This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Mainline broken on alpha and x86-64. SPEC2000's eon does not build.
- From: Jan Hubicka <jh at suse dot cz>
- To: Diego Novillo <dnovillo at redhat dot com>
- Cc: stevenb at suse dot de, jh at suse dot cz, ebotcazou at act-europe dot fr,mckinlay at redhat dot com, jlquinn at optonline dot net,"gcc at gcc dot gnu dot org" <gcc at gcc dot gnu dot org>
- Date: Thu, 27 May 2004 16:44:28 +0200
- Subject: Re: Mainline broken on alpha and x86-64. SPEC2000's eon does not build.
- References: <1085667394.9713.217.camel@localhost.localdomain>
>
> One of the attached patches broke mainline bootstraps on alpha and
> x86-64. We fail to build libjava with:
>
> /home/cygnus/dnovillo/perf/sbox/gcc/local.x86_64/src/libjava/javax/swing/JWindow.java: In class `javax.swing.JWindow':
> /home/cygnus/dnovillo/perf/sbox/gcc/local.x86_64/src/libjava/javax/swing/JWindow.java: In constructor `()':
> /home/cygnus/dnovillo/perf/sbox/gcc/local.x86_64/src/libjava/javax/swing/JWindow.java:83: error: Can't access package-private constructor `java.awt.Window.<init>' from `javax.swing.JWindow'.
> super();
> ^
>
> We also fail to build 252.eon at -O3. We die with a segmentation fault
> in tree-cfg.c:find_taken_edge
>
> Program received signal SIGSEGV, Segmentation fault.
> tree_int_cst_lt (t1=0x0, t2=0xb3391858)
> at /home/cygnus/dnovillo/perf/sbox/gcc/local.i686/src/gcc/tree.c:3497
> 3497 if (TYPE_UNSIGNED (TREE_TYPE (t1)) != TYPE_UNSIGNED (TREE_TYPE (t2)))
> (gdb) up
> #1 0x0833abce in tree_int_cst_compare (t1=0x0, t2=0xb3391858)
> at /home/cygnus/dnovillo/perf/sbox/gcc/local.i686/src/gcc/tree.c:3521
> 3521 if (tree_int_cst_lt (t1, t2))
> (gdb) up
> #2 0x080f6431 in find_taken_edge (bb=0xb37b9e0c, val=0xb3391858)
> at /home/cygnus/dnovillo/perf/sbox/gcc/local.i686/src/gcc/tree-cfg.c:2052
> 2052 cmp = tree_int_cst_compare (CASE_LOW (t), val);
>
>
> Steven, this failure seems related to your changes. I will send you a
> .ii file in a separate message.
I am seeing same failure on i686 and ia64 testers too, but not on my
home machine, so it is not architeture related, but something else.
I am checking if reverting of my ehreturn patch fixes it (it has been
bootstrapped about week and half ago, so things might have changed)
Honza
>
>
> Thanks. Diego.