[PATCH] PR fortran/104127 - [9/10/11/12 Regression] ICE in get_array_charlen, at fortran/trans-array.c:7244

[PATCH] PR fortran/104127 - [9/10/11/12 Regression] ICE in get_array_charlen, at fortran/trans-array.c:7244

[Harald Anlauf]

[-- Attachment #1: Type: text/plain, Size: 738 bytes --]

Dear Fortranners,

when simplifying TRANSFER with a MOLD argument of type character
and with SIZE=0 we lose the character length.

This happens in all gfortran versions and results in wrong code.
The purported regression is that at some point in the 9-development
this lead to a (previously possibly latent) ICE.

The attached patch sets up the character length in the typespec and
fixes the ICE.  There is another generic hidden/latent problem with
array constructors of size 0 passed to procedures (see e.g. pr86277)
which will remain and is beyond the scope of this fix.

Regtested on x86_64-pc-linux-gnu.  I also fixed a minor logic bug
in testcase transfer_simplify_11.f90.

OK for mainline?  Backports to branches?

Thanks,
Harald


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-Fortran-fix-simplification-of-TRANSFER-for-zero-size.patch --]
[-- Type: text/x-patch, Size: 4412 bytes --]

From c9882ace6199e2a327b69449f825e0366b442cba Mon Sep 17 00:00:00 2001
From: Harald Anlauf <anlauf@gmx.de>
Date: Thu, 20 Jan 2022 22:36:50 +0100
Subject: [PATCH] Fortran: fix simplification of TRANSFER for zero-sized
 character array result

gcc/fortran/ChangeLog:

	PR fortran/104127
	* simplify.cc (gfc_simplify_transfer): Ensure that the result
	typespec is set up for TRANSFER with MOLD of type CHARACTER
	including character length even if the result is a zero-sized
	array.

gcc/testsuite/ChangeLog:

	PR fortran/104127
	* gfortran.dg/transfer_simplify_11.f90: Fix logic.
	* gfortran.dg/transfer_simplify_13.f90: New test.
---
 gcc/fortran/simplify.cc                       | 13 ++++++-
 .../gfortran.dg/transfer_simplify_11.f90      |  2 +-
 .../gfortran.dg/transfer_simplify_13.f90      | 34 +++++++++++++++++++
 3 files changed, 47 insertions(+), 2 deletions(-)
 create mode 100644 gcc/testsuite/gfortran.dg/transfer_simplify_13.f90

diff --git a/gcc/fortran/simplify.cc b/gcc/fortran/simplify.cc
index 3881370d947..8604162cfd5 100644
--- a/gcc/fortran/simplify.cc
+++ b/gcc/fortran/simplify.cc
@@ -8162,7 +8162,18 @@ gfc_simplify_transfer (gfc_expr *source, gfc_expr *mold, gfc_expr *size)
      set even for array expressions, in order to pass this information into
      gfc_target_interpret_expr.  */
   if (result->ts.type == BT_CHARACTER && gfc_is_constant_expr (mold_element))
-    result->value.character.length = mold_element->value.character.length;
+    {
+      result->value.character.length = mold_element->value.character.length;
+
+      /* Let the typespec of the result inherit the string length.
+	 This is crucial if a resulting array has size zero.  */
+      if (mold_element->ts.u.cl->length)
+	result->ts.u.cl->length = gfc_copy_expr (mold_element->ts.u.cl->length);
+      else
+	result->ts.u.cl->length =
+	  gfc_get_int_expr (gfc_charlen_int_kind, NULL,
+			    mold_element->value.character.length);
+    }

   /* Set the number of elements in the result, and determine its size.  */

diff --git a/gcc/testsuite/gfortran.dg/transfer_simplify_11.f90 b/gcc/testsuite/gfortran.dg/transfer_simplify_11.f90
index 0911f9dba3a..409e4768a10 100644
--- a/gcc/testsuite/gfortran.dg/transfer_simplify_11.f90
+++ b/gcc/testsuite/gfortran.dg/transfer_simplify_11.f90
@@ -4,5 +4,5 @@
    integer, parameter :: N = 2
    character(len=1) :: chr(N)
    chr = transfer(repeat("x",ncopies=N),[character(len=1) ::], N)
-   if (chr(1) /= 'x' .and. chr(2) /= 'x') STOP 1
+   if (chr(1) /= 'x' .or. chr(2) /= 'x') STOP 1
 end
diff --git a/gcc/testsuite/gfortran.dg/transfer_simplify_13.f90 b/gcc/testsuite/gfortran.dg/transfer_simplify_13.f90
new file mode 100644
index 00000000000..59109c6029d
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/transfer_simplify_13.f90
@@ -0,0 +1,34 @@
+! { dg-do compile }
+! { dg-options "-fdump-tree-original" }
+! PR fortran/104127 - ICE in get_array_charlen
+! Contributed by G.Steinmetz
+
+program p
+  character(4) :: mold = "XYZ"
+  integer      :: i = 0
+  integer, parameter :: l1 = len  (transfer('ab', 'xyz', size=0))
+  integer, parameter :: s1 = size (transfer('ab', 'xyz', size=0))
+  integer, parameter :: l4 = len  (transfer(4_'abcd', 4_'xy', size=0))
+  integer, parameter :: s4 = size (transfer(4_'abcd', 4_'xy', size=0))
+  integer, parameter :: l2 = len  (transfer('ab', mold,  size=0))
+  integer, parameter :: l3 = len  (transfer('ab', mold,  size=1))
+  integer, parameter :: l5 = len  (transfer('ab',['xyz'], size=0))
+  integer, parameter :: s5 = size (transfer('ab',['xyz'], size=0))
+  call sub0 ( transfer('a', 'y', size=0) )
+  call sub1 ([transfer('a', 'y', size=0)])
+  call sub2 ([transfer('a',['y'],size=0)])
+  call sub3 ( transfer('a', 'y', size=1) )
+  call sub4 ([transfer('a', 'y', size=1)])
+  call sub5 ( transfer('a', 'y', size=i) )
+  call sub6 ( transfer(1_'abcd', 1_'xy' , size=0))
+  call sub7 ( transfer(1_'abcd',[1_'xy'], size=0))
+  call sub8 ( transfer(4_'abcd', 4_'xy' , size=0))
+  call sub9 ( transfer(4_'abcd',[4_'xy'], size=0))
+  print *, transfer('abcd', 'xy', size=0)
+  if (l1 /= 3 .or. s1 /= 0) stop 1
+  if (l4 /= 2 .or. s4 /= 0) stop 2
+  if (l2 /= 4 .or. l3 /= 4) stop 3
+  if (l5 /= 3 .or. s5 /= 0) stop 1
+end
+
+! { dg-final { scan-tree-dump-not "_gfortran_stop_numeric" "original" } }
--
2.31.1

Re: [PATCH] PR fortran/104127 - [9/10/11/12 Regression] ICE in get_array_charlen, at fortran/trans-array.c:7244

[Thomas Koenig]

Hello Harald,

> when simplifying TRANSFER with a MOLD argument of type character
> and with SIZE=0 we lose the character length.
> 
> This happens in all gfortran versions and results in wrong code.
> The purported regression is that at some point in the 9-development
> this lead to a (previously possibly latent) ICE.
> 
> The attached patch sets up the character length in the typespec and
> fixes the ICE.  There is another generic hidden/latent problem with
> array constructors of size 0 passed to procedures (see e.g. pr86277)
> which will remain and is beyond the scope of this fix.
> 
> Regtested on x86_64-pc-linux-gnu.  I also fixed a minor logic bug
> in testcase transfer_simplify_11.f90.
> 
> OK for mainline?  Backports to branches?

OK for both.

Thanks for the patch!

Regards

	Thomas