public inbox for fortran@gcc.gnu.org
 help / color / mirror / Atom feed
From: Andre Vehreschild <vehre@gmx.de>
To: GCC-Fortran-ML <fortran@gcc.gnu.org>,
	GCC-Patches-ML <gcc-patches@gcc.gnu.org>
Cc: Damian Rouson <damian@sourceryinstitute.org>,
	Katherine Rasmussen <krasmussen@lbl.gov>,
	Brad Richardson <brad@archaeologic.codes>
Subject: [Fortran, Patch, Coarray, PR 37336] Fix crash in finalizer when derived type coarray is already freed.
Date: Thu, 28 Sep 2023 14:45:05 +0200	[thread overview]
Message-ID: <20230928144505.77e03a8e@vepi2> (raw)

[-- Attachment #1: Type: text/plain, Size: 489 bytes --]

Hi all,

attached patch fixes a crash in coarray programs when an allocatable derived
typed coarray was freed explicitly. The generated cleanup code did not take
into account, that the coarray may have been deallocated already. The patch
fixes this by moving the statements accessing components inside the derived type
into the block guard by its allocated check.

Regtested ok on f37/x86_64. Ok for master?

Regards,
	Andre
--
Andre Vehreschild * Email: vehre ad gmx dot de

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: caf_finalizer_fix.patch --]
[-- Type: text/x-patch, Size: 2601 bytes --]

diff --git a/gcc/fortran/trans-array.cc b/gcc/fortran/trans-array.cc
index e0fc8ebc46b..8e94a9a469f 100644
--- a/gcc/fortran/trans-array.cc
+++ b/gcc/fortran/trans-array.cc
@@ -9320,6 +9320,12 @@ structure_alloc_comps (gfc_symbol * der_type, tree decl, tree dest,
       gfc_add_expr_to_block (&fnblock, tmp);
     }

+  /* Still having a descriptor array of rank == 0 here, indicates an
+     allocatable coarrays.  Dereference it correctly.  */
+  if (GFC_DESCRIPTOR_TYPE_P (decl_type))
+    {
+      decl = build_fold_indirect_ref (gfc_conv_array_data (decl));
+    }
   /* Otherwise, act on the components or recursively call self to
      act on a chain of components.  */
   for (c = der_type->components; c; c = c->next)
@@ -11507,7 +11513,11 @@ gfc_trans_deferred_array (gfc_symbol * sym, gfc_wrapped_block * block)
     {
       int rank;
       rank = sym->as ? sym->as->rank : 0;
-      tmp = gfc_deallocate_alloc_comp (sym->ts.u.derived, descriptor, rank);
+      tmp = gfc_deallocate_alloc_comp (sym->ts.u.derived, descriptor, rank,
+				       (sym->attr.codimension
+					&& flag_coarray == GFC_FCOARRAY_LIB)
+				       ? GFC_STRUCTURE_CAF_MODE_IN_COARRAY
+				       : 0);
       gfc_add_expr_to_block (&cleanup, tmp);
     }

@@ -11521,9 +11531,11 @@ gfc_trans_deferred_array (gfc_symbol * sym, gfc_wrapped_block * block)
 					NULL_TREE, NULL_TREE, true, e,
 					sym->attr.codimension
 					? GFC_CAF_COARRAY_DEREGISTER
-					: GFC_CAF_COARRAY_NOCOARRAY);
+					: GFC_CAF_COARRAY_NOCOARRAY,
+					NULL_TREE, gfc_finish_block (&cleanup));
       if (e)
 	gfc_free_expr (e);
+      gfc_init_block (&cleanup);
       gfc_add_expr_to_block (&cleanup, tmp);
     }

diff --git a/gcc/testsuite/gfortran.dg/coarray/alloc_comp_6.f90 b/gcc/testsuite/gfortran.dg/coarray/alloc_comp_6.f90
new file mode 100644
index 00000000000..e8a74db2c18
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/coarray/alloc_comp_6.f90
@@ -0,0 +1,29 @@
+! { dg-do run }
+
+program alloc_comp_6
+
+  implicit none
+
+  type :: foo
+    real :: x
+    integer, allocatable :: y(:)
+  end type
+
+  call check()
+
+contains
+
+  subroutine check()
+    block
+      type(foo), allocatable :: example[:] ! needs to be a coarray
+
+      allocate(example[*])
+      allocate(example%y(10))
+      example%x = 3.4
+      example%y = 4
+
+      deallocate(example)
+    end block  ! example%y shall not be accessed here by the finalizer,
+               ! because example is already deallocated
+  end subroutine check
+end program alloc_comp_6

[-- Attachment #3: caf_finalizer_fix.clog --]
[-- Type: text/plain, Size: 670 bytes --]

    Fortran: Free alloc. comp. in allocated coarrays only.

    When freeing allocatable components of an allocatable coarray, add
    a check that the coarray is still allocated, before accessing the
    components.

    This patch adds to PR fortran/37336, but does not fix it completely.

    gcc/fortran/ChangeLog:
            PR fortran/37336
            * trans-array.cc (structure_alloc_comps): Deref coarray.
            (gfc_trans_deferred_array): Add freeing of components after
            check for allocated coarray.

    gcc/testsuite/ChangeLog:
            PR fortran/37336
            * gfortran.dg/coarray/alloc_comp_6.f90: New test.


             reply	other threads:[~2023-09-28 12:45 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-09-28 12:45 Andre Vehreschild [this message]
2023-09-28 18:21 ` Paul Richard Thomas
2023-09-29 10:01   ` Andre Vehreschild
2023-09-29 12:38     ` Paul Richard Thomas
2023-09-29 13:13       ` Andre Vehreschild
2023-09-30 13:16         ` Andre Vehreschild
2023-10-01 19:38           ` Paul Richard Thomas

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230928144505.77e03a8e@vepi2 \
    --to=vehre@gmx.de \
    --cc=brad@archaeologic.codes \
    --cc=damian@sourceryinstitute.org \
    --cc=fortran@gcc.gnu.org \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=krasmussen@lbl.gov \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).