[Bug fortran/58058] New: Memory leak with transfer function

[Bug fortran/58058] New: Memory leak with transfer function

[thomas.jourdan at orange dot fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

            Bug ID: 58058
           Summary: Memory leak with transfer function
           Product: gcc
           Version: 4.8.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: thomas.jourdan at orange dot fr

Hello,

Using gfortran 4.8.1, the following code runs fine but according to valgrind it
produces a memory leak:

program test1

  implicit none

  integer, dimension(3) :: t1, t2
  character(len=64) :: str

  t1 = (/1,2,3/)

  !str = transfer((/1,2,3/),str) ! works
  str = transfer(t1,str) ! memory leak
  t2 = transfer(str,t1)

  write(*,*) 't2 = ',t2

end program test1

The output is:
==7005== Memcheck, a memory error detector
==7005== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al.
==7005== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
==7005== Command: ./test1
==7005== 
 t2 =            1           2           3
==7005== 
==7005== HEAP SUMMARY:
==7005==     in use at exit: 64 bytes in 1 blocks
==7005==   total heap usage: 23 allocs, 22 frees, 11,897 bytes allocated
==7005== 
==7005== LEAK SUMMARY:
==7005==    definitely lost: 64 bytes in 1 blocks
==7005==    indirectly lost: 0 bytes in 0 blocks
==7005==      possibly lost: 0 bytes in 0 blocks
==7005==    still reachable: 0 bytes in 0 blocks
==7005==         suppressed: 0 bytes in 0 blocks
==7005== Rerun with --leak-check=full to see details of leaked memory
==7005== 
==7005== For counts of detected and suppressed errors, rerun with: -v
==7005== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 2 from 2)


If I use directly "str = transfer((/1,2,3/),str)" instead, no memory problem
occurs. I experienced the same problem with the following test:


program test2

  implicit none

  type tuple
    integer :: nn
    integer :: pp
    integer :: qq
  end type tuple

  character(len=64) :: str
  type(tuple) :: mt1, mt2

  mt1%nn = 1
  mt1%pp = 2
  mt1%qq = 3

  str = transfer(mt1,str)
  mt2 = transfer(str,mt2)

end program test2

Thanks,

Thomas

[Bug fortran/58058] [4.7/4.8/4.9 Regression] Memory leak with transfer function

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

janus at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-08-02
                 CC|                            |janus at gcc dot gnu.org,
                   |                            |pault at gcc dot gnu.org
            Summary|Memory leak with transfer   |[4.7/4.8/4.9 Regression]
                   |function                    |Memory leak with transfer
                   |                            |function
     Ever confirmed|0                           |1

--- Comment #1 from janus at gcc dot gnu.org ---
I can confirm the leak with gfortran versions 4.7, 4.8 and trunk. However,
there is no leak in 4.3, so it is a regression (haven't tried other versions).

Indeed the dump produced by -fdump-tree-original shows a __builtin_malloc,
which is never freed. The regression seems to be due to this commit:

http://gcc.gnu.org/viewcvs/gcc?view=revision&revision=143462

[Bug fortran/58058] [4.7/4.8/4.9 Regression] Memory leak with transfer function

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

--- Comment #2 from janus at gcc dot gnu.org ---
I think the following should be enough to fix it:


Index: gcc/fortran/trans-intrinsic.c
===================================================================
--- gcc/fortran/trans-intrinsic.c    (revision 201331)
+++ gcc/fortran/trans-intrinsic.c    (working copy)
@@ -5652,8 +5652,7 @@ scalar_transfer:

   if (expr->ts.type == BT_CHARACTER)
     {
-      tree direct;
-      tree indirect;
+      tree direct, indirect, free;

       ptr = convert (gfc_get_pchar_type (expr->ts.kind), source);
       tmpdecl = gfc_create_var (gfc_get_pchar_type (expr->ts.kind),
@@ -5686,6 +5685,13 @@ scalar_transfer:
       tmp = build3_v (COND_EXPR, tmp, direct, indirect);
       gfc_add_expr_to_block (&se->pre, tmp);

+      /* Free the temporary string, if necessary.  */
+      free = gfc_call_free (tmpdecl);
+      tmp = fold_build2_loc (input_location, GT_EXPR, boolean_type_node,
+                 dest_word_len, source_bytes);
+      tmp = build3_v (COND_EXPR, tmp, free, build_empty_stmt
(input_location));
+      gfc_add_expr_to_block (&se->post, tmp);
+
       se->expr = tmpdecl;
       se->string_length = fold_convert (gfc_charlen_type_node, dest_word_len);
     }

[Bug fortran/58058] [4.7/4.8/4.9 Regression] Memory leak with transfer function

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

janus at gcc dot gnu.org changed:

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

--- Comment #3 from janus at gcc dot gnu.org ---
(In reply to janus from comment #2)
> I think the following should be enough to fix it:

It also regtests cleanly.

[Bug fortran/58058] [4.7/4.8/4.9 Regression] Memory leak with transfer function

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

--- Comment #4 from janus at gcc dot gnu.org ---
Fixed on trunk with this commit:


Author: janus
Date: Fri Aug  9 16:57:58 2013
New Revision: 201633

URL: http://gcc.gnu.org/viewcvs?rev=201633&root=gcc&view=rev
Log:
2013-08-09  Janus Weil  <janus@gcc.gnu.org>

    PR fortran/58058
    * trans-intrinsic.c (gfc_conv_intrinsic_transfer): Free the temporary
    string, if necessary.

2013-08-09  Janus Weil  <janus@gcc.gnu.org>

    PR fortran/58058
    * gfortran.dg/transfer_intrinsic_6.f90: New.

Added:
    trunk/gcc/testsuite/gfortran.dg/transfer_intrinsic_6.f90
Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/trans-intrinsic.c
    trunk/gcc/testsuite/ChangeLog

[Bug fortran/58058] [4.7/4.8/4.9 Regression] Memory leak with transfer function

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

--- Comment #5 from janus at gcc dot gnu.org ---
Fixed on 4.8 with:

Author: janus
Date: Sun Aug 11 11:31:41 2013
New Revision: 201652

URL: http://gcc.gnu.org/viewcvs?rev=201652&root=gcc&view=rev
Log:
2013-08-11  Janus Weil  <janus@gcc.gnu.org>

    Backport from trunk:
    2013-08-09  Janus Weil  <janus@gcc.gnu.org>

    PR fortran/58058
    * trans-intrinsic.c (gfc_conv_intrinsic_transfer): Free the temporary
    string, if necessary.

2013-08-11  Janus Weil  <janus@gcc.gnu.org>

    Backport from trunk:
    2013-08-09  Janus Weil  <janus@gcc.gnu.org>

    PR fortran/58058
    * gfortran.dg/transfer_intrinsic_6.f90: New.

Added:
    branches/gcc-4_8-branch/gcc/testsuite/gfortran.dg/transfer_intrinsic_6.f90
Modified:
    branches/gcc-4_8-branch/gcc/fortran/ChangeLog
    branches/gcc-4_8-branch/gcc/fortran/trans-intrinsic.c
    branches/gcc-4_8-branch/gcc/testsuite/ChangeLog

[Bug fortran/58058] [4.7/4.8/4.9 Regression] Memory leak with transfer function

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

--- Comment #6 from janus at gcc dot gnu.org ---
Fixed on 4.7 with:


Author: janus
Date: Sun Aug 11 14:08:12 2013
New Revision: 201653

URL: http://gcc.gnu.org/viewcvs?rev=201653&root=gcc&view=rev
Log:
2013-08-11  Janus Weil  <janus@gcc.gnu.org>

    Backport from trunk:
    2013-08-09  Janus Weil  <janus@gcc.gnu.org>

    PR fortran/58058
    * trans-intrinsic.c (gfc_conv_intrinsic_transfer): Free the temporary
    string, if necessary.

2013-08-11  Janus Weil  <janus@gcc.gnu.org>

    Backport from trunk:
    2013-08-09  Janus Weil  <janus@gcc.gnu.org>

    PR fortran/58058
    * gfortran.dg/transfer_intrinsic_6.f90: New.

Added:
    branches/gcc-4_7-branch/gcc/testsuite/gfortran.dg/transfer_intrinsic_6.f90
Modified:
    branches/gcc-4_7-branch/gcc/fortran/ChangeLog
    branches/gcc-4_7-branch/gcc/fortran/trans-intrinsic.c
    branches/gcc-4_7-branch/gcc/testsuite/ChangeLog

[Bug fortran/58058] [4.7/4.8/4.9 Regression] Memory leak with transfer function

[janus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58058

janus at gcc dot gnu.org changed:

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

--- Comment #7 from janus at gcc dot gnu.org ---
Fixed on all active branches. Closing.

Thanks for the report!