#16 g++ 4.5.0, internal compiler error

[Anonymous]

Don't know if this is the right place to post this - please excuse, if not. We're having some trouble with g++ 4.5.0... seems to be a compiler error (pl. see below). Is it save to use the 4.4.1 toolchain, instead? I read somewhere that the 4.4.1 compiler produces bad code, and shouldn't be used!

Also, do you automatically apply new patches into the 4.4.1 toolchain (e.g. libgen.h support)?!

Thanks.

" In file included from /server/devel/sdv/univ/uwe/win64/func/func_unop.h:20:0,
from /server/devel/sdv/univ/uwe/win64/func/func.h:59,
from /server/devel/sdv/univ/uwe/win64/ph4_lib/ph_path_filter.h:5,
from ph_postfilter.C:3:
/server/devel/sdv/univ/uwe/win64/func/func_cplxfft.h: In member function 'void func::cfft<V, C, R>::fft_func(V*, int) [with V = umath::vektor<std::complex<double>, 3>, C = std::complex<double>, R = double]':
/server/devel/sdv/univ/uwe/win64/func/func_cplxfft.h:183:2: error: invalid rhs for gimple memory store
z1$_x0$_M_value

c$_x0$_M_value

# .MEM_201 = VDEF <.MEM_476>
z1$_x0$_M_value = c$_x0$_M_value;

/server/devel/sdv/univ/uwe/win64/func/func_cplxfft.h:183:2: error: invalid rhs for gimple memory store
z1$_x1$_M_value

c$_x1$_M_value

# .MEM_202 = VDEF <.MEM_201>
z1$_x1$_M_value = c$_x1$_M_value;

/server/devel/sdv/univ/uwe/win64/func/func_cplxfft.h:183:2: error: invalid rhs for gimple memory store
z1$_x2$_M_value

c$_x2$_M_value

# .MEM_203 = VDEF <.MEM_202>
z1$_x2$_M_value = c$_x2$_M_value;

/server/devel/sdv/univ/uwe/win64/func/func_cplxfft.h:183:2: internal compiler error: verify_stmts failed
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

Error. "

[Kai Tietz]

Hello,

yes we will update our source packages and pre-build toolchains for 4.4.2 (released as 4.4.1-1 at the moment).

This kind of failures should be best reported to gcc's bugzilla at http://gcc.gnu.org/bugzilla/ so it get fixed. Please add to such a bug report there pre-compiled source, or a small testcase showing the issue.

Cheers,
Kai

[Nobody/Anonymous]

Great!

Do you have an idea when a pre-build toolchain for 4.4.2 including the latest patches will be available?

By the way, thanks a lot for your speedy replies!

[Nobody/Anonymous]

4.4.1a compiles this code without generating an internal error!

Could you provide a new build for:

mingw-w64-bin_i686-cygwin-1.5.25-15_4.4.1a.tar.bz2

...which includes the latest changes (e.g. libgen.h, etc.)?!

Thanks!

[Kai Tietz]

We already build 4.4.1a now in automated build for 32-bit and 64-bit. Eg see http://sourceforge.net/projects/mingw-w64/files/Toolchains%20targetting%20Win32/Automated%20Builds/mingw-w32-1.0-bin_i686-darwin_20091005.tar.bz2/download

Cheers,
Kai

[Kai Tietz]

The 4.4.1a targets have the prefix mingw-w32-1.0-bin_... and mingw-w64-1.0-bin_... to indicate, that they are build with branch v1.0. The builds for v1.0 branch, using gcc 4.4 and binutils head version, are build by our automated bot at Monday and Friday.

Kai

[Kai Tietz]

Thanks for telling this issue. I assume (as I build gcc 4.5.0 some hours ago) , that this bug is solved.
In general report such kind of errors on gcc's bugzilla tracker, or sent a e-mail to gcc's ML.

Cheers,
Kai