[Bug debug/65521] New: [5 Regression] nondeterministic -fcompare-debug failures

[Bug debug/65521] New: [5 Regression] nondeterministic -fcompare-debug failures

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

            Bug ID: 65521
           Summary: [5 Regression] nondeterministic -fcompare-debug
                    failures
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: debug
          Assignee: unassigned at gcc dot gnu.org
          Reporter: trippels at gcc dot gnu.org
              Host: powerpc64le-unknown-linux-gnu
            Target: powerpc64le-unknown-linux-gnu
             Build: powerpc64le-unknown-linux-gnu

on ppc64le I get nondeterministic -fcompare-debug failures during Linux
kernel build, e.g:

 % gcc -c -fcompare-debug -std=gnu89 -O2 ipmi_watchdog.i
gcc: error: ipmi_watchdog.i: -fcompare-debug failure
 % gcc -c -fcompare-debug -std=gnu89 -O2 ipmi_watchdog.i
 % gcc -c -fcompare-debug -std=gnu89 -O2 ipmi_watchdog.i
gcc: error: ipmi_watchdog.i: -fcompare-debug failure

 % diff -u ipmi_watchdog.gkd ipmi_watchdog.gk.gkd
--- ipmi_watchdog.gkd   2015-03-23 07:51:04.510095107 +0000
+++ ipmi_watchdog.gk.gkd        2015-03-23 07:51:05.100104842 +0000
@@ -90,7 +90,7 @@
 Declarations used by panic_smi_free, sorted by DECL_UID:
 0:   static void panic_recv_free (struct ipmi_recv_msg *);

-;; Function panic_smi_free (panic_smi_free, funcdef_no=1412, cgraph_uid=1422,
symbol_order=1513)
+;; Function panic_smi_free (panic_smi_free, funcdef_no=1408, cgraph_uid=1422,
symbol_order=1513)

 (note # 0 0 NOTE_INSN_DELETED)
 (note # 0 0 [bb 2] NOTE_INSN_BASIC_BLOCK)
@@ -531,7 +531,7 @@
 Declarations used by heartbeat_free_smi, sorted by DECL_UID:
 0:   static void heartbeat_free_recv (struct ipmi_recv_msg *);

-;; Function heartbeat_free_smi (heartbeat_free_smi, funcdef_no=1408,
cgraph_uid=1426, symbol_order=1526)
+;; Function heartbeat_free_smi (heartbeat_free_smi, funcdef_no=1412,
cgraph_uid=1426, symbol_order=1526)

 (note # 0 0 NOTE_INSN_DELETED)
 (note # 0 0 [bb 2] NOTE_INSN_BASIC_BLOCK)

[Bug debug/65521] [5 Regression] nondeterministic -fcompare-debug failures

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

--- Comment #1 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
Created attachment 35102
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35102&action=edit
unreduced testcase

[Bug debug/65521] [5 Regression] nondeterministic -fcompare-debug failures

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

--- Comment #2 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
Created attachment 35104
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35104&action=edit
testcase for x86_64

Also happens on x86_64:

 % gcc -c -fcompare-debug -std=gnu89 -O2 hooks.i
gcc: error: hooks.i: -fcompare-debug failure

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-03-23
                 CC|                            |marxin at gcc dot gnu.org
          Component|debug                       |ipa
   Target Milestone|---                         |5.0
     Ever confirmed|0                           |1

--- Comment #3 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
I suspect r221576 is responsible for this issue, because -fno-ipa-icf "fixes"
the problem.

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

--- Comment #4 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
struct socket {
  int sk;
};
selinux_task_getioprio(struct task_struct *p) {
  return current_has_perm(p, 256UL);
}
selinux_task_getscheduler(struct task_struct *p) {
  return current_has_perm(p, 256UL);
}
selinux_socket_getsockname(struct socket *sock) {
  return sock_has_perm(get_current(), &sock->sk, 16UL);
}
selinux_socket_getpeername(struct socket *sock) {
  return sock_has_perm(get_current(), &sock->sk, 16UL);
}
selinux_ops() { security_module_enable(selinux_ops); }

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Feels like hashing of pointers and walking hash-tables where ordering changes
code-generation (well, DECL_UID in this case).  That's a no-no, of course
(think address-space randomization).

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
But even hashing DECL_UIDs rather than hashing pointers is a no-go here,
pointers can be arbitrary, for DECL_UIDs we at least have the guarantee that -g
and -g0 DECL_UIDs compare the same, but there can be bigger gaps in between
them for -g.  Dunno if symtab_node's order field can be used reliably for this
or not (or it could hash ->name () string).

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 35105
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35105&action=edit
gcc5-pr65521.patch

This fixes this for me.  But I haven't done any in-depth analysis whether the
symtab order fields are always identical between -g and -g0.

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Though, judging from the fact that we print symtab_node::order into the
-fdump-final-insns= dumps:
;; Function selinux_ops (selinux_ops, funcdef_no=4, cgraph_uid=4,
symbol_order=4)
it should be ok (otherwise we'd have -fcompare-debug failures already).
For functions we could also use node->uid (either just that, or hash in both),
but we don't have those in the varpool.

So, I think I'll just add the testcase into the testsuite and bootstrap/regtest
this.

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #35105|0                           |1
        is obsolete|                            |
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 35106
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35106&action=edit
gcc5-pr65521.patch

Patch I'm going to bootstrap/regtest.

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Mon Mar 23 15:17:20 2015
New Revision: 221596

URL: https://gcc.gnu.org/viewcvs?rev=221596&root=gcc&view=rev
Log:
    PR ipa/65521
    * ipa-icf.c (sem_item::update_hash_by_addr_refs): Hash
    ultimate_alias_target ()->order ints instead of
    ultimate_alias_target () pointers.

    * gcc.dg/pr65521.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/pr65521.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ipa-icf.c
    trunk/gcc/testsuite/ChangeLog

[Bug ipa/65521] [5 Regression] nondeterministic -fcompare-debug failures

[trippels at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65521

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #11 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
Fixed. Thanks!