[Bug fortran/103766] New: [12 Regression] Initialization of variable passed via static chain is lost.

[Bug fortran/103766] New: [12 Regression] Initialization of variable passed via static chain is lost.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

            Bug ID: 103766
           Summary: [12 Regression] Initialization of variable passed via
                    static chain is lost.
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: iains at gcc dot gnu.org
  Target Milestone: ---

In the following code (by FX)

program crash
  implicit none
  real :: a
  a = 1.
  call sub()
contains
  subroutine sub()
    print *, a
  end subroutine sub
end program crash

when we build with O0, we get the correct result
for O >= 1 we get a random number

What appears to be happening is that the crash () routine is inlined into main
() but the code that initialises the frame (including the 'a' variable) is
dropped which means that sub() sees and prints an uninitialised var.

[Bug fortran/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |10.3.0, 11.1.0
   Target Milestone|---                         |12.0
             Target|                            |x86_64-pc-linux-gnu,
                   |                            |x86_64-apple-darwin,
                   |                            |aarch64-apple-darwin
           Keywords|                            |wrong-code

[Bug fortran/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #1 from Iain Sandoe <iains at gcc dot gnu.org> ---
example optimised tree dump:
O0

;; Function sub (sub.0, funcdef_no=1, decl_uid=2979, cgraph_uid=1,
symbol_order=0)

__attribute__((fn spec (". ")))
void sub ()
{
  struct __st_parameter_dt dt_parm.0;
  real(kind=4) * a.2_1;

  <bb 2> :
  dt_parm.0.common.filename = &"t.f90"[1]{lb: 1 sz: 1};
  dt_parm.0.common.line = 8;
  dt_parm.0.common.flags = 128;
  dt_parm.0.common.unit = 6;
  _gfortran_st_write (&dt_parm.0);
  a.2_1 = &CHAIN.4_8(D)->a;
  _gfortran_transfer_real_write (&dt_parm.0, a.2_1, 4);
  _gfortran_st_write_done (&dt_parm.0);
  dt_parm.0 ={v} {CLOBBER};
  return;

}

;; Function crash (MAIN__, funcdef_no=0, decl_uid=2977, cgraph_uid=2,
symbol_order=1)

__attribute__((fn spec (". ")))
void crash ()
{
  struct FRAME.crash FRAME.3;
  real(kind=4) a [value-expr: FRAME.3.a];
  real(kind=4) D.2991;
  real(kind=4) _3;

  <bb 2> :
  FRAME.3.FRAME_BASE.PARENT = 0B;
  _3 = 1.0e+0;
  FRAME.3.a = _3;
  sub (); [static-chain: &FRAME.3]
  GIMPLE_NOP
  return;

}

;; Function main (main, funcdef_no=2, decl_uid=2983, cgraph_uid=3,
symbol_order=2)

__attribute__((externally_visible))
integer(kind=4) main (integer(kind=4) argc, character(kind=1) * * argv)
{
  static integer(kind=4) options.1[7] = {2116, 4095, 0, 1, 1, 0, 31};
  integer(kind=4) D.2993;
  integer(kind=4) _7;

  <bb 2> :
  _gfortran_set_args (argc_2(D), argv_3(D));
  _gfortran_set_options (7, &options.1[0]);
  crash ();
  _7 = 0;

  <bb 3> :
<L0>:
  return _7;

}

O1

;; Function sub (sub.0, funcdef_no=1, decl_uid=2979, cgraph_uid=1,
symbol_order=0) (executed once)

__attribute__((fn spec (". ")))
void sub ()
{
  struct __st_parameter_dt dt_parm.0;
  real(kind=4) * a.2_1;

  <bb 2> [local count: 1073741824]:
  dt_parm.0.common.filename = &"t.f90"[1]{lb: 1 sz: 1};
  dt_parm.0.common.line = 8;
  dt_parm.0.common.flags = 128;
  dt_parm.0.common.unit = 6;
  _gfortran_st_write (&dt_parm.0);
  a.2_1 = &CHAIN.4_8(D)->a;
  _gfortran_transfer_real_write (&dt_parm.0, a.2_1, 4);
  _gfortran_st_write_done (&dt_parm.0);
  dt_parm.0 ={v} {CLOBBER};
  return;

}

;; Function main (main, funcdef_no=2, decl_uid=2983, cgraph_uid=3,
symbol_order=2) (executed once)

__attribute__((externally_visible))
integer(kind=4) main (integer(kind=4) argc, character(kind=1) * * argv)
{
  struct FRAME.crash FRAME.3;
  static integer(kind=4) options.1[7] = {2116, 4095, 0, 1, 1, 0, 31};

  <bb 2> [local count: 1073741824]:
  _gfortran_set_args (argc_2(D), argv_3(D));
  _gfortran_set_options (7, &options.1[0]);
  sub (); [static-chain: &FRAME.3]
  return 0;
}

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I tried to reproduce it with the following GNU C code but it still
worked:struct __st_parameter_dt
{
  const char *filename ;
  int linenum;
  int flags;
  int unit;
};
[[gnu::noinline, gnu::access(read_only, 1), gnu::access(read_only, 2, 3),
gnu::noclone, gnu::leaf]]
void print(struct __st_parameter_dt *b, void *a, int);
[[gnu::noinline, gnu::access(read_only, 1), gnu::noclone, gnu::leaf]]
void print1(struct __st_parameter_dt *b);
[[gnu::noinline, gnu::access(read_only, 1), gnu::noclone, gnu::leaf]]
void print2(struct __st_parameter_dt *b);

int main(void)
{
  double a;
  a = 1;
  [[gnu::noinline]]
  void f()
  {
      struct __st_parameter_dt b;
      b.filename = "a.c";
      b.linenum = 20;
      b.flags =128;
      b.unit = 1;
      print1(&b);
      print(&b, &a, sizeof(a));
      print2(&b);
  }
  f();
}

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-12-19
             Status|UNCONFIRMED                 |NEW

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
-fno-ipa-modref fixes the issue though. 

I can't figure out why the store to the frame is being removed.

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
                 CC|                            |kargl at gcc dot gnu.org

--- Comment #4 from kargl at gcc dot gnu.org ---
Host association is fundamental to Fortran.
This has worked for 20 years.  Looks like a P1
issue.

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #5 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
There seems to be wrong no_direct_read flag on the static chain parameter
comming out of the ipa propagation.  I will take a look.

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
This is yet another stupid early exit in the propagation code (next stage1 i
guess I will want to go through them and make it more systematic - those
originate from quite early versions of modref when it was easier to rule out
cases where propagation is no longer useful)

I would welcome a testuite friendly version of the fortran testcase
diff --git a/gcc/ipa-modref.c b/gcc/ipa-modref.c
index d3590f0b62b..7152d66c251 100644
--- a/gcc/ipa-modref.c
+++ b/gcc/ipa-modref.c
@@ -5020,8 +5020,14 @@ modref_merge_call_site_flags (escape_summary *sum,
   bool ignore_stores = ignore_stores_p (caller, callee_ecf_flags);

   /* If we have no useful info to propagate.  */
-  if ((!cur_summary || !cur_summary->arg_flags.length ())
-      && (!cur_summary_lto || !cur_summary_lto->arg_flags.length ()))
+  if ((!cur_summary
+       || (!cur_summary->arg_flags.length ()
+          && !cur_summary->static_chain_flags
+          && !cur_summary->retslot_flags))
+      && (!cur_summary_lto
+         || (!cur_summary_lto->arg_flags.length ()
+             && !cur_summary_lto->static_chain_flags
+             && !cur_summary_lto->retslot_flags)))
     return false;

   FOR_EACH_VEC_ELT (sum->esc, i, ee)

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #7 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Jan Hubicka from comment #6)

> I would welcome a testuite friendly version of the fortran testcase

Both Andrew and I failed to make a C reproducer - what about just taking the
-fdump-tree-gimple, as input would that work ?

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[hubicka at kam dot mff.cuni.cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #8 from hubicka at kam dot mff.cuni.cz ---
> > I would welcome a testuite friendly version of the fortran testcase
> 
> Both Andrew and I failed to make a C reproducer - what about just taking the
> -fdump-tree-gimple, as input would that work ?
Me too.  The testcase depends on fnspec annotations produced by Fortran
so we get both noescape and noread to trigger the wrong code.  I guess
we may try to feed fnspec to gimple frontend but I am not even sure it
is possible, since fnspec is internal attribute....

I suppose we can also live without a testcase - I hope I will not do
again such a stupid oversight.

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jan Hubicka <hubicka@gcc.gnu.org>:

https://gcc.gnu.org/g:fcbf94a5be9e0c1ecad92da773a6632b86b7f70a

commit r12-6067-gfcbf94a5be9e0c1ecad92da773a6632b86b7f70a
Author: Jan Hubicka <jh@suse.cz>
Date:   Sun Dec 19 22:28:40 2021 +0100

    Fix early exit in modref_merge_call_site_flags

    When adding support for static chain and return slot flags I forgot to
update
    early exit condition in modref_merge_call_site_flags.  This yields to wrong
    code as demonstrated by the Fortran testcase attached to PR (which I hope
    someone will help me to turn into testuite one).

    gcc/ChangeLog:

    2021-12-19  Jan Hubicka  <hubicka@ucw.cz>

            PR ipa/103766
            * ipa-modref.c (modref_merge_call_site_flags): Fix early exit
condition

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #10 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Fixed.

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Iain Sandoe from comment #7)
> (In reply to Jan Hubicka from comment #6)
> 
> > I would welcome a testuite friendly version of the fortran testcase
> 
> Both Andrew and I failed to make a C reproducer - what about just taking the
> -fdump-tree-gimple, as input would that work ?

Even trying to find a Fortran testsuite friendly version is hard because the
issue can only happen with print, I tried even doing internal write to a string
it is passing without any changes.

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[hubicka at kam dot mff.cuni.cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

--- Comment #12 from hubicka at kam dot mff.cuni.cz ---
> Even trying to find a Fortran testsuite friendly version is hard because the
> issue can only happen with print, I tried even doing internal write to a string
> it is passing without any changes.
Yep, any reads that happen in the code visible to compiler will prevent
the bug from happening.   It is a hard case - but as I said I think
there is quite low chance of it going back.

[Bug ipa/103766] [12 Regression] Initialization of variable passed via static chain is lost.

[fxcoudert at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103766

Francois-Xavier Coudert <fxcoudert at gcc dot gnu.org> changed:

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

--- Comment #13 from Francois-Xavier Coudert <fxcoudert at gcc dot gnu.org> ---
I think this is a Fortran reproducer that does not depend on printing to
stdout:

program crash
  implicit none
  real :: a
  a = 1.
  call sub()
contains
  subroutine sub()
    character(len=100) :: s
    real :: x

    write(s,*) a
    read(s,*) x
    if (x /= 1.) stop 1
  end subroutine sub
end program crash


It fails for me with trunk (before patch), but passed with 11.2.0
Could it be added to the testsuite please?