[PATCH 1/2] libstdc++: case-sensitivity in hexfloat std::from_chars [PR105441]

[PATCH 1/2] libstdc++: case-sensitivity in hexfloat std::from_chars [PR105441]

[Patrick Palka]

The hexfloat parser for binary32/64 added in r12-6645-gcc3bf3404e4b1c
overlooked that the exponent part can also begin with an uppercase 'P'.

Tested on x86_64-pc-linux-gnu, does this look OK for trunk/11, and possibly the
12 branch now for 12.1?

	PR libstdc++/105441

libstdc++-v3/ChangeLog:

	* src/c++17/floating_from_chars.cc (__floating_from_chars_hex):
	Also accept 'P' as the start of the exponent.
	* testsuite/20_util/from_chars/7.cc: Add corresponding testcase.
---
 libstdc++-v3/src/c++17/floating_from_chars.cc  | 2 +-
 libstdc++-v3/testsuite/20_util/from_chars/7.cc | 1 +
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/libstdc++-v3/src/c++17/floating_from_chars.cc b/libstdc++-v3/src/c++17/floating_from_chars.cc
index 13de1e346ab..e7f3a58cf18 100644
--- a/libstdc++-v3/src/c++17/floating_from_chars.cc
+++ b/libstdc++-v3/src/c++17/floating_from_chars.cc
@@ -664,7 +664,7 @@ namespace
 
     // Parse the written exponent.
     int written_exponent = 0;
-    if (first != last && *first == 'p')
+    if (first != last && (*first == 'p' || *first == 'P'))
       {
 	// Tentatively consume the 'p' and try to parse a decimal number.
 	const char* const fallback_first = first;
diff --git a/libstdc++-v3/testsuite/20_util/from_chars/7.cc b/libstdc++-v3/testsuite/20_util/from_chars/7.cc
index 2a78c7441e2..1aa9b230531 100644
--- a/libstdc++-v3/testsuite/20_util/from_chars/7.cc
+++ b/libstdc++-v3/testsuite/20_util/from_chars/7.cc
@@ -96,6 +96,7 @@ constexpr testcase testcases[] = {
   { "1p-1", 4, {}, 0x1p-1 },
   { "0", 1, {}, 0.0 },
   { "A", 1, {}, 0xA },
+  { "1.ABCDEFP+10", 12, {}, 0x1.ABCDEFP+10 },
   { "-1", 2, {}, -1.0 },
   { "-0", 2, {}, -0.0 },
   { "42", 2, {}, 0x42p0 },
-- 
2.36.0.44.g0f828332d5

[PATCH 2/2] libstdc++: Don't use std::tolower in <charconv> [PR103911]

[Patrick Palka]

As in r12-6281-gc83ecfbe74a5cf for std::isdigit, <charconv> shouldn't
use std::tolower either I think.

Tested on x86_64-pc-linux-gnu, does this look OK for trunk/11 and the
12 branch after it's thawed?

	PR libstdc++/103911

libstdc++-v3/ChangeLog:

	* src/c++17/floating_from_chars.cc (find_end_of_float): Accept
	two possible delimeters for the exponent part in the form of a
	(possibly NULL) string of length two.  Don't use std::tolower.
	(pattern): Adjust calls to find_end_of_float accordingly.
---
 libstdc++-v3/src/c++17/floating_from_chars.cc | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/libstdc++-v3/src/c++17/floating_from_chars.cc b/libstdc++-v3/src/c++17/floating_from_chars.cc
index e7f3a58cf18..5d2a931d5dd 100644
--- a/libstdc++-v3/src/c++17/floating_from_chars.cc
+++ b/libstdc++-v3/src/c++17/floating_from_chars.cc
@@ -40,7 +40,6 @@
 #include <cmath>
 #include <cstdlib>
 #include <cstring>
-#include <cctype>
 #include <locale.h>
 #include <bits/functexcept.h>
 #if _GLIBCXX_HAVE_XLOCALE_H
@@ -142,10 +141,10 @@ namespace
 
   // Find initial portion of [first, last) containing a floating-point number.
   // The string `digits` is either `dec_digits` or `hex_digits`
-  // and `exp` is 'e' or 'p' or '\0'.
+  // and `exp` is "eE", "pP" or NULL.
   const char*
   find_end_of_float(const char* first, const char* last, const char* digits,
-		    char exp)
+		    const char *exp)
   {
     while (first < last && strchr(digits, *first) != nullptr)
       ++first;
@@ -155,7 +154,7 @@ namespace
 	while (first < last && strchr(digits, *first))
 	  ++first;
       }
-    if (first < last && exp != 0 && std::tolower((unsigned char)*first) == exp)
+    if (first < last && exp != nullptr && (*first == exp[0] || *first == exp[1]))
       {
 	++first;
 	if (first < last && (*first == '-' || *first == '+'))
@@ -237,7 +236,7 @@ namespace
 
 	if ((last - first + 2) > buffer_resource::guaranteed_capacity())
 	  {
-	    last = find_end_of_float(first + neg, last, digits, 'p');
+	    last = find_end_of_float(first + neg, last, digits, "pP");
 #ifndef __cpp_exceptions
 	    if ((last - first + 2) > buffer_resource::guaranteed_capacity())
 	      {
@@ -261,7 +260,7 @@ namespace
 	if ((last - first) > buffer_resource::guaranteed_capacity())
 	  {
 	    last = find_end_of_float(first + neg, last, digits,
-				     "e"[fmt == chars_format::fixed]);
+				     fmt == chars_format::fixed ? nullptr : "eE");
 #ifndef __cpp_exceptions
 	    if ((last - first) > buffer_resource::guaranteed_capacity())
 	      {
-- 
2.36.0.44.g0f828332d5

Re: [PATCH 2/2] libstdc++: Don't use std::tolower in <charconv> [PR103911]

[Jonathan Wakely]

On Sun, 1 May 2022 at 19:21, Patrick Palka via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
> As in r12-6281-gc83ecfbe74a5cf for std::isdigit, <charconv> shouldn't
> use std::tolower either I think.
>
> Tested on x86_64-pc-linux-gnu, does this look OK for trunk/11 and the
> 12 branch after it's thawed?

OK, thanks.

>
>         PR libstdc++/103911
>
> libstdc++-v3/ChangeLog:
>
>         * src/c++17/floating_from_chars.cc (find_end_of_float): Accept
>         two possible delimeters for the exponent part in the form of a
>         (possibly NULL) string of length two.  Don't use std::tolower.
>         (pattern): Adjust calls to find_end_of_float accordingly.
> ---
>  libstdc++-v3/src/c++17/floating_from_chars.cc | 11 +++++------
>  1 file changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/libstdc++-v3/src/c++17/floating_from_chars.cc b/libstdc++-v3/src/c++17/floating_from_chars.cc
> index e7f3a58cf18..5d2a931d5dd 100644
> --- a/libstdc++-v3/src/c++17/floating_from_chars.cc
> +++ b/libstdc++-v3/src/c++17/floating_from_chars.cc
> @@ -40,7 +40,6 @@
>  #include <cmath>
>  #include <cstdlib>
>  #include <cstring>
> -#include <cctype>
>  #include <locale.h>
>  #include <bits/functexcept.h>
>  #if _GLIBCXX_HAVE_XLOCALE_H
> @@ -142,10 +141,10 @@ namespace
>
>    // Find initial portion of [first, last) containing a floating-point number.
>    // The string `digits` is either `dec_digits` or `hex_digits`
> -  // and `exp` is 'e' or 'p' or '\0'.
> +  // and `exp` is "eE", "pP" or NULL.
>    const char*
>    find_end_of_float(const char* first, const char* last, const char* digits,
> -                   char exp)
> +                   const char *exp)
>    {
>      while (first < last && strchr(digits, *first) != nullptr)
>        ++first;
> @@ -155,7 +154,7 @@ namespace
>         while (first < last && strchr(digits, *first))
>           ++first;
>        }
> -    if (first < last && exp != 0 && std::tolower((unsigned char)*first) == exp)
> +    if (first < last && exp != nullptr && (*first == exp[0] || *first == exp[1]))
>        {
>         ++first;
>         if (first < last && (*first == '-' || *first == '+'))
> @@ -237,7 +236,7 @@ namespace
>
>         if ((last - first + 2) > buffer_resource::guaranteed_capacity())
>           {
> -           last = find_end_of_float(first + neg, last, digits, 'p');
> +           last = find_end_of_float(first + neg, last, digits, "pP");
>  #ifndef __cpp_exceptions
>             if ((last - first + 2) > buffer_resource::guaranteed_capacity())
>               {
> @@ -261,7 +260,7 @@ namespace
>         if ((last - first) > buffer_resource::guaranteed_capacity())
>           {
>             last = find_end_of_float(first + neg, last, digits,
> -                                    "e"[fmt == chars_format::fixed]);
> +                                    fmt == chars_format::fixed ? nullptr : "eE");
>  #ifndef __cpp_exceptions
>             if ((last - first) > buffer_resource::guaranteed_capacity())
>               {
> --
> 2.36.0.44.g0f828332d5
>

Re: [PATCH 1/2] libstdc++: case-sensitivity in hexfloat std::from_chars [PR105441]

[Jonathan Wakely]

On Sun, 1 May 2022 at 19:22, Patrick Palka via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
> The hexfloat parser for binary32/64 added in r12-6645-gcc3bf3404e4b1c
> overlooked that the exponent part can also begin with an uppercase 'P'.
>
> Tested on x86_64-pc-linux-gnu, does this look OK for trunk/11, and possibly the
> 12 branch now for 12.1?

OK for trunk, and as it's a regression since gcc-11 I'd recommend
release manager approval for 12.1 as well, but it's the RMs call.



>
>         PR libstdc++/105441
>
> libstdc++-v3/ChangeLog:
>
>         * src/c++17/floating_from_chars.cc (__floating_from_chars_hex):
>         Also accept 'P' as the start of the exponent.
>         * testsuite/20_util/from_chars/7.cc: Add corresponding testcase.
> ---
>  libstdc++-v3/src/c++17/floating_from_chars.cc  | 2 +-
>  libstdc++-v3/testsuite/20_util/from_chars/7.cc | 1 +
>  2 files changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/libstdc++-v3/src/c++17/floating_from_chars.cc b/libstdc++-v3/src/c++17/floating_from_chars.cc
> index 13de1e346ab..e7f3a58cf18 100644
> --- a/libstdc++-v3/src/c++17/floating_from_chars.cc
> +++ b/libstdc++-v3/src/c++17/floating_from_chars.cc
> @@ -664,7 +664,7 @@ namespace
>
>      // Parse the written exponent.
>      int written_exponent = 0;
> -    if (first != last && *first == 'p')
> +    if (first != last && (*first == 'p' || *first == 'P'))
>        {
>         // Tentatively consume the 'p' and try to parse a decimal number.
>         const char* const fallback_first = first;
> diff --git a/libstdc++-v3/testsuite/20_util/from_chars/7.cc b/libstdc++-v3/testsuite/20_util/from_chars/7.cc
> index 2a78c7441e2..1aa9b230531 100644
> --- a/libstdc++-v3/testsuite/20_util/from_chars/7.cc
> +++ b/libstdc++-v3/testsuite/20_util/from_chars/7.cc
> @@ -96,6 +96,7 @@ constexpr testcase testcases[] = {
>    { "1p-1", 4, {}, 0x1p-1 },
>    { "0", 1, {}, 0.0 },
>    { "A", 1, {}, 0xA },
> +  { "1.ABCDEFP+10", 12, {}, 0x1.ABCDEFP+10 },
>    { "-1", 2, {}, -1.0 },
>    { "-0", 2, {}, -0.0 },
>    { "42", 2, {}, 0x42p0 },
> --
> 2.36.0.44.g0f828332d5
>

Re: [PATCH 1/2] libstdc++: case-sensitivity in hexfloat std::from_chars [PR105441]

[Jakub Jelinek]

On Sun, May 01, 2022 at 08:09:05PM +0100, Jonathan Wakely via Gcc-patches wrote:
> On Sun, 1 May 2022 at 19:22, Patrick Palka via Libstdc++
> <libstdc++@gcc.gnu.org> wrote:
> >
> > The hexfloat parser for binary32/64 added in r12-6645-gcc3bf3404e4b1c
> > overlooked that the exponent part can also begin with an uppercase 'P'.
> >
> > Tested on x86_64-pc-linux-gnu, does this look OK for trunk/11, and possibly the
> > 12 branch now for 12.1?
> 
> OK for trunk, and as it's a regression since gcc-11 I'd recommend
> release manager approval for 12.1 as well, but it's the RMs call.

Ok for 12.1.

> >         PR libstdc++/105441
> >
> > libstdc++-v3/ChangeLog:
> >
> >         * src/c++17/floating_from_chars.cc (__floating_from_chars_hex):
> >         Also accept 'P' as the start of the exponent.
> >         * testsuite/20_util/from_chars/7.cc: Add corresponding testcase.

	Jakub