Le 08/10/2026 à 14:32, Paul Richard Thomas a écrit :
bug4-dealloc-scope-exit.f90 fails in runtime with:
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
#0 0x7f2b83809803 in ???
#1 0x7f2b838083b5 in ???
#2 0x7f2b834dcfaf in ???
#3 0x40066f in n1
at /home/pault/prs/pdt_conformance/reducers/bug4-dealloc-scope-exit.f90:14
#4 0x4006c9 in main
at /home/pault/prs/pdt_conformance/reducers/bug4-dealloc-scope-exit.f90:14
Segmentation fault (core dumped) ./a.out
The patch is straightforward enough as not to need explanation.
Passes regtesting on FC44/x86_64. OK for mainline.
Cheers
Paul
bug4.patch
From b308d107d4ac32e348e2d728764e3910af9c3a92 Mon Sep 17 00:00:00 2001
From: Paul Thomas <[email protected]>
Date: Thu, 8 Oct 2026 09:58:52 +0100
Subject: [PATCH] Fortran: [meta-bug] [PDT] Parameterized derived type errors
[PR82173]
2026-10-06 Paul Thomas <[email protected]>
gcc/fortran
PR fortran/82173
* trans-decl.cc (gfc_trans_deferred_vars): Small formatting
adjustment followed by a guarding of the pdt component dealloc
for allocatable variables.
gcc/testsuite/
PR fortran/82173
* gfortran.dg/pdt_bug4.f90: New test.
---
gcc/fortran/trans-decl.cc | 16 ++++++++++++++++
gcc/testsuite/gfortran.dg/pdt_bug4.f90 | 14 ++++++++++++++
2 files changed, 30 insertions(+)
create mode 100644 gcc/testsuite/gfortran.dg/pdt_bug4.f90
diff --git a/gcc/fortran/trans-decl.cc b/gcc/fortran/trans-decl.cc
index a614e71eced..0229cc7ab82 100644
--- a/gcc/fortran/trans-decl.cc
+++ b/gcc/fortran/trans-decl.cc
@@ -5179,6 +5179,22 @@ gfc_trans_deferred_vars (gfc_symbol * proc_sym,
gfc_wrapped_block * block)
else
tmp = NULL_TREE;
+ if (sym->attr.allocatable && !sym->attr.result
+ && !GFC_DESCRIPTOR_TYPE_P (TREE_TYPE (sym->backend_decl)))
+ {
+ /* Only deallocate the PDT comonents if the variable has not
+ been deallocated. Arrays are guarded elsewhere. */
+ tree null_cond, zero;
+ null_cond = sym->backend_decl;
+ zero = build_int_cst (TREE_TYPE (null_cond), 0);
+ null_cond = fold_build2_loc (input_location, NE_EXPR,
+ logical_type_node,
+ null_cond, zero);
+ tmp = build3_loc (input_location, COND_EXPR, void_type_node,
+ null_cond, tmp,
+ build_empty_stmt (input_location));
+ }
+
From the context, it seems that `tmp' can be NULL_TREE. So please add
a check excluding that case (unless there is something that already
provides that guarantee).
OK with that change.
gfc_add_init_cleanup (block, gfc_finish_block (&tmpblock), tmp);
}
else if (sym->attr.dummy)
