno-rtti -fno-exceptions)

 /usr/src/i686-pc-gcc297/include/g++-v3/bits/std_ostream.h: In function int 
    main(int, char**)':
 /usr/src/i686-pc-gcc297/include/g++-v3/bits/std_ostream.h:684: Internal 
    compiler error in allocate_dynamic_stack_space, at explow.c:1388
 Please submit a full bug report,
 [...]
 
 Machine is not out of memory, but maybe the compiler's memory allocator has
 hit some limit? (The same might be true for example (2), of course ...)
 

I know that the above info is not really useful as a bug report (excpet for
(1) maybe), I'll try to prepare testcases and submit proper bug reports ASAP.
Unfortunately, this will unlikely happen before monday.

I just wanted to let you know right now, maybe somebody knows the problems
and happens to have some good idea what's going on.

Regards,
-- 
Kurt Garloff                   <kurt@garloff.de>         [Eindhoven, NL]
Physics: Plasma simulations  <K.Garloff@Phys.TUE.NL>  [TU Eindhoven, NL]
Linux: SCSI, Security          <garloff@suse.de>   [SuSE Nuernberg, FRG]
 (See mail header or public key servers for PGP2 and GPG public keys.)

PGP signature