[libcxx-commits] [libcxx] [libc++] LWG2381: Inconsistency in parsing floating point numbers (PR #77948)
A. Jiang via libcxx-commits
libcxx-commits at lists.llvm.org
Sat Jan 13 09:59:59 PST 2024
https://github.com/frederick-vs-ja updated https://github.com/llvm/llvm-project/pull/77948
>From c35da18ab546a36d3993cc94c6d7a0a1d8a5e714 Mon Sep 17 00:00:00 2001
From: "A. Jiang" <de34 at live.cn>
Date: Sun, 14 Jan 2024 01:59:13 +0800
Subject: [PATCH] [libc++] LWG2381: Inconsistency in parsing floating point
numbers
This PR implements LWG2381 by rejecting 'i', 'I', 'n', 'N' in FP
parsing, as inf and NaN are intendedly rejected by that LWG issue.
The source character array used for parsing is
"0123456789abcdefABCDEFxX+-pPiInN", whose first 26 or 28 characters
are used for parsing integers or floating-point values respectively.
Previously, libc++ used 32 characters, including 'i', 'I', 'n', 'N',
for FP parsing, which was inconsistent with LWG2381. This PR also
replaces magic numbers 26 and 28 (formerly 32) with named constants.
As a driven-by change, some parts of get_long_double.pass.cpp are
guarded with `LDBL_MANT_DIG >= 64`, because these parts fail on
platforms where long double is of binary64 format (e.g. MSVC).
Apple back-deployment targets remain broken due to dylib. XFAIL is
marked in related tests.
---
libcxx/docs/Status/Cxx23Issues.csv | 2 +-
libcxx/include/locale | 27 ++++++-----
.../facet.num.get.members/get_double.pass.cpp | 42 +++++++++--------
.../facet.num.get.members/get_float.pass.cpp | 42 +++++++++--------
.../get_long_double.pass.cpp | 45 +++++++++++--------
5 files changed, 91 insertions(+), 67 deletions(-)
diff --git a/libcxx/docs/Status/Cxx23Issues.csv b/libcxx/docs/Status/Cxx23Issues.csv
index b24ecc5525a149..a835a57ba90c54 100644
--- a/libcxx/docs/Status/Cxx23Issues.csv
+++ b/libcxx/docs/Status/Cxx23Issues.csv
@@ -98,7 +98,7 @@
`3555 <https://wg21.link/LWG3555>`__,"``{transform,elements}_view::iterator::iterator_concept`` should consider const-qualification of the underlying range","June 2021","","","|ranges|"
"","","","","",""
`2191 <https://wg21.link/LWG2191>`__,"Incorrect specification of ``match_results(match_results&&)``","October 2021","|Nothing To Do|",""
-`2381 <https://wg21.link/LWG2381>`__,"Inconsistency in parsing floating point numbers","October 2021","",""
+`2381 <https://wg21.link/LWG2381>`__,"Inconsistency in parsing floating point numbers","October 2021","|Complete|","18.0"
`2762 <https://wg21.link/LWG2762>`__,"``unique_ptr operator*()`` should be ``noexcept``","October 2021","",""
`3121 <https://wg21.link/LWG3121>`__,"``tuple`` constructor constraints for ``UTypes&&...`` overloads","October 2021","",""
`3123 <https://wg21.link/LWG3123>`__,"``duration`` constructor from representation shouldn't be effectively non-throwing","October 2021","","","|chrono|"
diff --git a/libcxx/include/locale b/libcxx/include/locale
index 70d22ff95e1ee1..021077c765693c 100644
--- a/libcxx/include/locale
+++ b/libcxx/include/locale
@@ -366,9 +366,11 @@ _LIBCPP_HIDE_FROM_ABI _ForwardIterator __scan_keyword(
struct _LIBCPP_EXPORTED_FROM_ABI __num_get_base {
static const int __num_get_buf_sz = 40;
+ static const size_t __int_chr_cnt = 26; // count of leading characters in __src used for parsing integers
+ static const size_t __fp_chr_cnt = 28; // count of leading characters in __src used for parsing floating-point values
static int __get_base(ios_base&);
- static const char __src[33];
+ static const char __src[33]; // "0123456789abcdefABCDEFxX+-pPiInN"
};
_LIBCPP_EXPORTED_FROM_ABI void
@@ -431,7 +433,7 @@ private:
template <typename _Tp>
const _Tp* __do_widen_p(ios_base& __iob, _Tp* __atoms) const {
locale __loc = __iob.getloc();
- use_facet<ctype<_Tp> >(__loc).widen(__src, __src + 26, __atoms);
+ use_facet<ctype<_Tp> >(__loc).widen(__src, __src + __int_chr_cnt, __atoms);
return __atoms;
}
@@ -447,7 +449,7 @@ private:
template <class _CharT>
string __num_get<_CharT>::__stage2_int_prep(ios_base& __iob, _CharT* __atoms, _CharT& __thousands_sep) {
locale __loc = __iob.getloc();
- std::use_facet<ctype<_CharT> >(__loc).widen(__src, __src + 26, __atoms);
+ std::use_facet<ctype<_CharT> >(__loc).widen(__src, __src + __int_chr_cnt, __atoms);
const numpunct<_CharT>& __np = std::use_facet<numpunct<_CharT> >(__loc);
__thousands_sep = __np.thousands_sep();
return __np.grouping();
@@ -458,7 +460,7 @@ template <class _CharT>
string __num_get<_CharT>::__stage2_float_prep(
ios_base& __iob, _CharT* __atoms, _CharT& __decimal_point, _CharT& __thousands_sep) {
locale __loc = __iob.getloc();
- std::use_facet<ctype<_CharT> >(__loc).widen(__src, __src + 32, __atoms);
+ std::use_facet<ctype<_CharT> >(__loc).widen(__src, __src + __fp_chr_cnt, __atoms);
const numpunct<_CharT>& __np = std::use_facet<numpunct<_CharT> >(__loc);
__decimal_point = __np.decimal_point();
__thousands_sep = __np.thousands_sep();
@@ -490,7 +492,7 @@ __num_get<_CharT>::__stage2_int_loop(_CharT __ct, int __base, char* __a, char*&
}
return 0;
}
- ptrdiff_t __f = std::find(__atoms, __atoms + 26, __ct) - __atoms;
+ ptrdiff_t __f = std::find(__atoms, __atoms + __int_chr_cnt, __ct) - __atoms;
if (__f >= 24)
return -1;
switch (__base) {
@@ -546,8 +548,8 @@ int __num_get<_CharT>::__stage2_float_loop(
}
return 0;
}
- ptrdiff_t __f = std::find(__atoms, __atoms + 32, __ct) - __atoms;
- if (__f >= 32)
+ ptrdiff_t __f = std::find(__atoms, __atoms + __num_get_base::__fp_chr_cnt, __ct) - __atoms;
+ if (__f >= static_cast<ptrdiff_t>(__num_get_base::__fp_chr_cnt))
return -1;
char __x = __src[__f];
if (__x == '-' || __x == '+') {
@@ -846,7 +848,7 @@ _InputIterator num_get<_CharT, _InputIterator>::__do_get_signed(
int __base = this->__get_base(__iob);
// Stage 2
char_type __thousands_sep;
- const int __atoms_size = 26;
+ const int __atoms_size = __num_get_base::__int_chr_cnt;
#ifdef _LIBCPP_ABI_OPTIMIZED_LOCALE_NUM_GET
char_type __atoms1[__atoms_size];
const char_type* __atoms = this->__do_widen(__iob, __atoms1);
@@ -895,7 +897,7 @@ _InputIterator num_get<_CharT, _InputIterator>::__do_get_unsigned(
int __base = this->__get_base(__iob);
// Stage 2
char_type __thousands_sep;
- const int __atoms_size = 26;
+ const int __atoms_size = __num_get_base::__int_chr_cnt;
#ifdef _LIBCPP_ABI_OPTIMIZED_LOCALE_NUM_GET
char_type __atoms1[__atoms_size];
const char_type* __atoms = this->__do_widen(__iob, __atoms1);
@@ -942,7 +944,7 @@ _InputIterator num_get<_CharT, _InputIterator>::__do_get_floating_point(
iter_type __b, iter_type __e, ios_base& __iob, ios_base::iostate& __err, _Fp& __v) const {
// Stage 1, nothing to do
// Stage 2
- char_type __atoms[32];
+ char_type __atoms[__num_get_base::__fp_chr_cnt];
char_type __decimal_point;
char_type __thousands_sep;
string __grouping = this->__stage2_float_prep(__iob, __atoms, __decimal_point, __thousands_sep);
@@ -996,10 +998,11 @@ _InputIterator num_get<_CharT, _InputIterator>::do_get(
// Stage 1
int __base = 16;
// Stage 2
- char_type __atoms[26];
+ char_type __atoms[__num_get_base::__int_chr_cnt];
char_type __thousands_sep = char_type();
string __grouping;
- std::use_facet<ctype<_CharT> >(__iob.getloc()).widen(__num_get_base::__src, __num_get_base::__src + 26, __atoms);
+ std::use_facet<ctype<_CharT> >(__iob.getloc())
+ .widen(__num_get_base::__src, __num_get_base::__src + __num_get_base::__int_chr_cnt, __atoms);
string __buf;
__buf.resize(__buf.capacity());
char* __a = &__buf[0];
diff --git a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_double.pass.cpp b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_double.pass.cpp
index c802ab787682aa..6b919946e8eaad 100644
--- a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_double.pass.cpp
+++ b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_double.pass.cpp
@@ -6,6 +6,12 @@
//
//===----------------------------------------------------------------------===//
+// The fix for LWG2381 (https://github.com/llvm/llvm-project/pull/77948) changed
+// behavior of FP parsing, while Apple back-deployment targets remain broken due
+// to the dylib.
+// XFAIL: stdlib=apple-libc++ && target={{.+}}-apple-macosx10.{{.+}}
+// XFAIL: stdlib=apple-libc++ && target={{.+}}-apple-macosx11.{{.+}}
+
// <locale>
// class num_get<charT, InputIterator>
@@ -116,9 +122,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == INFINITY);
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0);
}
{
const char str[] = "INF";
@@ -128,9 +134,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == INFINITY);
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0);
}
{
const char str[] = "-inf";
@@ -140,9 +146,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == -INFINITY);
+ assert(base(iter) == str + 1);
+ assert(err == ios.failbit);
+ assert(v == 0.0);
}
{
const char str[] = "-INF";
@@ -152,9 +158,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == -INFINITY);
+ assert(base(iter) == str + 1);
+ assert(err == ios.failbit);
+ assert(v == 0.0);
}
{
const char str[] = "nan";
@@ -164,9 +170,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(std::isnan(v));
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0);
}
{
const char str[] = "NAN";
@@ -176,9 +182,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(std::isnan(v));
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0);
}
{
v = -1;
diff --git a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_float.pass.cpp b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_float.pass.cpp
index 79c8480d0699b0..f7553f611640fd 100644
--- a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_float.pass.cpp
+++ b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_float.pass.cpp
@@ -6,6 +6,12 @@
//
//===----------------------------------------------------------------------===//
+// The fix for LWG2381 (https://github.com/llvm/llvm-project/pull/77948) changed
+// behavior of FP parsing, while Apple back-deployment targets remain broken due
+// to the dylib.
+// XFAIL: stdlib=apple-libc++ && target={{.+}}-apple-macosx10.{{.+}}
+// XFAIL: stdlib=apple-libc++ && target={{.+}}-apple-macosx11.{{.+}}
+
// <locale>
// class num_get<charT, InputIterator>
@@ -105,9 +111,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == INFINITY);
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0f);
}
{
const char str[] = "INF";
@@ -117,9 +123,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == INFINITY);
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0f);
}
{
const char str[] = "-inf";
@@ -129,9 +135,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == -INFINITY);
+ assert(base(iter) == str + 1);
+ assert(err == ios.failbit);
+ assert(v == 0.0f);
}
{
const char str[] = "-INF";
@@ -141,9 +147,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == -INFINITY);
+ assert(base(iter) == str + 1);
+ assert(err == ios.failbit);
+ assert(v == 0.0f);
}
{
const char str[] = "nan";
@@ -153,9 +159,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(std::isnan(v));
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0f);
}
{
const char str[] = "NAN";
@@ -165,9 +171,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(std::isnan(v));
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0f);
}
{
v = -1;
diff --git a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long_double.pass.cpp b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long_double.pass.cpp
index e2b2aeafd1ef92..e6d08a6cc2533c 100644
--- a/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long_double.pass.cpp
+++ b/libcxx/test/std/localization/locale.categories/category.numeric/locale.num.get/facet.num.get.members/get_long_double.pass.cpp
@@ -6,6 +6,12 @@
//
//===----------------------------------------------------------------------===//
+// The fix for LWG2381 (https://github.com/llvm/llvm-project/pull/77948) changed
+// behavior of FP parsing, while Apple back-deployment targets remain broken due
+// to the dylib.
+// XFAIL: stdlib=apple-libc++ && target={{.+}}-apple-macosx10.{{.+}}
+// XFAIL: stdlib=apple-libc++ && target={{.+}}-apple-macosx11.{{.+}}
+
// <locale>
// class num_get<charT, InputIterator>
@@ -16,6 +22,7 @@
#include <locale>
#include <ios>
#include <cassert>
+#include <cfloat>
#include <streambuf>
#include <cmath>
#include "test_macros.h"
@@ -105,9 +112,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == INFINITY);
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0l);
}
{
const char str[] = "INF";
@@ -117,9 +124,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == INFINITY);
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0l);
}
{
const char str[] = "-inf";
@@ -129,9 +136,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == -INFINITY);
+ assert(base(iter) == str + 1);
+ assert(err == ios.failbit);
+ assert(v == 0.0l);
}
{
const char str[] = "-INF";
@@ -141,9 +148,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(v == -INFINITY);
+ assert(base(iter) == str + 1);
+ assert(err == ios.failbit);
+ assert(v == 0.0l);
}
{
const char str[] = "nan";
@@ -153,9 +160,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(std::isnan(v));
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0l);
}
{
const char str[] = "NAN";
@@ -165,9 +172,9 @@ int main(int, char**)
f.get(cpp17_input_iterator<const char*>(str),
cpp17_input_iterator<const char*>(str+sizeof(str)),
ios, err, v);
- assert(base(iter) == str+sizeof(str)-1);
- assert(err == ios.goodbit);
- assert(std::isnan(v));
+ assert(base(iter) == str);
+ assert(err == ios.failbit);
+ assert(v == 0.0l);
}
{
const char str[] = "1.189731495357231765021264e+49321";
@@ -193,6 +200,7 @@ int main(int, char**)
assert(err == ios.failbit);
assert(v == INFINITY);
}
+#if LDBL_MANT_DIG >= 64
{
const char str[] = "11.189731495357231765021264e+4932";
std::ios_base::iostate err = ios.goodbit;
@@ -229,6 +237,7 @@ int main(int, char**)
assert(err != ios.failbit);
assert(v == 304888344611713860501504000000.0L);
}
+#endif // LDBL_MANT_DIG >= 64
{
v = -1;
const char str[] = "1.19973e+4933"; // unrepresentable
More information about the libcxx-commits
mailing list