This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PROPOSED PATCH] Fix integer overflow when adjusting posixrules
- From: Paul Eggert <eggert at cs dot ucla dot edu>
- To: Florian Weimer <fweimer at redhat dot com>, libc-alpha at sourceware dot org
- Date: Wed, 13 Apr 2016 00:41:44 -0700
- Subject: Re: [PROPOSED PATCH] Fix integer overflow when adjusting posixrules
- Authentication-results: sourceware.org; auth=none
- References: <1460317379-31635-1-git-send-email-eggert at cs dot ucla dot edu> <570DD877 dot 1010709 at redhat dot com>
Florian Weimer wrote:
On 04/10/2016 09:42 PM, Paul Eggert wrote:
+/* intprops.h -- properties of integer types
I'm afraid I have serious doubts about the maintainability of this file.
The file has been used in core GNU utilities for a decade or so, so there is
some evidence for maintainability. It can always be improved of course.
> In this case, the argument for the outer cast is not within the range of
type t, so the result of the cast is implementation-defined. For ones'
complement, ~0 can be undefined, it does not have to be 0.
Good catch. That code dates back to 2005, when ones' complement platforms were
still arguably plausible porting targets for GNU code. Unisys stopped selling
ones' complement mainframes late last year, though. ~0 should work on Unisys
mainframes; but really, it's time to remove that gingerbread, if only to make
the code easier to review.
There's a similar issue in time/mktime.c in glibc (also shared with gnulib).
Revised patches attached. The first is an updated version of the tzfile fix, and
the second fixes mktime. The proposed intprops.h and mktime.c files match what's
now in gnulib (which I've already fixed).
>From 967f35b75a43bd215b8d4457d39a6863732bda76 Mon Sep 17 00:00:00 2001
From: Paul Eggert <eggert@cs.ucla.edu>
Date: Sun, 10 Apr 2016 12:30:55 -0700
Subject: [PATCH 1/2] Fix integer overflow when adjusting posixrules
[BZ #19738]
This bug can occur on hosts with 32-bit time_t when reading tz files
generated by tz 2014c or later. Problem reported by Aurelien
Jarno in: http://mm.icann.org/pipermail/tz/2016-March/023321.html
* include/intprops.h: New file, copied verbatim from Gnulib.
This newer version has an LGPLv2.1+ notice and a single year range.
* time/tzfile.c: Include "intprops.h".
(__tzfile_default): Check for time_t overflow when parsing
TZ='PST+8PDT' when the initial transition is at or near the
minimum time_t value.
---
ChangeLog | 14 ++
include/intprops.h | 413 +++++++++++++++++++++++++++++++++++++++++++++++++++++
time/tzfile.c | 35 +++--
3 files changed, 448 insertions(+), 14 deletions(-)
create mode 100644 include/intprops.h
diff --git a/ChangeLog b/ChangeLog
index a32946a..0a6d632 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,17 @@
+2016-04-13 Paul Eggert <eggert@cs.ucla.edu>
+
+ Fix integer overflow when adjusting posixrules
+ [BZ #19738]
+ This bug can occur on hosts with 32-bit time_t when reading tz files
+ generated by tz 2014c or later. Problem reported by Aurelien
+ Jarno in: http://mm.icann.org/pipermail/tz/2016-March/023321.html
+ * include/intprops.h: New file, copied verbatim from Gnulib.
+ This newer version has an LGPLv2.1+ notice and a single year range.
+ * time/tzfile.c: Include "intprops.h".
+ (__tzfile_default): Check for time_t overflow when parsing
+ TZ='PST+8PDT' when the initial transition is at or near the
+ minimum time_t value.
+
2016-04-12 H.J. Lu <hongjiu.lu@intel.com>
[BZ #19928]
diff --git a/include/intprops.h b/include/intprops.h
new file mode 100644
index 0000000..de6c324
--- /dev/null
+++ b/include/intprops.h
@@ -0,0 +1,413 @@
+/* intprops.h -- properties of integer types
+
+ Copyright (C) 2001-2016 Free Software Foundation, Inc.
+
+ This program is free software: you can redistribute it and/or modify it
+ under the terms of the GNU Lesser General Public License as published
+ by the Free Software Foundation; either version 2.1 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 Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+/* Written by Paul Eggert. */
+
+#ifndef _GL_INTPROPS_H
+#define _GL_INTPROPS_H
+
+#include <limits.h>
+
+/* Return a value with the common real type of E and V and the value of V. */
+#define _GL_INT_CONVERT(e, v) (0 * (e) + (v))
+
+/* Act like _GL_INT_CONVERT (E, -V) but work around a bug in IRIX 6.5 cc; see
+ <http://lists.gnu.org/archive/html/bug-gnulib/2011-05/msg00406.html>. */
+#define _GL_INT_NEGATE_CONVERT(e, v) (0 * (e) - (v))
+
+/* The extra casts in the following macros work around compiler bugs,
+ e.g., in Cray C 5.0.3.0. */
+
+/* True if the arithmetic type T is an integer type. bool counts as
+ an integer. */
+#define TYPE_IS_INTEGER(t) ((t) 1.5 == 1)
+
+/* True if the real type T is signed. */
+#define TYPE_SIGNED(t) (! ((t) 0 < (t) -1))
+
+/* Return 1 if the real expression E, after promotion, has a
+ signed or floating type. */
+#define EXPR_SIGNED(e) (_GL_INT_NEGATE_CONVERT (e, 1) < 0)
+
+
+/* Minimum and maximum values for integer types and expressions.
+ These macros have undefined behavior for signed types that either
+ have padding bits or do not use two's complement. If this is a
+ problem for you, please let us know how to fix it for your host. */
+
+/* The maximum and minimum values for the integer type T. */
+#define TYPE_MINIMUM(t) ((t) ~ TYPE_MAXIMUM (t))
+#define TYPE_MAXIMUM(t) \
+ ((t) (! TYPE_SIGNED (t) \
+ ? (t) -1 \
+ : ((((t) 1 << (sizeof (t) * CHAR_BIT - 2)) - 1) * 2 + 1)))
+
+/* The maximum and minimum values for the type of the expression E,
+ after integer promotion. E should not have side effects. */
+#define _GL_INT_MINIMUM(e) \
+ (EXPR_SIGNED (e) \
+ ? ~ _GL_SIGNED_INT_MAXIMUM (e) \
+ : _GL_INT_CONVERT (e, 0))
+#define _GL_INT_MAXIMUM(e) \
+ (EXPR_SIGNED (e) \
+ ? _GL_SIGNED_INT_MAXIMUM (e) \
+ : _GL_INT_NEGATE_CONVERT (e, 1))
+#define _GL_SIGNED_INT_MAXIMUM(e) \
+ (((_GL_INT_CONVERT (e, 1) << (sizeof ((e) + 0) * CHAR_BIT - 2)) - 1) * 2 + 1)
+
+
+/* Return 1 if the __typeof__ keyword works. This could be done by
+ 'configure', but for now it's easier to do it by hand. */
+#if (2 <= __GNUC__ || defined __IBM__TYPEOF__ \
+ || (0x5110 <= __SUNPRO_C && !__STDC__))
+# define _GL_HAVE___TYPEOF__ 1
+#else
+# define _GL_HAVE___TYPEOF__ 0
+#endif
+
+/* Return 1 if the integer type or expression T might be signed. Return 0
+ if it is definitely unsigned. This macro does not evaluate its argument,
+ and expands to an integer constant expression. */
+#if _GL_HAVE___TYPEOF__
+# define _GL_SIGNED_TYPE_OR_EXPR(t) TYPE_SIGNED (__typeof__ (t))
+#else
+# define _GL_SIGNED_TYPE_OR_EXPR(t) 1
+#endif
+
+/* Bound on length of the string representing an unsigned integer
+ value representable in B bits. log10 (2.0) < 146/485. The
+ smallest value of B where this bound is not tight is 2621. */
+#define INT_BITS_STRLEN_BOUND(b) (((b) * 146 + 484) / 485)
+
+/* Bound on length of the string representing an integer type or expression T.
+ Subtract 1 for the sign bit if T is signed, and then add 1 more for
+ a minus sign if needed.
+
+ Because _GL_SIGNED_TYPE_OR_EXPR sometimes returns 0 when its argument is
+ signed, this macro may overestimate the true bound by one byte when
+ applied to unsigned types of size 2, 4, 16, ... bytes. */
+#define INT_STRLEN_BOUND(t) \
+ (INT_BITS_STRLEN_BOUND (sizeof (t) * CHAR_BIT \
+ - _GL_SIGNED_TYPE_OR_EXPR (t)) \
+ + _GL_SIGNED_TYPE_OR_EXPR (t))
+
+/* Bound on buffer size needed to represent an integer type or expression T,
+ including the terminating null. */
+#define INT_BUFSIZE_BOUND(t) (INT_STRLEN_BOUND (t) + 1)
+
+
+/* Range overflow checks.
+
+ The INT_<op>_RANGE_OVERFLOW macros return 1 if the corresponding C
+ operators might not yield numerically correct answers due to
+ arithmetic overflow. They do not rely on undefined or
+ implementation-defined behavior. Their implementations are simple
+ and straightforward, but they are a bit harder to use than the
+ INT_<op>_OVERFLOW macros described below.
+
+ Example usage:
+
+ long int i = ...;
+ long int j = ...;
+ if (INT_MULTIPLY_RANGE_OVERFLOW (i, j, LONG_MIN, LONG_MAX))
+ printf ("multiply would overflow");
+ else
+ printf ("product is %ld", i * j);
+
+ Restrictions on *_RANGE_OVERFLOW macros:
+
+ These macros do not check for all possible numerical problems or
+ undefined or unspecified behavior: they do not check for division
+ by zero, for bad shift counts, or for shifting negative numbers.
+
+ These macros may evaluate their arguments zero or multiple times,
+ so the arguments should not have side effects. The arithmetic
+ arguments (including the MIN and MAX arguments) must be of the same
+ integer type after the usual arithmetic conversions, and the type
+ must have minimum value MIN and maximum MAX. Unsigned types should
+ use a zero MIN of the proper type.
+
+ These macros are tuned for constant MIN and MAX. For commutative
+ operations such as A + B, they are also tuned for constant B. */
+
+/* Return 1 if A + B would overflow in [MIN,MAX] arithmetic.
+ See above for restrictions. */
+#define INT_ADD_RANGE_OVERFLOW(a, b, min, max) \
+ ((b) < 0 \
+ ? (a) < (min) - (b) \
+ : (max) - (b) < (a))
+
+/* Return 1 if A - B would overflow in [MIN,MAX] arithmetic.
+ See above for restrictions. */
+#define INT_SUBTRACT_RANGE_OVERFLOW(a, b, min, max) \
+ ((b) < 0 \
+ ? (max) + (b) < (a) \
+ : (a) < (min) + (b))
+
+/* Return 1 if - A would overflow in [MIN,MAX] arithmetic.
+ See above for restrictions. */
+#define INT_NEGATE_RANGE_OVERFLOW(a, min, max) \
+ ((min) < 0 \
+ ? (a) < - (max) \
+ : 0 < (a))
+
+/* Return 1 if A * B would overflow in [MIN,MAX] arithmetic.
+ See above for restrictions. Avoid && and || as they tickle
+ bugs in Sun C 5.11 2010/08/13 and other compilers; see
+ <http://lists.gnu.org/archive/html/bug-gnulib/2011-05/msg00401.html>. */
+#define INT_MULTIPLY_RANGE_OVERFLOW(a, b, min, max) \
+ ((b) < 0 \
+ ? ((a) < 0 \
+ ? (a) < (max) / (b) \
+ : (b) == -1 \
+ ? 0 \
+ : (min) / (b) < (a)) \
+ : (b) == 0 \
+ ? 0 \
+ : ((a) < 0 \
+ ? (a) < (min) / (b) \
+ : (max) / (b) < (a)))
+
+/* Return 1 if A / B would overflow in [MIN,MAX] arithmetic.
+ See above for restrictions. Do not check for division by zero. */
+#define INT_DIVIDE_RANGE_OVERFLOW(a, b, min, max) \
+ ((min) < 0 && (b) == -1 && (a) < - (max))
+
+/* Return 1 if A % B would overflow in [MIN,MAX] arithmetic.
+ See above for restrictions. Do not check for division by zero.
+ Mathematically, % should never overflow, but on x86-like hosts
+ INT_MIN % -1 traps, and the C standard permits this, so treat this
+ as an overflow too. */
+#define INT_REMAINDER_RANGE_OVERFLOW(a, b, min, max) \
+ INT_DIVIDE_RANGE_OVERFLOW (a, b, min, max)
+
+/* Return 1 if A << B would overflow in [MIN,MAX] arithmetic.
+ See above for restrictions. Here, MIN and MAX are for A only, and B need
+ not be of the same type as the other arguments. The C standard says that
+ behavior is undefined for shifts unless 0 <= B < wordwidth, and that when
+ A is negative then A << B has undefined behavior and A >> B has
+ implementation-defined behavior, but do not check these other
+ restrictions. */
+#define INT_LEFT_SHIFT_RANGE_OVERFLOW(a, b, min, max) \
+ ((a) < 0 \
+ ? (a) < (min) >> (b) \
+ : (max) >> (b) < (a))
+
+
+/* The _GL*_OVERFLOW macros have the same restrictions as the
+ *_RANGE_OVERFLOW macros, except that they do not assume that operands
+ (e.g., A and B) have the same type as MIN and MAX. Instead, they assume
+ that the result (e.g., A + B) has that type. */
+#define _GL_ADD_OVERFLOW(a, b, min, max) \
+ ((min) < 0 ? INT_ADD_RANGE_OVERFLOW (a, b, min, max) \
+ : (a) < 0 ? (b) <= (a) + (b) \
+ : (b) < 0 ? (a) <= (a) + (b) \
+ : (a) + (b) < (b))
+#define _GL_SUBTRACT_OVERFLOW(a, b, min, max) \
+ ((min) < 0 ? INT_SUBTRACT_RANGE_OVERFLOW (a, b, min, max) \
+ : (a) < 0 ? 1 \
+ : (b) < 0 ? (a) - (b) <= (a) \
+ : (a) < (b))
+#define _GL_MULTIPLY_OVERFLOW(a, b, min, max) \
+ (((min) == 0 && (((a) < 0 && 0 < (b)) || ((b) < 0 && 0 < (a)))) \
+ || INT_MULTIPLY_RANGE_OVERFLOW (a, b, min, max))
+#define _GL_DIVIDE_OVERFLOW(a, b, min, max) \
+ ((min) < 0 ? (b) == _GL_INT_NEGATE_CONVERT (min, 1) && (a) < - (max) \
+ : (a) < 0 ? (b) <= (a) + (b) - 1 \
+ : (b) < 0 && (a) + (b) <= (a))
+#define _GL_REMAINDER_OVERFLOW(a, b, min, max) \
+ ((min) < 0 ? (b) == _GL_INT_NEGATE_CONVERT (min, 1) && (a) < - (max) \
+ : (a) < 0 ? (a) % (b) != ((max) - (b) + 1) % (b) \
+ : (b) < 0 && ! _GL_UNSIGNED_NEG_MULTIPLE (a, b, max))
+
+/* Return a nonzero value if A is a mathematical multiple of B, where
+ A is unsigned, B is negative, and MAX is the maximum value of A's
+ type. A's type must be the same as (A % B)'s type. Normally (A %
+ -B == 0) suffices, but things get tricky if -B would overflow. */
+#define _GL_UNSIGNED_NEG_MULTIPLE(a, b, max) \
+ (((b) < -_GL_SIGNED_INT_MAXIMUM (b) \
+ ? (_GL_SIGNED_INT_MAXIMUM (b) == (max) \
+ ? (a) \
+ : (a) % (_GL_INT_CONVERT (a, _GL_SIGNED_INT_MAXIMUM (b)) + 1)) \
+ : (a) % - (b)) \
+ == 0)
+
+/* Check for integer overflow, and report low order bits of answer.
+
+ The INT_<op>_OVERFLOW macros return 1 if the corresponding C operators
+ might not yield numerically correct answers due to arithmetic overflow.
+ The INT_<op>_WRAPV macros also store the low-order bits of the answer.
+ These macros work correctly on all known practical hosts, and do not rely
+ on undefined behavior due to signed arithmetic overflow.
+
+ Example usage, assuming A and B are long int:
+
+ if (INT_MULTIPLY_OVERFLOW (a, b))
+ printf ("result would overflow\n");
+ else
+ printf ("result is %ld (no overflow)\n", a * b);
+
+ Example usage with WRAPV flavor:
+
+ long int result;
+ bool overflow = INT_MULTIPLY_WRAPV (a, b, &result);
+ printf ("result is %ld (%s)\n", result,
+ overflow ? "after overflow" : "no overflow");
+
+ Restrictions on these macros:
+
+ These macros do not check for all possible numerical problems or
+ undefined or unspecified behavior: they do not check for division
+ by zero, for bad shift counts, or for shifting negative numbers.
+
+ These macros may evaluate their arguments zero or multiple times, so the
+ arguments should not have side effects.
+
+ The WRAPV macros are not constant expressions. They support only
+ +, binary -, and *. The result type must be signed.
+
+ These macros are tuned for their last argument being a constant.
+
+ Return 1 if the integer expressions A * B, A - B, -A, A * B, A / B,
+ A % B, and A << B would overflow, respectively. */
+
+#define INT_ADD_OVERFLOW(a, b) \
+ _GL_BINARY_OP_OVERFLOW (a, b, _GL_ADD_OVERFLOW)
+#define INT_SUBTRACT_OVERFLOW(a, b) \
+ _GL_BINARY_OP_OVERFLOW (a, b, _GL_SUBTRACT_OVERFLOW)
+#define INT_NEGATE_OVERFLOW(a) \
+ INT_NEGATE_RANGE_OVERFLOW (a, _GL_INT_MINIMUM (a), _GL_INT_MAXIMUM (a))
+#define INT_MULTIPLY_OVERFLOW(a, b) \
+ _GL_BINARY_OP_OVERFLOW (a, b, _GL_MULTIPLY_OVERFLOW)
+#define INT_DIVIDE_OVERFLOW(a, b) \
+ _GL_BINARY_OP_OVERFLOW (a, b, _GL_DIVIDE_OVERFLOW)
+#define INT_REMAINDER_OVERFLOW(a, b) \
+ _GL_BINARY_OP_OVERFLOW (a, b, _GL_REMAINDER_OVERFLOW)
+#define INT_LEFT_SHIFT_OVERFLOW(a, b) \
+ INT_LEFT_SHIFT_RANGE_OVERFLOW (a, b, \
+ _GL_INT_MINIMUM (a), _GL_INT_MAXIMUM (a))
+
+/* Return 1 if the expression A <op> B would overflow,
+ where OP_RESULT_OVERFLOW (A, B, MIN, MAX) does the actual test,
+ assuming MIN and MAX are the minimum and maximum for the result type.
+ Arguments should be free of side effects. */
+#define _GL_BINARY_OP_OVERFLOW(a, b, op_result_overflow) \
+ op_result_overflow (a, b, \
+ _GL_INT_MINIMUM (0 * (b) + (a)), \
+ _GL_INT_MAXIMUM (0 * (b) + (a)))
+
+/* Compute A + B, A - B, A * B, respectively, storing the result into *R.
+ Return 1 if the result overflows. See above for restrictions. */
+#define INT_ADD_WRAPV(a, b, r) \
+ _GL_INT_OP_WRAPV (a, b, r, +, __builtin_add_overflow, INT_ADD_OVERFLOW)
+#define INT_SUBTRACT_WRAPV(a, b, r) \
+ _GL_INT_OP_WRAPV (a, b, r, -, __builtin_sub_overflow, INT_SUBTRACT_OVERFLOW)
+#define INT_MULTIPLY_WRAPV(a, b, r) \
+ _GL_INT_OP_WRAPV (a, b, r, *, __builtin_mul_overflow, INT_MULTIPLY_OVERFLOW)
+
+#ifndef __has_builtin
+# define __has_builtin(x) 0
+#endif
+
+/* Nonzero if this compiler has GCC bug 68193 or Clang bug 25390. See:
+ https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68193
+ https://llvm.org/bugs/show_bug.cgi?id=25390
+ For now, assume all versions of GCC-like compilers generate bogus
+ warnings for _Generic. This matters only for older compilers that
+ lack __builtin_add_overflow. */
+#if __GNUC__
+# define _GL__GENERIC_BOGUS 1
+#else
+# define _GL__GENERIC_BOGUS 0
+#endif
+
+/* Store A <op> B into *R, where OP specifies the operation.
+ BUILTIN is the builtin operation, and OVERFLOW the overflow predicate.
+ See above for restrictions. */
+#if 5 <= __GNUC__ || __has_builtin (__builtin_add_overflow)
+# define _GL_INT_OP_WRAPV(a, b, r, op, builtin, overflow) builtin (a, b, r)
+#elif 201112 <= __STDC_VERSION__ && !_GL__GENERIC_BOGUS
+# define _GL_INT_OP_WRAPV(a, b, r, op, builtin, overflow) \
+ (_Generic \
+ (*(r), \
+ signed char: \
+ _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned char, \
+ signed char, SCHAR_MIN, SCHAR_MAX), \
+ short int: \
+ _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned short int, \
+ short int, SHRT_MIN, SHRT_MAX), \
+ int: \
+ _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned int, \
+ int, INT_MIN, INT_MAX), \
+ long int: \
+ _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long int, \
+ long int, LONG_MIN, LONG_MAX), \
+ long long int: \
+ _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long long int, \
+ long long int, LLONG_MIN, LLONG_MAX)))
+#else
+# define _GL_INT_OP_WRAPV(a, b, r, op, builtin, overflow) \
+ (sizeof *(r) == sizeof (signed char) \
+ ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned char, \
+ signed char, SCHAR_MIN, SCHAR_MAX) \
+ : sizeof *(r) == sizeof (short int) \
+ ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned short int, \
+ short int, SHRT_MIN, SHRT_MAX) \
+ : sizeof *(r) == sizeof (int) \
+ ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned int, \
+ int, INT_MIN, INT_MAX) \
+ : _GL_INT_OP_WRAPV_LONGISH(a, b, r, op, overflow))
+# ifdef LLONG_MAX
+# define _GL_INT_OP_WRAPV_LONGISH(a, b, r, op, overflow) \
+ (sizeof *(r) == sizeof (long int) \
+ ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long int, \
+ long int, LONG_MIN, LONG_MAX) \
+ : _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long long int, \
+ long long int, LLONG_MIN, LLONG_MAX))
+# else
+# define _GL_INT_OP_WRAPV_LONGISH(a, b, r, op, overflow) \
+ _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long int, \
+ long int, LONG_MIN, LONG_MAX))
+# endif
+#endif
+
+/* Store the low-order bits of A <op> B into *R, where the operation
+ is given by OP. Use the unsigned type UT for calculation to avoid
+ overflow problems. *R's type is T, with extremal values TMIN and
+ TMAX. T must be a signed integer type. */
+#define _GL_INT_OP_CALC(a, b, r, op, overflow, ut, t, tmin, tmax) \
+ (sizeof ((a) op (b)) < sizeof (t) \
+ ? _GL_INT_OP_CALC1 ((t) (a), (t) (b), r, op, overflow, ut, t, tmin, tmax) \
+ : _GL_INT_OP_CALC1 (a, b, r, op, overflow, ut, t, tmin, tmax))
+#define _GL_INT_OP_CALC1(a, b, r, op, overflow, ut, t, tmin, tmax) \
+ ((overflow (a, b) \
+ || (EXPR_SIGNED ((a) op (b)) && ((a) op (b)) < (tmin)) \
+ || (tmax) < ((a) op (b))) \
+ ? (*(r) = _GL_INT_OP_WRAPV_VIA_UNSIGNED (a, b, op, ut, t, tmin, tmax), 1) \
+ : (*(r) = _GL_INT_OP_WRAPV_VIA_UNSIGNED (a, b, op, ut, t, tmin, tmax), 0))
+
+/* Return A <op> B, where the operation is given by OP. Use the
+ unsigned type UT for calculation to avoid overflow problems.
+ Convert the result to type T without overflow by subtracting TMIN
+ from large values before converting, and adding it afterwards.
+ Compilers can optimize all the operations except OP. */
+#define _GL_INT_OP_WRAPV_VIA_UNSIGNED(a, b, op, ut, t, tmin, tmax) \
+ (((ut) (a) op (ut) (b)) <= (tmax) \
+ ? (t) ((ut) (a) op (ut) (b)) \
+ : ((t) (((ut) (a) op (ut) (b)) - (tmin)) + (tmin)))
+
+#endif /* _GL_INTPROPS_H */
diff --git a/time/tzfile.c b/time/tzfile.c
index 9049878..f988d39 100644
--- a/time/tzfile.c
+++ b/time/tzfile.c
@@ -65,6 +65,7 @@ static char *tzspec;
#include <endian.h>
#include <byteswap.h>
+#include <intprops.h>
/* Decode the four bytes at PTR as a signed integer in network byte order. */
static inline int
@@ -586,20 +587,26 @@ __tzfile_default (const char *std, const char *dst,
the flavor of its original type. */
type_idxs[i] = trans_type->isdst;
- if (trans_type->isgmt)
- /* The transition time is in GMT. No correction to apply. */ ;
- else if (isdst && !trans_type->isstd)
- /* The type says this transition is in "local wall clock time", and
- wall clock time as of the previous transition was DST. Correct
- for the difference between the rule's DST offset and the user's
- DST offset. */
- transitions[i] += dstoff - rule_dstoff;
- else
- /* This transition is in "local wall clock time", and wall clock
- time as of this iteration is non-DST. Correct for the
- difference between the rule's standard offset and the user's
- standard offset. */
- transitions[i] += stdoff - rule_stdoff;
+ if (! trans_type->isgmt)
+ {
+ long int correction;
+ if (isdst && !trans_type->isstd)
+ /* The type says this transition is in "local wall clock time", and
+ wall clock time as of the previous transition was DST. Correct
+ for the difference between the rule's DST offset and the user's
+ DST offset. */
+ correction = dstoff - rule_dstoff;
+ else
+ /* This transition is in "local wall clock time", and wall clock
+ time as of this iteration is non-DST. Correct for the
+ difference between the rule's standard offset and the user's
+ standard offset. */
+ correction = stdoff - rule_stdoff;
+ if (INT_ADD_WRAPV (correction, transitions[i], &transitions[i]))
+ transitions[i] = (correction < 0
+ ? TYPE_MINIMUM (time_t)
+ : TYPE_MAXIMUM (time_t));
+ }
/* The DST state of "local wall clock time" for the next iteration is
as specified by this transition. */
--
2.5.5
>From 83be322adc0d4ac94aaa46b157aff21447f0755d Mon Sep 17 00:00:00 2001
From: Paul Eggert <eggert@cs.ucla.edu>
Date: Wed, 13 Apr 2016 00:26:06 -0700
Subject: [PATCH 2/2] mktime assumes two's complement now
The code attempted to verify that the current platform is two's
complement, but in theory the verification would not have
correctly rejected some non-two's-complement hosts, and anyway the
verification effort was not worth the hassle.
Reported by Florian Weimer in:
https://sourceware.org/ml/libc-alpha/2016-04/msg00295.html
* time/mktime.c (TYPE_TWOS_COMPLEMENT):
Remove. All uses rewritten to assume two's complement, which is
all we can reasonably test nowadays anyway.
---
ChangeLog | 11 +++++++++++
time/mktime.c | 31 +++++++++++--------------------
2 files changed, 22 insertions(+), 20 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index 0a6d632..6f5efb4 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,16 @@
2016-04-13 Paul Eggert <eggert@cs.ucla.edu>
+ mktime assumes two's complement now
+ The code attempted to verify that the current platform is two's
+ complement, but in theory the verification would not have
+ correctly rejected some non-two's-complement hosts, and anyway the
+ verification effort was not worth the hassle.
+ Reported by Florian Weimer in:
+ https://sourceware.org/ml/libc-alpha/2016-04/msg00295.html
+ * time/mktime.c (TYPE_TWOS_COMPLEMENT):
+ Remove. All uses rewritten to assume two's complement, which is
+ all we can reasonably test nowadays anyway.
+
Fix integer overflow when adjusting posixrules
[BZ #19738]
This bug can occur on hosts with 32-bit time_t when reading tz files
diff --git a/time/mktime.c b/time/mktime.c
index bc7ed56..46f3e06 100644
--- a/time/mktime.c
+++ b/time/mktime.c
@@ -103,25 +103,20 @@ verify (long_int_is_wide_enough, INT_MAX == INT_MAX * (long_int) 2 / 2);
an integer. */
#define TYPE_IS_INTEGER(t) ((t) 1.5 == 1)
-/* True if negative values of the signed integer type T use two's
- complement, or if T is an unsigned integer type. */
-#define TYPE_TWOS_COMPLEMENT(t) ((t) ~ (t) 0 == (t) -1)
-
/* True if the arithmetic type T is signed. */
#define TYPE_SIGNED(t) (! ((t) 0 < (t) -1))
-/* The maximum and minimum values for the integer type T. These
- macros have undefined behavior if T is signed and has padding bits.
- If this is a problem for you, please let us know how to fix it for
- your host. */
-#define TYPE_MINIMUM(t) \
- ((t) (! TYPE_SIGNED (t) \
- ? (t) 0 \
- : ~ TYPE_MAXIMUM (t)))
-#define TYPE_MAXIMUM(t) \
- ((t) (! TYPE_SIGNED (t) \
- ? (t) -1 \
- : ((((t) 1 << (sizeof (t) * CHAR_BIT - 2)) - 1) * 2 + 1)))
+/* Minimum and maximum values for integer types.
+ These macros have undefined behavior for signed types that either
+ have padding bits or do not use two's complement. If this is a
+ problem for you, please let us know how to fix it for your host. */
+
+/* The maximum and minimum values for the integer type T. */
+#define TYPE_MINIMUM(t) ((t) ~ TYPE_MAXIMUM (t))
+#define TYPE_MAXIMUM(t) \
+ ((t) (! TYPE_SIGNED (t) \
+ ? (t) -1 \
+ : ((((t) 1 << (sizeof (t) * CHAR_BIT - 2)) - 1) * 2 + 1)))
#ifndef TIME_T_MIN
# define TIME_T_MIN TYPE_MINIMUM (time_t)
@@ -132,10 +127,6 @@ verify (long_int_is_wide_enough, INT_MAX == INT_MAX * (long_int) 2 / 2);
#define TIME_T_MIDPOINT (SHR (TIME_T_MIN + TIME_T_MAX, 1) + 1)
verify (time_t_is_integer, TYPE_IS_INTEGER (time_t));
-verify (twos_complement_arithmetic,
- (TYPE_TWOS_COMPLEMENT (int)
- && TYPE_TWOS_COMPLEMENT (long_int)
- && TYPE_TWOS_COMPLEMENT (time_t)));
#define EPOCH_YEAR 1970
#define TM_YEAR_BASE 1900
--
2.5.5