[Bug fortran/102366] New: [10/11/12 Regression] Illegal instruction with large arrays

[Bug fortran/102366] New: [10/11/12 Regression] Illegal instruction with large arrays

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

            Bug ID: 102366
           Summary: [10/11/12 Regression] Illegal instruction with large
                    arrays
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dominiq at lps dot ens.fr
  Target Milestone: ---

The following test

REAL(KIND=4) :: a(16776325), s
a=1.0_8
END

gives at run time

Illegal instruction

a(16775301) to a(16776324) gives

Segmentation fault

and below a(16776323) the code run as expected.

This occurred between r12-3046 (OK) and r12-3430 and r10-10049 (OK) and
r10-10122. It also affects r11-8981.

Note that

REAL(KIND=4) :: a(16776325), s
a(16776325)=1.0_8
END

compiles and runs witout problem.

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

Iain Sandoe <iains at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |x86_64-darwin,
                   |                            |x86_64-linux-gnu
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2021-09-16
     Ever confirmed|0                           |1
           Keywords|                            |wrong-code
                 CC|                            |iains at gcc dot gnu.org

--- Comment #1 from Iain Sandoe <iains at gcc dot gnu.org> ---
also fails on x86_64 linux (with a segv rather than a ill).

Linux:

(gdb) r
Starting program: /home/iains/t 

Program received signal SIGSEGV, Segmentation fault.
0x000000000040118c in MAIN__ () at dom1.f90:2
2       a=1.0_8


Darwin:

Process 14098 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS
(code=1, address=0x7ffeebbff900)
    frame #0: 0x0000000100000ec2 t`MAIN__ at dom1.f90:2:7
   1    REAL(KIND=4) :: a(16,1024,1024), s
-> 2    a=1.0_8
   3    END
Target 0: (t) stopped.

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #2 from anlauf at gcc dot gnu.org ---
What is your stack size?

Does it help if you declare a  SAVEd?

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #3 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> What is your stack size?

65532 kbytes

> Does it help if you declare a  SAVEd?

The illegal instruction is gone.

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

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

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

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
And you are trying to use 65532 kbytes long array on the stack, leaving no
stack space for anything else.  Clearly user error.

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #5 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #4)
> And you are trying to use 65532 kbytes long array on the stack, leaving no
> stack space for anything else.  Clearly user error.

out of curiosity, why did this work previously? (was the array previously
allocated on the heap?)

we are not being very helpful to the user who made the error (neither Linux nor
Darwin give any stack overflow), but I guess that's hard to arrange robustly.

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

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

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

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Seems it changed with r12-3129-gf95946afd160e2a1f4beac4ee5e6d5633307f39a

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #7 from anlauf at gcc dot gnu.org ---
(In reply to Jakub Jelinek from comment #6)
> Seems it changed with r12-3129-gf95946afd160e2a1f4beac4ee5e6d5633307f39a

Looking at the tree dump, it appears that there is a latent issue.


void MAIN__ ()
{
  real(kind=4) a[16776325];


However, that standard says:

F2018  8.5.16  SAVE attribute

(4) A variable, common block, or procedure pointer declared in the scoping unit
of a main program, [...] implicitly has the SAVE attribute


My interpretations of the issue is that we miss to set the SAVE_IMPLICIT,
which is confirmed by running under the debugger, which shows that a
has SAVE_NONE.

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #8 from anlauf at gcc dot gnu.org ---
Potential fix for comment#0:

diff --git a/gcc/fortran/trans-decl.c b/gcc/fortran/trans-decl.c
index bed61e2325d..54309646aad 100644
--- a/gcc/fortran/trans-decl.c
+++ b/gcc/fortran/trans-decl.c
@@ -709,6 +709,11 @@ gfc_finish_var_decl (tree decl, gfc_symbol * sym)
        }
     }

+  /* Implicit SAVE (F2018, 8.5.16 SAVE attribute, clause 4).  */
+  if (sym->ns->proc_name && sym->ns->proc_name->attr.is_main_program
+      && sym->attr.save == SAVE_NONE)
+    sym->attr.save = SAVE_IMPLICIT;
+
   /* Derived types are a bit peculiar because of the possibility of
      a default initializer; this must be applied each time the variable
      comes into scope it therefore need not be static.  These variables

Works for me, but not regtested yet.

Does that help?

Of course that handles only part of that clause, and we need extended
testcases.

[Bug fortran/102366] [10/11/12 Regression] Illegal instruction with large arrays

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #9 from anlauf at gcc dot gnu.org ---
(In reply to anlauf from comment #8)
> Potential fix for comment#0:

I'm getting many regressions for this change.  Investigating.

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[10/11/12 Regression]       |[10/11/12 Regression] large
                   |Illegal instruction with    |arrays no longer become
                   |large arrays                |static
   Target Milestone|---                         |10.4

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #10 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> Seems it changed with r12-3129-gf95946afd160e2a1f4beac4ee5e6d5633307f39a

The problem is gone if I revert r12-3129.

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #11 from anlauf at gcc dot gnu.org ---
(In reply to Dominique d'Humieres from comment #10)

> The problem is gone if I revert r12-3129.

But then it regresses on pr98411.  See for yourself compiling with -Wall.

Something like

diff --git a/gcc/fortran/trans-decl.c b/gcc/fortran/trans-decl.c
index bed61e2325d..b022d818e82 100644
--- a/gcc/fortran/trans-decl.c
+++ b/gcc/fortran/trans-decl.c
@@ -743,7 +743,6 @@ gfc_finish_var_decl (tree decl, gfc_symbol * sym)

   /* Keep variables larger than max-stack-var-size off stack.  */
   if (!(sym->ns->proc_name && sym->ns->proc_name->attr.recursive)
-      && !(sym->ns->proc_name && sym->ns->proc_name->attr.is_main_program)
       && !sym->attr.automatic
       && sym->attr.save != SAVE_EXPLICIT
       && sym->attr.save != SAVE_IMPLICIT
@@ -757,7 +756,9 @@ gfc_finish_var_decl (tree decl, gfc_symbol * sym)
          || sym->attr.allocatable)
       && !DECL_ARTIFICIAL (decl))
     {
-      if (flag_max_stack_var_size > 0)
+      if (flag_max_stack_var_size > 0
+         && !(sym->ns->proc_name &&
+              sym->ns->proc_name->attr.is_main_program))
        gfc_warning (OPT_Wsurprising,
                     "Array %qs at %L is larger than limit set by "
                     "%<-fmax-stack-var-size=%>, moved from stack to static "

disables the warning for arrays in the main.

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
That LGTM, except formatting - && shouldn't be at the end of line per coding
conventions, but on the start of next line.

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #13 from anlauf at gcc dot gnu.org ---
In decl.c:match_attr_spec we have:

5818      /* Since Fortran 2008 module variables implicitly have the SAVE
attribute.  */
5819      if ((gfc_current_state () == COMP_MODULE
5820           || gfc_current_state () == COMP_SUBMODULE)
5821          && !current_attr.save
5822          && (gfc_option.allow_std & GFC_STD_F2008) != 0)
5823        current_attr.save = SAVE_IMPLICIT;

I tried to extend that to the case of COMP_PROGRAM, but:

(gdb) p gfc_state_stack->state 
$162 = COMP_NONE

But when I add

program test

at the beginning of the file, I find

(gdb) p gfc_state_stack->state 
$164 = COMP_PROGRAM

So this may be too early to handle arrays in the main.

Given Jakub's comment, I'll fix that and regtest.

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #14 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Harald Anlauf <anlauf@gcc.gnu.org>:

https://gcc.gnu.org/g:51166eb2c534692c3c7779def24f83c8c3811b98

commit r12-3639-g51166eb2c534692c3c7779def24f83c8c3811b98
Author: Harald Anlauf <anlauf@gmx.de>
Date:   Fri Sep 17 21:45:33 2021 +0200

    Fortran - (large) arrays in the main shall be static

    gcc/fortran/ChangeLog:

            PR fortran/102366
            * trans-decl.c (gfc_finish_var_decl): Disable the warning message
            for variables moved from stack to static storange if they are
            declared in the main, but allow the move to happen.

    gcc/testsuite/ChangeLog:

            PR fortran/102366
            * gfortran.dg/pr102366.f90: New test.

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

anlauf at gcc dot gnu.org changed:

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

--- Comment #15 from anlauf at gcc dot gnu.org ---
Committed as per Jakub's LGTM and announced here:

https://gcc.gnu.org/pipermail/fortran/2021-September/056540.html

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #16 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-11 branch has been updated by Harald Anlauf
<anlauf@gcc.gnu.org>:

https://gcc.gnu.org/g:7fbab3afca910c055676ebc566bf87c4d5a7372f

commit r11-9020-g7fbab3afca910c055676ebc566bf87c4d5a7372f
Author: Harald Anlauf <anlauf@gmx.de>
Date:   Fri Sep 17 21:45:33 2021 +0200

    Fortran - (large) arrays in the main shall be static

    gcc/fortran/ChangeLog:

            PR fortran/102366
            * trans-decl.c (gfc_finish_var_decl): Disable the warning message
            for variables moved from stack to static storange if they are
            declared in the main, but allow the move to happen.

    gcc/testsuite/ChangeLog:

            PR fortran/102366
            * gfortran.dg/pr102366.f90: New test.

    (cherry picked from commit 51166eb2c534692c3c7779def24f83c8c3811b98)

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

--- Comment #17 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Harald Anlauf
<anlauf@gcc.gnu.org>:

https://gcc.gnu.org/g:d3db5b1d23cbe4e4569e688a6dbc8b5b2c38588e

commit r10-10140-gd3db5b1d23cbe4e4569e688a6dbc8b5b2c38588e
Author: Harald Anlauf <anlauf@gmx.de>
Date:   Fri Sep 17 21:45:33 2021 +0200

    Fortran - (large) arrays in the main shall be static

    gcc/fortran/ChangeLog:

            PR fortran/102366
            * trans-decl.c (gfc_finish_var_decl): Disable the warning message
            for variables moved from stack to static storange if they are
            declared in the main, but allow the move to happen.

    gcc/testsuite/ChangeLog:

            PR fortran/102366
            * gfortran.dg/pr102366.f90: New test.

    (cherry picked from commit 51166eb2c534692c3c7779def24f83c8c3811b98)

[Bug fortran/102366] [10/11/12 Regression] large arrays no longer become static

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102366

anlauf at gcc dot gnu.org changed:

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

--- Comment #18 from anlauf at gcc dot gnu.org ---
Fixed on all affected branches.  Closing.

Thanks for the report!