[Bug tree-optimization/65818] New: libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[Bug tree-optimization/65818] New: libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[danglin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

            Bug ID: 65818
           Summary: libiberty/vprintf-support.c:41:1: ICE: in expand_i
                    fn_va_arg_1, at tree-stdarg.c:1095
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: danglin at gcc dot gnu.org
              Host: hppa2.0w-hp-hpux11.11
            Target: hppa2.0w-hp-hpux11.11
             Build: hppa2.0w-hp-hpux11.11

if [ x"-fPIC" != x ]; then \
          /test/gnu/gcc/objdir/./prev-gcc/xgcc
-B/test/gnu/gcc/objdir/./prev-gcc
/ -B/opt/gnu/gcc/gcc-6.0/hppa2.0w-hp-hpux11.11/bin/
-B/opt/gnu/gcc/gcc-6.0/hppa2
.0w-hp-hpux11.11/bin/ -B/opt/gnu/gcc/gcc-6.0/hppa2.0w-hp-hpux11.11/lib/
-isystem
 /opt/gnu/gcc/gcc-6.0/hppa2.0w-hp-hpux11.11/include -isystem
/opt/gnu/gcc/gcc-6.
0/hppa2.0w-hp-hpux11.11/sys-include    -c -DHAVE_CONFIG_H -g -O2  -I.
-I../../gc
c/libiberty/../include  -W -Wall -Wwrite-strings -Wc++-compat
-Wstrict-prototype
s -pedantic  -fPIC ../../gcc/libiberty/vprintf-support.c -o
pic/vprintf-support.
o; \
        else true; fi
../../gcc/libiberty/vprintf-support.c: In function
'libiberty_vprintf_buffer_siz
e':
../../gcc/libiberty/vprintf-support.c:41:1: internal compiler error: in
expand_i
fn_va_arg_1, at tree-stdarg.c:1095
 libiberty_vprintf_buffer_size (const char *format, va_list args)
 ^

[Bug tree-optimization/65818] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[danglin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #1 from John David Anglin <danglin at gcc dot gnu.org> ---
Created attachment 35371
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35371&action=edit
Preprocessed source

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |build
                 CC|                            |tom at codesourcery dot com
   Target Milestone|---                         |6.0
            Summary|libiberty/vprintf-support.c |[6 Regression]
                   |:41:1: ICE: in expand_i     |libiberty/vprintf-support.c
                   |fn_va_arg_1, at             |:41:1: ICE: in expand_i
                   |tree-stdarg.c:1095          |fn_va_arg_1, at
                   |                            |tree-stdarg.c:1095

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

vries at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-04-21
                 CC|                            |vries at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from vries at gcc dot gnu.org ---
Reproduced.

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #3 from vries at gcc dot gnu.org ---
The code snippet:
...
     case 'f':
     case 'e':
     case 'E':
     case 'g':
     case 'G':
       (void) __builtin_va_arg(ap, double);
       total_width += 307;
       break;
...

translates to:
...
<L23>:
  VA_ARG (&ap, 0B);
  total_width_70 = total_width_31 + 337;
  # DEBUG total_width => total_width_70
  goto <bb 21> (<L31>);
...

at the end of hppa_gimplify_va_arg_expr, the result looks like:
...
(gdb) p *pre_p
$2 = (gimple_seq) 0x0
(gdb) p *post_p
$3 = (gimple_seq) 0x0
(gdb) call debug_generic_expr (t)
*(double *) (ap = ap + 4294967288 & 4294967288B)
...

Note that pre_p and post_p are empty, and t contains the side-effect of
assigning to ap.

Since ifn_va_arg has no lhs, the returned value is not assigned to any var in
expand_ifn_va_arg_1, and the side-effect is lost.

We need to further gimplify the expression to fix the error. [ I wonder whether
that's the responsibility of TARGET_GIMPLIFY_VA_ARG_EXPR. ]

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

vries at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |vries at gcc dot gnu.org

--- Comment #4 from vries at gcc dot gnu.org ---
Created attachment 35383
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35383&action=edit
Tentative patch

This patch seems to fix the error:
...
diff --git a/gcc/gimplify.c b/gcc/gimplify.c
index c68bd47..5f1dd1a 100644
--- a/gcc/gimplify.c
+++ b/gcc/gimplify.c
@@ -9352,7 +9352,9 @@ gimplify_va_arg_internal (tree valist, tree type,
location_t loc,
   else
     gimplify_expr (&valist, pre_p, post_p, is_gimple_min_lval, fb_lvalue);

-  return targetm.gimplify_va_arg_expr (valist, type, pre_p, post_p);
+  tree expr = targetm.gimplify_va_arg_expr (valist, type, pre_p, post_p);
+  gimplify_expr (&expr, pre_p, post_p, is_gimple_val, fb_rvalue);
+  return expr;
 }

 /* Gimplify __builtin_va_arg, aka VA_ARG_EXPR, which is not really a
...

Before gimplify_expr:
...
(gdb) call debug_generic_expr (expr)
*(double *) (ap = ap + 4294967288 & 4294967288B)
...

And after gimplify_expr:
...
(gdb) call debug_generic_expr (expr)
D.3458
(gdb) call debug_gimple_seq (*pre_p)
ap.13 = ap;
D.3456 = ap.13 + 4294967288;
ap.14 = D.3456 & 4294967288B;
ap = ap.14;
ap.13 = ap;
D.3458 = MEM[(double *)ap.13];
...

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #5 from vries at gcc dot gnu.org ---
Dave,

could you please test the patch on hppa?

Thanks,
- Tom

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[dave.anglin at bell dot net]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #6 from dave.anglin at bell dot net ---
On 2015-04-22, at 3:15 AM, vries at gcc dot gnu.org wrote:

> could you please test the patch on hppa?

Started.  Thanks for the patch.

Dave
--
John David Anglin    dave.anglin@bell.net

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #7 from vries at gcc dot gnu.org ---
Hmm, this patch has failures for x86_64:

- va_arg (ap, void):
  gcc.c-torture/compile/pr48767.c 
- vla:
  gcc.c-torture/execute/20020412-1.c
  gcc.dg/compat/struct-by-value-22
  gcc.dg/lto/20090706-1
  gcc.target/i386/memcpy-strategy-4.c

this patch is more conservative:
...
diff --git a/gcc/gimplify.c b/gcc/gimplify.c
index 7786e16..e23f510 100644
--- a/gcc/gimplify.c
+++ b/gcc/gimplify.c
@@ -9336,7 +9336,12 @@ gimplify_va_arg_internal (tree valist, tree type,
location_t loc,
   else
     gimplify_expr (&valist, pre_p, post_p, is_gimple_min_lval, fb_lvalue);

-  return targetm.gimplify_va_arg_expr (valist, type, pre_p, post_p);
+  tree expr = targetm.gimplify_va_arg_expr (valist, type, pre_p, post_p);
+  gcc_assert (TREE_CODE (expr) == MEM_REF);
+  if (!is_gimple_mem_ref_addr (TREE_OPERAND (expr, 0)))
+    gimplify_expr (&TREE_OPERAND (expr, 0), pre_p, post_p,
+                  is_gimple_mem_ref_addr, fb_rvalue);
+  return expr;
 }

 /* Gimplify __builtin_va_arg, aka VA_ARG_EXPR, which is not really a
...

It allows at least pr48767.c, 20020412-1.c and memcpy-strategy-4.c to pass.

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[dave.anglin at bell dot net]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #8 from dave.anglin at bell dot net ---
On 2015-04-22 12:40 PM, vries at gcc dot gnu.org wrote:
> this patch is more conservative
Restarted test with updated patch.

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[dave.anglin at bell dot net]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #9 from dave.anglin at bell dot net ---
On 2015-04-22, at 12:40 PM, vries at gcc dot gnu.org wrote:

> this patch is more conservative:
> ...
> diff --git a/gcc/gimplify.c b/gcc/gimplify.c
> index 7786e16..e23f510 100644
> --- a/gcc/gimplify.c
> +++ b/gcc/gimplify.c
> @@ -9336,7 +9336,12 @@ gimplify_va_arg_internal (tree valist, tree type,
> location_t loc,
>   else
>     gimplify_expr (&valist, pre_p, post_p, is_gimple_min_lval, fb_lvalue);
> 
> -  return targetm.gimplify_va_arg_expr (valist, type, pre_p, post_p);
> +  tree expr = targetm.gimplify_va_arg_expr (valist, type, pre_p, post_p);
> +  gcc_assert (TREE_CODE (expr) == MEM_REF);
> +  if (!is_gimple_mem_ref_addr (TREE_OPERAND (expr, 0)))
> +    gimplify_expr (&TREE_OPERAND (expr, 0), pre_p, post_p,
> +                  is_gimple_mem_ref_addr, fb_rvalue);
> +  return expr;
> }
> 
> /* Gimplify __builtin_va_arg, aka VA_ARG_EXPR, which is not really a

This patch has the following new failure:

/test/gnu/gcc/gcc/gcc/testsuite/g++.dg/pr65802.C: In function 'void fn1()':
/test/gnu/gcc/gcc/gcc/testsuite/g++.dg/pr65802.C:20:23: error: impossible
constraint in 'asm'

FAIL: g++.dg/pr65802.C  -std=gnu++98 (test for excess errors)
FAIL: g++.dg/pr65802.C  -std=gnu++11 (test for excess errors)
FAIL: g++.dg/pr65802.C  -std=gnu++14 (test for excess errors)

--
John David Anglin    dave.anglin@bell.net

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #10 from vries at gcc dot gnu.org ---
(In reply to dave.anglin from comment #9)
> This patch has the following new failure:
> 
> /test/gnu/gcc/gcc/gcc/testsuite/g++.dg/pr65802.C: In function 'void fn1()':
> /test/gnu/gcc/gcc/gcc/testsuite/g++.dg/pr65802.C:20:23: error: impossible
> constraint in 'asm'
> 
> FAIL: g++.dg/pr65802.C  -std=gnu++98 (test for excess errors)
> FAIL: g++.dg/pr65802.C  -std=gnu++11 (test for excess errors)
> FAIL: g++.dg/pr65802.C  -std=gnu++14 (test for excess errors)
> 

That failure was probably already present without the patch.  (see f.i.
https://gcc.gnu.org/ml/gcc-patches/2015-04/msg01482.html ).

I've rewritten the test-case in C and cleaned it up. It's now called
gcc.dg/pr65802.c.

I tested the new test-case with and without the updated patch, it passes in
both cases.

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #11 from vries at gcc dot gnu.org ---
Bootstrap and reg-test on x86_64 also went ok for for the updated patch. I'll
submit.

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

vries at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch

--- Comment #12 from vries at gcc dot gnu.org ---
https://gcc.gnu.org/ml/gcc-patches/2015-04/msg01543.html

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #13 from vries at gcc dot gnu.org ---
*** Bug 65824 has been marked as a duplicate of this bug. ***

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

--- Comment #14 from vries at gcc dot gnu.org ---
Author: vries
Date: Tue Apr 28 06:05:14 2015
New Revision: 222509

URL: https://gcc.gnu.org/viewcvs?rev=222509&root=gcc&view=rev
Log:
Evaluate side-effects in expand_ifn_va_arg_1

2015-04-28  Tom de Vries  <tom@codesourcery.com>

        PR tree-optimization/65818
        * tree-stdarg.c (expand_ifn_va_arg_1): Ensure that side-effects are
        evaluated.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-stdarg.c

[Bug tree-optimization/65818] [6 Regression] libiberty/vprintf-support.c:41:1: ICE: in expand_i fn_va_arg_1, at tree-stdarg.c:1095

[vries at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65818

vries at gcc dot gnu.org changed:

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

--- Comment #15 from vries at gcc dot gnu.org ---
patch committed, marking resolved fixed