This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [3.3] [powerpc-linux] libjava testsuite failures with --enable-debug
- From: Daniel Jacobowitz <drow at mvista dot com>
- To: Matthias Klose <doko at cs dot tu-berlin dot de>
- Cc: gcc at gcc dot gnu dot org, java at gcc dot gnu dot org
- Date: Wed, 30 Apr 2003 12:58:32 -0400
- Subject: Re: [3.3] [powerpc-linux] libjava testsuite failures with --enable-debug
- References: <16047.65054.59393.220515@gargle.gargle.HOWL>
On Wed, Apr 30, 2003 at 06:47:26PM +0200, Matthias Klose wrote:
> Configuring with --enable-debug on powerpc-linux shows about 230
> additional failures in the testsuite, all of them of the form:
>
> ASSERTION FAILURE: ../../../src/libffi/src/powerpc/ffi.c line 118
>
> Looking at libffi/include/ffi_common.h, the assertion triggers, if the
> macro argument is not properly aligned.
>
> ffi.c line 118:
> /* Check that everything starts aligned properly. */
> FFI_ASSERT(((unsigned)(char *)stack & 0xF) == 0);
>
> Configuring without --enable-debug results in an empty FFI_ASSERT
> definition.
Thanks for tracking this down, at last!
> There seem to be two bugs:
>
> - the meaning of --enable-debug is not well defined. For libstdc++
> an extra debugging library is built, for libffi, only one library
> with debugging information is built.
>
> - On powerpc-linux, the assertion triggers for a wrong reason, or
> it reveals an alignment problem.
>
> Append are the test-summery of libjava configure with --enable-debug.
FWIW, this is an alignment problem - the stack should be 16-byte
aligned. Probably the wrong one of the multitude of stack alignment
macros is being used.
--
Daniel Jacobowitz
MontaVista Software Debian GNU/Linux Developer