Squashed 'lib/fmt/' changes from 17dda583..b6f4ceae

REVERT: 17dda583 constexpr -> const for portability REVERT: 7ffe87c0 Fix docs REVERT: 3c4273dd Simplify UDL REVERT: 36d95c9f Fix docs REVERT: 44abd1f4 Update signatures in docs and ostream.h REVERT: db745986 Workaround broken std::numeric_limits REVERT: 8271e43e Improve __float128 support and use constexpr REVERT: 3f9b7433 Improve __float128 support REVERT: 71778e8b Specialize float_info for __float128 REVERT: f024565c Improve exponent handling in Dragon REVERT: e7f31f5c Cleanup format_dragon REVERT: 3c61799f Cleanup fuzzing mode REVERT: 4e39e130 Remove xchar.h include from ostream.h REVERT: ac0d9d5f Issue #2816: also strip named-arg for the fallback formatter REVERT: 4ad90578 Fix #2818: diagnose unformattable arguments in unpacked case REVERT: 17ba99c1 Fix #2817: add compile-time checking to ostream overloads of fmt::print REVERT: 3d19be28 Fix #2816: strip named argument wrappers for compile-time checking REVERT: c076a54a Move snprintf_float to format.h REVERT: 0419d238 Add FMT_USE_FLOAT128 REVERT: 69396347 Update color.h (#2815) REVERT: c51604a0 Reduce the number of configs REVERT: 587dc994 Remove windows-2016 env no longer suppported by GA REVERT: 1f3d44b8 Update std::tm/chrono docs REVERT: bc654faf Add is_floating_point that works with __float128 REVERT: 26bffce6 Simplify basic_memory_buffer REVERT: ed18ca3e Implement isnan REVERT: a204b8dd Add initial __float128 support REVERT: b6b003b0 Cleanup test REVERT: f2543b0a Add initial support for 128-bit floats REVERT: 72f48756 Simplify float_info REVERT: f91f61cd Reuse num_significand_bits REVERT: 9a1beab5 Workaround Windows API garbage REVERT: a8fe8bec Fix compilation error for ranges with ADL `begin`/`end` (#2807) REVERT: f6bcb25e Remove extra dot REVERT: b4a4189d Fix handling of implicit bit REVERT: 32d477e5 Add `styled` in documentation (#2805) REVERT: 0b7c045a Simplify _cf REVERT: c10fffec Make _cf visible in the doc build REVERT: dcfbe4a7 Document output_file default behavior correctly (#2803) REVERT: 8c9bc070 Implement styled arguments (#2793) REVERT: 5bc39d36 Eliminate intel compiler warnings (#2802) REVERT: e3d688e7 Fix warning C4251: class fmt::v8::file needs to have dll-interface (#2797) REVERT: 8d4f3e91 Update docs REVERT: 0cef1f81 Fixing formatting of certain kinds of ranges of ranges. (#2787) REVERT: 5c0d6564 Fix apt install REVERT: d416a995 Update README.rst REVERT: 3f67a124 Update README.rst REVERT: cc57e359 Update godbolt link in the readme (#2789) REVERT: 86477f7e Fix size computation REVERT: 0742606f Fix Conversion Warning (#2782) REVERT: 1ba69fb5 Remove snprintf FP fallback REVERT: ea6f0bf0 Minor cleanup REVERT: 1a18a2f3 Fixing "C4127: conditional expression is constant" Visual Studio 2022 warning in pedantic mode (#2783) REVERT: 4fcacea3 Parameterized fp on significand type REVERT: cf940ae8 Simplify to_decimal REVERT: 70dc3de0 Update format.h REVERT: cbc59ca8 Clear moved from memory buffer REVERT: ea3d326c Fix clang -Wliteral-range warning (#2779) REVERT: aad44f28 Add fmt::enums::format_as REVERT: 1319719a Add underlying_t REVERT: af5d8004 Limit Dragonbox to supported FP formats REVERT: 7b964209 Remove unused include REVERT: a0b43bfa Add support for 96-bit long double REVERT: 2c8cd2db Fix handling of zero precision REVERT: b6d56170 Remove unnecessary inline REVERT: 05432e57 Use consistent indentation REVERT: 47da218c Remove uintptr_fallback REVERT: 4ddab890 Merge accumulator into int128_fallback REVERT: d38f72af Refactor fallback ints REVERT: 15c2a3ba int128_t -> int128_opt REVERT: 532a69a6 Fix handling of 96-bit long double with -m32 REVERT: d8e1dd4a improve installing headers REVERT: ae25f796 add ability to build Apple framework using CMAKE_FRAMEWORK REVERT: ce93a66d Implement a fallback uint128_t REVERT: 6a134640 Include 128-bit with other signed integers in specifier check REVERT: 70de324a Apply 2746 fix for NVidia compiler also (#2770) REVERT: a1ea3e01 Move built-in formatter specialization to core REVERT: 161059dd Add support for extended precision FP REVERT: c4c6b42d Bump version REVERT: 21785040 Fix markup REVERT: 2b6f7fc7 Add partial support for extended precision FP REVERT: 0a24a071 Clz builtin may be not constexpr (Issue #2761) (#2762) REVERT: ba6f89c7 Update .bazelversion (#2766) REVERT: 5594edaf Address fmtlib/fmt#2763 (#2765) REVERT: 10e3b83a Replace ``make_args_checked`` with ``make_format_args`` (#2760) REVERT: c48353cb Update docs REVERT: 083510f0 Add FMT_CONSTEXPR to rotr instead REVERT: dba99bc8 Revert adding constexpr to rotr to satisfy C++11 compilers REVERT: c04af4bf Simplify remove_trailing_zeros REVERT: b348caa9 Remove some C-style casts for consistency REVERT: c8bd1e64 Simplify remove_trailing_zeros REVERT: 9b23e9dc Fix wrong comment/refer to a correct reference REVERT: 69f2c550 Remove std:: infront of uint32_t/64_t & add constexpr to rotr REVERT: 9b62310f Fix some conversion issues REVERT: 08d12f31 Fix typo REVERT: dbddb1d0 Remove literal separator to satisfy some compilers REVERT: 7dbe3dcd Recover log10_2_significand REVERT: 10642e60 Optimize remove_trailing_zeros REVERT: 7b4323e1 Add rotr REVERT: f1bd6f77 Check r < deltai first, because that is the major branch chosen for short inputs REVERT: 5d8eb6a1 Reflect the new paper - Change constants appearing in log & division computations - Rename beta_minus_1 to beta REVERT: 8e2e4d40 Suppress a gcc warning REVERT: a44716f5 Workaround to Intel compiler (#2758) REVERT: c71b0701 Add missing const qualifier (#2755) REVERT: ecd6022c Update docs REVERT: afbcf1e8 Remove legacy C locale wrapper REVERT: 90325d09 Fix stored type detection REVERT: e2ba01fc Fix overload ambiguity in print REVERT: 17b362f7 Simplify ostream opt-in API REVERT: a5a7e3a2 Update docs REVERT: f055ebbd Make ostream operators opt in to reduce the risk of ODR violations REVERT: 8a21e328 Remove problematic constructibility check REVERT: 31e743d0 Don't use ostream for types convertible to string_view REVERT: 35c0286c Simplify byte handling REVERT: c7173a36 Drop :: and fix formatting REVERT: 3e8372b9 qualify unqualified calls to format in compile.h (#2742) REVERT: a34a97cc Supporting ? as a string presentation type (#2674) REVERT: ae1aaaee Fix access mode of files created (#2530) (#2733) REVERT: 1557ab76 Add format_as for enums REVERT: b00a1eac Fixes NVIDIA HPC compiler and Intel ICC compatibility (#2732) REVERT: a7aecbfc Remove an old mingw workaround REVERT: dfcc730c Making target_compile_options PRIVATE, fix #2726, fix #2507 REVERT: f7a809be Clarify the choice of magic numbers and compute the most magic one REVERT: 09fde7f4 Add fmt::underlying for enum classes REVERT: 0014024a Don't rely on transitive includes REVERT: c2850055 FMT_NOEXCEPT -> noexcept REVERT: 6240d020 Improve comments REVERT: 925b744a Improve comments REVERT: 22b14ff2 Simplify cache recovery REVERT: 3dc26b44 Make a fallback path more compiler-friendly REVERT: 2e4038bf Simplify lines with __builtin_addcll and friends REVERT: 76336b4f Replace noexcept with FMT_NOEXCEPT REVERT: 91819834 Fix syntax errors REVERT: 74097a14 Remove now-unused stuffs REVERT: 21a1c533 Fix typo REVERT: 04eea0f0 Remove now-unused stuffs REVERT: 35a468ed Simplify integer checks REVERT: 1882a7a2 Replace Dragonbox cache which allows simpler cache recovery & integer checks REVERT: f4dd1b1b Simplify Dragonbox Step 3. REVERT: 70561ed1 Minimize the usage of built-in 128-bit ints It usually generates slower code than manual handling. REVERT: cdf1a3b5 Fix codecvt warning (#2408) (#2725) REVERT: b8b037e9 Fix -Wconversion warning (#2724) REVERT: 5985f0a7 Fix overflow for chrono durations (#2722) REVERT: 8f8a1a02 Fix handling of formattable types implicitly convertible to pointers REVERT: b02e5af5 fmt::join support FMT_COMPILE (#2720) REVERT: 58fb7823 Improve docs REVERT: 4fe6129d Fix FMT_NOEXCEPT definition REVERT: c056a009 Docs: Fix link to "Compile-time Format String Checks" section (#2712) REVERT: 7c12118c Deprecate buffered_file::fileno REVERT: 2a09d468 Use noexcept unconditionally REVERT: a126b4d8 Check if right shift is arithmetic REVERT: 9ff91b18 Simplify write_fractional_seconds REVERT: d9f045fb Fix a UB in chrono REVERT: c06bef72 Adding comments for range formatting. (#2706) REVERT: 3c98f1a4 Comment style REVERT: 6e0f1399 Supporting nested format specs for ranges. (#2673) REVERT: 0102101a Make colored print handle UTF-8 (#2701) REVERT: 4ac5269b Update ChangeLog.rst git-subtree-dir: lib/fmt git-subtree-split: b6f4ceaed0a0a24ccf575fab6c56dd50ccf6f1a9
Nazar-Hordii · Mar 24, 2022 · 29eb8e8 · 29eb8e8
1 parent af531ef
commit 29eb8e8
Show file tree

Hide file tree

Showing 40 changed files with 2,263 additions and 2,458 deletions.
diff --git a/.github/workflows/linux.yml b/.github/workflows/linux.yml
@@ -52,7 +52,6 @@ jobs:
     - name: Create Build Environment
       run: |
         ${{matrix.install}}
-        sudo apt update
         sudo apt install locales-all
         cmake -E make_directory ${{runner.workspace}}/build
 

diff --git a/.github/workflows/windows.yml b/.github/workflows/windows.yml
@@ -7,21 +7,26 @@ jobs:
     runs-on: ${{matrix.os}}
     strategy:
       matrix:
-        # windows-2019 has MSVC 2019 installed:
-        # https://github.com/actions/virtual-environments.
-        os: [windows-2019]
+        # windows-2016 and windows-2019 have MSVC 2017 and 2019 installed
+        # respectively: https://github.com/actions/virtual-environments.
+        os: [windows-2016, windows-2019]
         platform: [Win32, x64]
         build_type: [Debug, Release]
         standard: [11, 17, 20]
         include:
-          - os: windows-2019
+          - os: windows-2016
             platform: Win32
             build_type: Debug
             shared: -DBUILD_SHARED_LIBS=ON
         exclude:
+          - os: windows-2016
+            platform: Win32
+          - os: windows-2016
+            standard: 17
+          - os: windows-2016
+            standard: 20
           - os: windows-2019
             standard: 11
-            platform: Win32
           - os: windows-2019
             standard: 20
             platform: Win32

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -209,6 +209,18 @@ if (FMT_MASTER_PROJECT AND CMAKE_GENERATOR MATCHES "Visual Studio")
     ${CMAKE_MAKE_PROGRAM} -p:FrameworkPathOverride=\"${netfxpath}\" %*")
 endif ()
 
+set(strtod_l_headers stdlib.h)
+if (APPLE)
+  set(strtod_l_headers ${strtod_l_headers} xlocale.h)
+endif ()
+
+include(CheckSymbolExists)
+if (WIN32)
+  check_symbol_exists(_strtod_l "${strtod_l_headers}" HAVE_STRTOD_L)
+else ()
+  check_symbol_exists(strtod_l "${strtod_l_headers}" HAVE_STRTOD_L)
+endif ()
+
 function(add_headers VAR)
   set(headers ${${VAR}})
   foreach (header ${ARGN})
@@ -232,6 +244,17 @@ endif ()
 add_library(fmt ${FMT_SOURCES} ${FMT_HEADERS} README.rst ChangeLog.rst)
 add_library(fmt::fmt ALIAS fmt)
 
+if (HAVE_STRTOD_L)
+  target_compile_definitions(fmt PUBLIC FMT_LOCALE)
+endif ()
+
+if (MINGW)
+  check_cxx_compiler_flag("-Wa,-mbig-obj" FMT_HAS_MBIG_OBJ)
+  if (${FMT_HAS_MBIG_OBJ})
+    target_compile_options(fmt PUBLIC "-Wa,-mbig-obj")
+  endif()
+endif ()
+
 if (FMT_WERROR)
   target_compile_options(fmt PRIVATE ${WERROR_FLAG})
 endif ()
@@ -252,7 +275,6 @@ set(FMT_DEBUG_POSTFIX d CACHE STRING "Debug library postfix.")
 
 set_target_properties(fmt PROPERTIES
   VERSION ${FMT_VERSION} SOVERSION ${CPACK_PACKAGE_VERSION_MAJOR}
-  PUBLIC_HEADER "${FMT_HEADERS}"
   DEBUG_POSTFIX "${FMT_DEBUG_POSTFIX}")
 
 # Set FMT_LIB_NAME for pkg-config fmt.pc. We cannot use the OUTPUT_NAME target
@@ -330,8 +352,6 @@ if (FMT_INSTALL)
   install(TARGETS ${INSTALL_TARGETS} EXPORT ${targets_export_name}
           LIBRARY DESTINATION ${FMT_LIB_DIR}
           ARCHIVE DESTINATION ${FMT_LIB_DIR}
-          PUBLIC_HEADER DESTINATION "${FMT_INC_DIR}/fmt"
-          FRAMEWORK DESTINATION "."
           RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 
   # Use a namespace because CMake provides better diagnostics for namespaced
@@ -348,6 +368,7 @@ if (FMT_INSTALL)
 
   install(FILES $<TARGET_PDB_FILE:${INSTALL_TARGETS}>
           DESTINATION ${FMT_LIB_DIR} OPTIONAL)
+  install(FILES ${FMT_HEADERS} DESTINATION "${FMT_INC_DIR}/fmt")
   install(FILES "${pkgconfig}" DESTINATION "${FMT_PKGCONFIG_DIR}")
 endif ()
 

diff --git a/ChangeLog.rst b/ChangeLog.rst
@@ -6,7 +6,7 @@
   `#2696 <https://github.com/fmtlib/fmt/pull/2696>`_).
   Thanks `@saraedum (Julian Rüth) <https://github.com/saraedum>`_.
 
-* Fixed chrono formatting on big endian systems
+* Fixed chorno formatting on big endian systems
   (`#2698 <https://github.com/fmtlib/fmt/issues/2698>`_,
   `#2699 <https://github.com/fmtlib/fmt/pull/2699>`_).
   Thanks `@phprus (Vladislav Shchapov) <https://github.com/phprus>`_ and

diff --git a/README.rst b/README.rst
@@ -1,7 +1,5 @@
-.. image:: https://user-images.githubusercontent.com/
-           576385/156254208-f5b743a9-88cf-439d-b0c0-923d53e8d551.png
-   :width: 25%
-   :alt: {fmt}
+{fmt}
+=====
 
 .. image:: https://github.com/fmtlib/fmt/workflows/linux/badge.svg
    :target: https://github.com/fmtlib/fmt/actions?query=workflow%3Alinux
@@ -28,8 +26,9 @@
 **{fmt}** is an open-source formatting library providing a fast and safe
 alternative to C stdio and C++ iostreams.
 
-If you like this project, please consider donating to one of the funds that
-help victims of the war in Ukraine: https://www.stopputin.net/.
+If you like this project, please consider donating to the BYSOL
+Foundation that helps victims of political repressions in Belarus:
+https://bysol.org/en/bs/general/.
 
 `Documentation <https://fmt.dev>`__
 
@@ -124,7 +123,7 @@ Output::
     Default format: 42s 100ms
     strftime-like format: 03:15:30
 
-**Print a container** (`run <https://godbolt.org/z/MxM1YqjE7>`_)
+**Print a container** (`run <https://godbolt.org/z/MjsY7c>`_)
 
 .. code:: c++
 

diff --git a/doc/api.rst b/doc/api.rst
@@ -113,7 +113,8 @@ binary footprint, for example (https://godbolt.org/z/oba4Mc):
 
     template <typename S, typename... Args>
     void log(const char* file, int line, const S& format, Args&&... args) {
-      vlog(file, line, format, fmt::make_format_args(args...));
+      vlog(file, line, format,
+          fmt::make_args_checked<Args...>(format, args...));
     }
 
     #define MY_LOG(format, ...) \
@@ -124,6 +125,8 @@ binary footprint, for example (https://godbolt.org/z/oba4Mc):
 Note that ``vlog`` is not parameterized on argument types which improves compile
 times and reduces binary code size compared to a fully parameterized version.
 
+.. doxygenfunction:: fmt::make_args_checked(const S&, const remove_reference_t<Args>&...)
+
 .. doxygenfunction:: fmt::make_format_args(const Args&...)
 
 .. doxygenclass:: fmt::format_arg_store
@@ -179,11 +182,6 @@ functions and locale support.
 Formatting User-defined Types
 -----------------------------
 
-The {fmt} library provides formatters for many standard C++ types.
-See :ref:`fmt/ranges.h <ranges-api>` for ranges and tuples including standard
-containers such as ``std::vector`` and :ref:`fmt/chrono.h <chrono-api>` for date
-and time formatting.
-
 To make a user-defined type formattable, specialize the ``formatter<T>`` struct
 template and implement ``parse`` and ``format`` methods::
 
@@ -224,7 +222,7 @@ template and implement ``parse`` and ``format`` methods::
     // Formats the point p using the parsed format specification (presentation)
     // stored in this formatter.
     template <typename FormatContext>
-    auto format(const point& p, FormatContext& ctx) const -> decltype(ctx.out()) {
+    auto format(const point& p, FormatContext& ctx) -> decltype(ctx.out()) {
       // ctx.out() is an output iterator to write to.
       return presentation == 'f'
                 ? format_to(ctx.out(), "({:.1f}, {:.1f})", p.x, p.y)
@@ -246,7 +244,7 @@ example::
   template <> struct fmt::formatter<color>: formatter<string_view> {
     // parse is inherited from formatter<string_view>.
     template <typename FormatContext>
-    auto format(color c, FormatContext& ctx) const {
+    auto format(color c, FormatContext& ctx) {
       string_view name = "unknown";
       switch (c) {
       case color::red:   name = "red"; break;
@@ -284,7 +282,7 @@ You can also write a formatter for a hierarchy of classes::
   struct fmt::formatter<T, std::enable_if_t<std::is_base_of<A, T>::value, char>> :
       fmt::formatter<std::string> {
     template <typename FormatCtx>
-    auto format(const A& a, FormatCtx& ctx) const {
+    auto format(const A& a, FormatCtx& ctx) {
       return fmt::formatter<std::string>::format(a.name(), ctx);
     }
   };
@@ -309,7 +307,7 @@ The following user-defined literals are defined in ``fmt/format.h``.
 
 .. doxygenfunction:: operator""_format(const char *s, size_t n) -> detail::udl_formatter<char> 
 
-.. doxygenfunction:: operator""_a()
+.. doxygenfunction:: operator""_a(const char *s, size_t) -> detail::udl_arg<char>
 
 Utilities
 ---------
@@ -318,8 +316,6 @@ Utilities
 .. doxygenfunction:: fmt::ptr(const std::unique_ptr<T> &p) -> const void*
 .. doxygenfunction:: fmt::ptr(const std::shared_ptr<T> &p) -> const void*
 
-.. doxygenfunction:: fmt::underlying(Enum e) -> typename std::underlying_type<Enum>::type
-
 .. doxygenfunction:: fmt::to_string(const T &value) -> std::string
 
 .. doxygenfunction:: fmt::to_string_view(const Char *s) -> basic_string_view<Char>
@@ -450,19 +446,16 @@ The format syntax is described in :ref:`chrono-specs`.
 Format string compilation
 =========================
 
-``fmt/compile.h`` provides format string compilation enabled via the
-``FMT_COMPILE`` macro or the ``_cf`` user-defined literal. Format strings
-marked with ``FMT_COMPILE`` or ``_cf`` are parsed, checked and converted into
-efficient formatting code at compile-time. This supports arguments of built-in
-and string types as well as user-defined types with ``constexpr`` ``parse``
-functions in their ``formatter`` specializations. Format string compilation can
-generate more binary code compared to the default API and is only recommended in
-places where formatting is a performance bottleneck.
+``fmt/compile.h`` provides format string compilation support when using
+``FMT_COMPILE``. Format strings are parsed, checked and converted into efficient
+formatting code at compile-time. This supports arguments of built-in and string
+types as well as user-defined types with ``constexpr`` ``parse`` functions in
+their ``formatter`` specializations. Format string compilation can generate more
+binary code compared to the default API and is only recommended in places where
+formatting is a performance bottleneck.
 
 .. doxygendefine:: FMT_COMPILE
 
-.. doxygenfunction:: operator""_cf()
-
 .. _color-api:
 
 Terminal color and text style
@@ -476,8 +469,6 @@ Terminal color and text style
 
 .. doxygenfunction:: bg(detail::color_type)
 
-.. doxygenfunction:: styled(const T& value, text_style ts)
-
 .. _os-api:
 
 System APIs
@@ -495,9 +486,7 @@ System APIs
 ========================
 
 ``fmt/ostream.h`` provides ``std::ostream`` support including formatting of
-user-defined types that have an overloaded insertion operator (``operator<<``).
-In order to make a type formattable via ``std::ostream`` you should provide a
-``formatter`` specialization inherited from ``ostream_formatter``::
+user-defined types that have an overloaded insertion operator (``operator<<``)::
 
   #include <fmt/ostream.h>
 
@@ -511,12 +500,13 @@ In order to make a type formattable via ``std::ostream`` you should provide a
     }
   };
 
-  template <> struct fmt::formatter<date> : ostream_formatter {};
-
   std::string s = fmt::format("The date is {}", date(2012, 12, 9));
   // s == "The date is 2012-12-9"
 
-.. doxygenfunction:: print(std::ostream &os, format_string<T...> fmt, T&&... args)
+{fmt} only supports insertion operators that are defined in the same namespaces
+as the types they format and can be found with the argument-dependent lookup.
+
+.. doxygenfunction:: print(std::basic_ostream<Char> &os, const S &format_str, Args&&... args)
 
 .. _printf-api:
 

diff --git a/doc/build.py b/doc/build.py
@@ -65,7 +65,6 @@ def build_docs(version='dev', **kwargs):
                           FMT_USE_RVALUE_REFERENCES=1 \
                           FMT_USE_USER_DEFINED_LITERALS=1 \
                           FMT_USE_ALIAS_TEMPLATES=1 \
-                          FMT_USE_NONTYPE_TEMPLATE_PARAMETERS=1 \
                           FMT_API= \
                           "FMT_BEGIN_NAMESPACE=namespace fmt {{" \
                           "FMT_END_NAMESPACE=}}" \

diff --git a/doc/index.rst b/doc/index.rst
@@ -101,7 +101,7 @@ The code
   format(FMT_STRING("The answer is {:d}"), "forty-two");
 
 reports a compile-time error on compilers that support relaxed ``constexpr``.
-See `here <api.html#compile-time-format-string-checks>`_ for details.
+See `here <api.html#c.fmt>`_ for details.
 
 The following code
 

diff --git a/doc/syntax.rst b/doc/syntax.rst
@@ -304,8 +304,7 @@ The available presentation types for pointers are:
 Chrono Format Specifications
 ============================
 
-Format specifications for chrono types and ``std::tm`` have the following
-syntax:
+Format specifications for chrono types have the following syntax:
 
 .. productionlist:: sf
    chrono_format_spec: [[`fill`]`align`][`width`]["." `precision`][`chrono_specs`]
@@ -349,35 +348,9 @@ points are:
 Specifiers that have a calendaric component such as `'d'` (the day of month)
 are valid only for ``std::tm`` and not durations or time points.
 
-.. range-specs:
-
-Range Format Specifications
-===========================
-
-Format specifications for range types have the following syntax:
-
-..productionlist:: sf
-  range_format_spec: [":" [`underlying_spec`]]
-
-The `underlying_spec` is parsed based on the formatter of the range's
-reference type.
-
-By default, a range of characters or strings is printed escaped and quoted. But
-if any `underlying_spec` is provided (even if it is empty), then the characters
-or strings are printed according to the provided specification.
-
-Examples:
-
-  fmt::format("{}", std::vector{10, 20, 30});
-  // Result: [10, 20, 30]
-  fmt::format("{::#x}", std::vector{10, 20, 30});
-  // Result: [0xa, 0x14, 0x13]
-  fmt::format("{}", vector{'h', 'e', 'l', 'l', 'o'});
-  // Result: ['h', 'e', 'l', 'l', 'o']
-  fmt::format("{::}", vector{'h', 'e', 'l', 'l', 'o'});
-  // Result: [h, e, l, l, o]
-  fmt::format("{::d}", vector{'h', 'e', 'l', 'l', 'o'});
-  // Result: [104, 101, 108, 108, 111]
+``std::tm`` uses the system's `strftime
+<https://en.cppreference.com/w/cpp/chrono/c/strftime>`_ so refer to its
+documentation for details on supported conversion specifiers. 
 
 .. _formatexamples:
 

diff --git a/include/fmt/args.h b/include/fmt/args.h
@@ -95,10 +95,10 @@ class dynamic_format_arg_store
   };
 
   template <typename T>
-  using stored_type = conditional_t<
-      std::is_convertible<T, std::basic_string<char_type>>::value &&
-          !detail::is_reference_wrapper<T>::value,
-      std::basic_string<char_type>, T>;
+  using stored_type = conditional_t<detail::is_string<T>::value &&
+                                        !has_formatter<T, Context>::value &&
+                                        !detail::is_reference_wrapper<T>::value,
+                                    std::basic_string<char_type>, T>;
 
   // Storage of basic_format_arg must be contiguous.
   std::vector<basic_format_arg<Context>> data_;