[Bug c/20277] New: -mcpu=power4 vs. -maltivec
Since 3.4.x, you can't use both -mcpu=power4 and -maltivec on ppc targets. This
is critical that it's fixed before 4.0 as the linux kernel will rely on this.
It's currently using -mcpu=970 as a workaround since it has one file using
-maltivec (the RAID6 code), but this workaround isn't suitable for 4.0 due to
implicit generation of vector code.
--
Summary: -mcpu=power4 vs. -maltivec
Product: gcc
Version: unknown
Status: UNCONFIRMED
Severity: critical
Priority: P2
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: benh at kernel dot crashing dot org
CC: gcc-bugs at gcc dot gnu dot org
GCC host triplet: powerpc64-linux-gcc
GCC target triplet: powerpc64-linux-gcc
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20277
[Bug c/21457] New: ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913
So I get the above ICE (gen_subprogram_die at dwarf2out.c:10913) with debian's 3.4.4 shapshot (20050314). I'm currently building CVS top of branch to double check. It also happens with an older 20041116 build I have here. -- Summary: ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913 Product: gcc Version: 3.4.4 Status: UNCONFIRMED Severity: normal Priority: P2 Component: c AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: benh at kernel dot crashing dot org CC: gcc-bugs at gcc dot gnu dot org GCC build triplet: powerpc-linux-gcc GCC host triplet: powerpc-linux-gcc GCC target triplet: powerpc-linux-gcc http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21457
[Bug c/21457] ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913
--- Additional Comments From benh at kernel dot crashing dot org 2005-05-08 22:46 --- Created an attachment (id=8841) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8841&action=view) Here is the .i file -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21457
[Bug c/21457] ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913
--- Additional Comments From benh at kernel dot crashing dot org 2005-05-08 22:47 --- Created an attachment (id=8842) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8842&action=view) And the half-generated .s file in case it's useful too -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21457
[Bug c/21457] ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913
--- Additional Comments From benh at kernel dot crashing dot org 2005-05-08 22:48 --- Oh, and sorry if I got the triplets wrong, I can never remember what are the "proper" ones ... -- What|Removed |Added Summary|ICE building glibc 2.3.5|ICE building glibc 2.3.5 |nptl in gen_subprogram_die |nptl in gen_subprogram_die |at dwarf2out.c:10913|at dwarf2out.c:10913 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21457
[Bug c/21457] ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913
--- Additional Comments From benh at kernel dot crashing dot org
2005-05-08 23:19 ---
Ok, this is with a CVS checkout of today, without the debian patches (that
changes the line number a bit). I now get:
../elf/tst-execstack-mod.c:24: internal compiler error: in gen_subprogram_die,
at dwarf2out.c:10887
The program is simple:
/* Test module for making nonexecutable stacks executable
on load of a DSO that requires executable stacks. */
#include
#include
#include
void callme (void (*callback) (void));
/* This is a function that makes use of executable stack by
using a local function trampoline. */
void
tryme (void)
{
bool ok = false;
void callback (void) { ok = true; }
callme (&callback);
if (ok)
printf ("DSO called ok (local %p, trampoline %p)\n", &ok, &callback);
else
abort ();
}
void
callme (void (*callback) (void))
{
(*callback) ();
}
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21457
[Bug c/21457] ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913
--- Additional Comments From benh at kernel dot crashing dot org 2005-05-08 23:23 --- Hrm... smells a bit like 16676 ... It really needs to be fixed asap, it prevents building glibc in it's current state. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21457
[Bug c/21457] ICE building glibc 2.3.5 nptl in gen_subprogram_die at dwarf2out.c:10913
--- Additional Comments From benh at kernel dot crashing dot org 2005-05-08 23:39 --- Same goes for iconvconfig.c:1244 in glibc too, so at least now 2 occurences of this bug preventing build of glibc. This one is a nested function too, so I suppose that is the real issue. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21457
[Bug debug/16676] [3.4 Regression] ICE with nested functions and -g1, blocks glibc
--- Additional Comments From benh at kernel dot crashing dot org 2005-05-09 00:29 --- Note that the use of -g1 on glibc NPTL build is a debian rule, I don't know if anybody else does that and I don't know why they do that, some obscure comment seem to imply it is to get unwind informations. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16676
[Bug c/16676] [3.4 Regression] ICE with nested functions and -g1, blocks glibc
--- Additional Comments From benh at kernel dot crashing dot org 2005-05-09 07:11 --- Ben Elliston just produced a patch for it that I tested. It fixed building of glibc on debian powerpc with -g1 (used by debian rules for nptl). The patch is on it's way to the patch list (which didn't catch up yet it seems). The patch is very simple, it just causes gcc to emit more debug infos in the case of nested functions (like the -g2 case) instead of trying to be smart. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16676
