[commit] off-by-one max exponent computation in convert_doublest_to_floatformat

[commit] off-by-one max exponent computation in convert_doublest_to_floatformat

[Joel Brobecker]

Hello,

This is a followup on a patch I checked in a while ago.

Assuming the following variable definition:

    long double inp = 2.0;

On platforms where "long double" is a double precision IEEE floating
point, GDB currently behaves as follow:

    (gdb) set variable inp = 1.6e+308l
    (gdb) p inp
    $2 = inf   <<<<---- !!!!

Instead, the value of "inp" should be printed as:

    (gdb) p inp
    $1 = 1.6e+308

The problem is due to a small error in the comparison of the exponent
versus the maximum value this exponent can take, causing us to think
that the value was too big to fit. But it isn't.

gdb/ChangeLog:

        * doublest.c (convert_doublest_to_floatformat): Fix comparison
        against maximum exponent value.

gdb/testsuite/ChangeLog:

        * gdb.base/ldbl_e308.c, gdb.base/ldbl_e308.exp: New files.

Tested on x86_64-linux. Checked in.

---
 gdb/ChangeLog                        |    5 +++++
 gdb/doublest.c                       |    2 +-
 gdb/testsuite/ChangeLog              |    4 ++++
 gdb/testsuite/gdb.base/ldbl_e308.c   |   26 ++++++++++++++++++++++++++
 gdb/testsuite/gdb.base/ldbl_e308.exp |   31 +++++++++++++++++++++++++++++++
 5 files changed, 67 insertions(+), 1 deletion(-)
 create mode 100644 gdb/testsuite/gdb.base/ldbl_e308.c
 create mode 100644 gdb/testsuite/gdb.base/ldbl_e308.exp

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index 4d3c3ec..20ab7a9 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,5 +1,10 @@
 2012-10-24  Joel Brobecker  <brobecker@adacore.com>
 
+	* doublest.c (convert_doublest_to_floatformat): Fix comparison
+	against maximum exponent value.
+
+2012-10-24  Joel Brobecker  <brobecker@adacore.com>
+
 	* ada-lang.h (ADA_KNOWN_RUNTIME_FILE_NAME_PATTERNS): Add entry for
 	"unwind-seh.c".
 
diff --git a/gdb/doublest.c b/gdb/doublest.c
index f683002..0308dac 100644
--- a/gdb/doublest.c
+++ b/gdb/doublest.c
@@ -483,7 +483,7 @@ convert_doublest_to_floatformat (CONST struct floatformat *fmt,
       goto finalize_byteorder;
     }
 
-  if (exponent + fmt->exp_bias >= (1 << fmt->exp_len) - 1)
+  if (exponent + fmt->exp_bias >= (1 << fmt->exp_len))
     {
       /* The value is too large to fit into the destination.
 	 Treat as infinity.  */
diff --git a/gdb/testsuite/ChangeLog b/gdb/testsuite/ChangeLog
index a7a1180..633c0ab 100644
--- a/gdb/testsuite/ChangeLog
+++ b/gdb/testsuite/ChangeLog
@@ -1,5 +1,9 @@
 2012-10-24  Joel Brobecker  <brobecker@adacore.com>
 
+	* gdb.base/ldbl_e308.c, gdb.base/ldbl_e308.exp: New files.
+
+2012-10-24  Joel Brobecker  <brobecker@adacore.com>
+
 	* gdb.ada/unc_arr_ptr_in_var_rec: New testcase.
 
 2012-10-24  Mark Kettenis  <kettenis@gnu.org>
diff --git a/gdb/testsuite/gdb.base/ldbl_e308.c b/gdb/testsuite/gdb.base/ldbl_e308.c
new file mode 100644
index 0000000..8bbe01b
--- /dev/null
+++ b/gdb/testsuite/gdb.base/ldbl_e308.c
@@ -0,0 +1,26 @@
+/* Copyright (C) 2012 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+long double ldbl_308 = 2.0;
+
+int
+main (void)
+{
+  ldbl_308 = ldbl_308 / 2;
+
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.base/ldbl_e308.exp b/gdb/testsuite/gdb.base/ldbl_e308.exp
new file mode 100644
index 0000000..89c54b5
--- /dev/null
+++ b/gdb/testsuite/gdb.base/ldbl_e308.exp
@@ -0,0 +1,31 @@
+# Copyright (C) 2012 Free Software Foundation, Inc.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+set testfile ldbl_e308
+set srcfile ${testfile}.c
+if { [prepare_for_testing ${testfile}.exp ${testfile} ${srcfile}] } {
+    return -1
+}
+
+if ![runto_main] {
+    untested ${testfile}.exp
+    return -1
+}
+
+gdb_test_no_output "set variable ldbl_308 = 1.6e+308l"
+
+gdb_test "print ldbl_308" \
+         "= (1\..*e\\+308|inf)"
+
-- 
1.7.9.5

Re: [commit] off-by-one max exponent computation in convert_doublest_to_floatformat

[Andreas Schwab]

Joel Brobecker <brobecker@adacore.com> writes:

> +gdb_test_no_output "set variable ldbl_308 = 1.6e+308l"
> +
> +gdb_test "print ldbl_308" \
> +         "= (1\..*e\\+308|inf)"

You are still allowing the bug to happen.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [commit] off-by-one max exponent computation in convert_doublest_to_floatformat

[Joel Brobecker]

[-- Attachment #1: Type: text/plain, Size: 455 bytes --]

On Thu, Oct 25, 2012 at 01:54:58AM +0200, Andreas Schwab wrote:
> You are still allowing the bug to happen.

Hmmm, yes, thank you. I was trying to handle the case where sizeof
long double is less than 8, but obviously, that's not the right
approach. I have applied the following patch that just removes "inf"
from the list of valid outputs. We'll make the output target-specific
if we find that we have to.

Tested on x86_64-linux.  Checked in.

-- 
Joel

[-- Attachment #2: 0001-gdb.base-ldbl_e308.exp-Do-not-allow-inf-in-value-out.patch --]
[-- Type: text/x-diff, Size: 1064 bytes --]

From e48d36c1af06cff4a602bc475c4acecc157b32dd Mon Sep 17 00:00:00 2001
From: Joel Brobecker <brobecker@adacore.com>
Date: Thu, 25 Oct 2012 07:11:32 -0400
Subject: [PATCH] gdb.base/ldbl_e308.exp: Do not allow "inf" in value output.

That was an attempt at handling the targets where sizeof(long double)
is less than 8, but the way it was implement allows the bug that
this testcase verifies to come back without being noticed.

gdb/testsuite/ChangeLog:

        * gdb.base/ldbl_e308.exp: Remove "inf" from possible expected
        output for "print inp" test.
---
 gdb/testsuite/gdb.base/ldbl_e308.exp |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/gdb/testsuite/gdb.base/ldbl_e308.exp b/gdb/testsuite/gdb.base/ldbl_e308.exp
index 89c54b5..ad9c5f4 100644
--- a/gdb/testsuite/gdb.base/ldbl_e308.exp
+++ b/gdb/testsuite/gdb.base/ldbl_e308.exp
@@ -27,5 +27,5 @@ if ![runto_main] {
 gdb_test_no_output "set variable ldbl_308 = 1.6e+308l"
 
 gdb_test "print ldbl_308" \
-         "= (1\..*e\\+308|inf)"
+         "= 1\..*e\\+308"
 
-- 
1.7.9.5