This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Current CVS bootstrap failure on i686-pc-linux-gnu
- To: gcc-bugs at gcc dot gnu dot org
- Subject: Current CVS bootstrap failure on i686-pc-linux-gnu
- From: Gabriel Dos Reis <gdr at codesourcery dot com>
- Date: 28 May 2000 22:34:38 +0200
- Organization: CodeSourcery, LLC
I think the patch
2000-05-27 Geoffrey Keating <geoffk@cygnus.com>
is causing the bootstrap failure I seeing on an i686-pc-linux-gnu.
The bootstrap fails with the following message:
[...]
/home/gaby/egcs/gcc/regclass.c: In function `record_operand_costs':
/home/gaby/egcs/gcc/regclass.c:864: `reg_changes_size' undeclared (first use in this function)
/home/gaby/egcs/gcc/regclass.c:864: (Each undeclared identifier is reported only once
/home/gaby/egcs/gcc/regclass.c:864: for each function it appears in.)
[...]
make[2]: *** [regclass.o] Error 1
make[2]: Leaving directory `/tmp/E/gcc'
make[1]: *** [bootstrap] Error 2
make[1]: Leaving directory `/tmp/E/gcc'
make: *** [bootstrap] Error 2
Indeed the variable reg_changes_size is conditionally defined as:
#ifdef CLASS_CANNOT_CHANGE_SIZE
/* These are the classes containing only registers that can be used in
a SUBREG expression that changes the size of the register. */
static int class_can_change_size[N_REG_CLASSES];
/* Registers, including pseudos, which change size. */
static regset reg_changes_size;
#endif /* CLASS_CANNOT_CHANGE_SIZE */
But it is used unconditionally.
-- Gaby
CodeSourcery, LLC http://www.codesourcery.com