C++17 Should Refer To C11 Instead Of C99 - Open
Doc. No.: | P0063R2 | |
---|---|---|
Date: | 2016-03-18 | |
Reply to: | Clark Nelson | Hans-J. Boehm |
Email: | clark.nelson@intel.com | hboehm@google.com |
Audience: | Library Evolution |
This document is a follow-on to P0063R0, from the pre-Kona mailing. It goes into the details of the “easy part” described therein. Much of its rationale is not repeated here.
It has also been revised as a result of LEWG review in Jacksonville. Changes from R1 are boxed.
Signal handler restrictions
18.10p10 effectively says that a signal handler must be written in the common subset of C and C++. But C++ and C11 both have thread-local storage. Referencing C11 instead of C99 might be taken to suggest that it would be OK for a signal handler to reference a thread-local object.
It turns out that there is a technicality/loophole by which this can be dodged. Because the associated keyword is spelled differently in C and C++, technically thread-local storage is not actually in the common subset of the two languages – even though both languages support the feature.
While this state would not provide a satisfying long-term resolution, it is considered adequate to separate the deep technical issues of signal handlers from the broader, more editorial issues of updating the normative reference. Hans has written P0270 to address signal handlers.
This document also doesn't go into interoperability of threads or atomics in a mixed C/C++ program.
Header coverage
C11 made the <complex.h> header optional. Currently in C++, the <ccomplex> header maps directly to the <complex> header. Simply deleting the requirement for a <ccomplex> header is simpler, and may be better, than making it optional. Of course an implementation would not need to change in order to conform to the new state of affairs, but programs that use <ccomplex> instead of <complex> would cease to be strictly conforming.
C11 also added five new headers:
<stdatomic.h> <threads.h> These are also optional in C11, and since their functionality is already provided in C++ by other means, it's not clear that inventing optional wrappers for them would be worth the effort. <stdalign.h> <uchar.h> C++ already covers these two, having inherited one of them from the C Unicode TR (later incorporated into C11), and having effectively jumped the gun on the other one. <stdnoreturn.h> All this provides is a single prettifying macro: #define noreturn _Noreturn In C++, [[noreturn]] is probably already considered pretty enough.So it's not clear that C++ needs to add anything for any of these new headers.
LEWG decided that <stdatomic.h>, <stdnoreturn.h> and <threads.h> should all be ignored – except that there should be a footnote pointing out that their omission is intentional.
It was also decided that the existing headers <ccomplex> and <ctgmath> should be deprecated. <ciso646> was called out as explicitly having no effect according to the standard, but apparently the real world is not so simple. <cstdalign> and <cstdbool> were mentioned as also being pointless in C++, but were not the subject of any straw poll; I have taken the liberty of assuming that perhaps they should be deprecated as well.
Annex K Bounds checking interfaces
C11 optionally added several dozen new functions to the standard library, as safer alternatives to existing standard C library functions. For example, strcat_s was added as an alternative to strcat.
The declarations of these functions were added to existing headers; backward compatibility was provided by requiring that they be controlled by the setting of a macro: __STDC_WANT_LIB_EXT1__. A program can suppress them by defining the macro to zero, or request them by defining the macro to one. Whether they are declared if the macro is not defined by the program is implementation-defined.
At a recent WG14 meeting, a proposal was considered (N1967) that these functions should be removed from a future revision of the standard.
LEWG decided that these functions should not be imported into the std namespace, but that it should be entirely implementation-defined whether they are declared in the global namespace. (Any implementation can document that it does the full “macro dance” described by the C standard.)
Working draft changes
1.2 Normative references
Change paragraph 1:
...
- ...
- ISO/IEC 9899:19992011, Programming languages — C
- ISO/IEC 9899:19992011/Cor 1:20012012(E), Programming languages — C, Technical Corrigendum 1
- ISO/IEC 9899:1999/Cor.2:2004(E), Programming languages — C, Technical Corrigendum 2
- ISO/IEC 9899:1999/Cor.3:2007(E), Programming languages — C, Technical Corrigendum 3
- ...
- ISO/IEC TR 19769:2004, Information technology — Programming languages, their environments and system software interfaces — Extensions for the programming language C to support new character data types
Change paragraph 2:
The library described in Clause 7 of ISO/IEC 9899:19992011 and Clause 7 of ISO/IEC 9899:1999/Cor.1:2001 and Clause 7 of ISO/IEC 9899:1999/Cor.2:2003 is hereinafter called the C standard library.1
Delete paragraph 3:
The library described in ISO/IEC TR 19769:2004 is hereinafter called the C Unicode TR.
17.5.1.5 C Library
Change paragraph 1:
Paragraphs labeled “See also:” contain cross-references to the relevant portions of this International Standard and the ISO C standard, which is incorporated into this International Standard by reference.
17.6.1.2 Headers
Add a footnote to paragraph 3:
The facilities of the C standard Library are provided in 26 additional headers, as shown in Table 15.N)
N) It is intentional that there is no C++ header for any of these C headers: <stdatomic.h>, <stdnoreturn.h>, <threads.h>.
Change paragraph 4:
Except as noted in Clauses 18 through 30 and Annex D, the contents of each header cname shall be the same as that of the corresponding header name.h, as specified in the C standard library (1.2) or the C Unicode TR, as appropriate, as if by inclusion. In the C++ standard library, however, the declarations (except for names which are defined as macros in C) are within namespace scope (3.3.6) of the namespace std. It is unspecified whether these names (including any overloads added in Clauses 18 through 30 and Annex D) are first declared within the global namespace scope and are then injected into namespace std by explicit using-declarations (7.3.3).
Add a new paragraph to this section:
Annex K of the C standard describes a large number of functions which “promote safer, more secure programming” than many of the traditional C library functions. The names of most of these functions have a suffix of _s; they provide the same service as the C library function with the unsuffixed name, but generally take an additional argument, whose value is the size of the result array. If a C++ header is included that corresponds to a C header that declares any of these functions, it is implementation-defined whether any of these functions is declared in the global namespace; none of them is declared in namespace std.
18.2 Types
Change reference to C standard for paragraph 9:
See also: ... ISO C 7.1.6 7.19.
18.3.3 C library
Change table 32:
DBL_DIG DBL_MIN_EXP FLT_MAX_EXP LDBL_MANT_DIG DBL_EPSILON DECIMAL_DIG FLT_MIN LDBL_MAX_10_EXP DBL_MANT_DIG FLT_DIG FLT_MIN_10_EXP LDBL_MAX_EXP DBL_MAX FLT_EPSILON FLT_MIN_EXP LDBL_MAX DBL_MAX_10_EXP FLT_EVAL_METHOD FLT_RADIX LDBL_MIN DBL_MAX_EXP FLT_MANT_DIG FLT_ROUNDS LDBL_MIN_10_EXP DBL_MIN FLT_MAX LDBL_DIG LDBL_MIN_EXP DBL_MIN_10_EXP FLT_MAX_10_EXP LDBL_EPSILON DBL_HAS_SUBNORM FLT_HAS_SUBNORM LDBL_HAS_SUBNORM DBL_DECIMAL_DIG FLT_DECIMAL_DIG LDBL_DECIMAL_DIG DBL_TRUE_MIN FLT_TRUE_MIN LDBL_TRUE_MIN
It looks like all the functionality represented by these new macros is already available from numeric_limits.
Change references to C standard:
See also: ISO C 7.1.5, 5.2.4.2.2, 5.2.4.2.1.
18.4.1 Header <cstdint> synopsys
<cstdint> is the only other C header wrapper, besides <cfenv>, whose contents are not synopsized in a table of names. But until it has been decided whether that is the desired direction, as opposed to full synopses (as in P0175), I'm not going to try to make a table-form synopsis for this.
Change paragraph 2:
The header defines all types and macros the same as 7.18 in the C standard the Standard C library header<stdint.h>. [ Note: The macros defined by <cstdint> are provided unconditionally. In particular, the symbols __STDC_LIMIT_MACROS and __STDC_CONSTANT_MACROS (mentioned in footnotes 219, 220, and 222 in the C standard) play no role in C++. —end note ]
18.5 Start and termination
Change paragraph 12:
... [ Note: The standard file buffers are not flushed. See: ISO C 7.20.4.4 7.22.4.5. —end note ]
Change reference to C standard for paragraph 4:
See also: ... ISO C 7.10.4 7.22.4.
18.6.1.4 Data races
Change paragraph 1:
For purposes of determining the existence of data races, the library versions of operator new, user replacement versions of global operator new, the C standard library functions aligned_alloc, calloc and malloc, the library versions of operator delete, user replacement versions of operator delete, the C standard library function free, and the C standard library function realloc shall not introduce a data race (17.6.5.9). Calls to these functions that allocate or deallocate a particular unit of storage shall occur in a single total order, and each such deallocation call shall happen before (1.10) the next allocation (if any) in this order.
18.6.1.1 also references malloc, and reference to aligned_alloc should be made as well; that issue is addressed in P0035.
18.10 Other runtime support
Change reference to C standard for paragraph 3:
See also: ISO C 4.8.1.1 7.16.1.1.
Change references to C standard for paragraph 4:
See also: ISO C 7.10.4, 7.8, 7.6, 7.13 7.12.
7.10.4 referred to the exit family of functions; 7.8 to <stdarg.h>; and 7.12 (possibly) to <time.h>. These other references may be relevant elsewhere, but not to this paragraph, which is solely about setjmp and longjmp.
19.3 Assertions
Change paragraph 3:
The contents are the same as the Standard C library header <assert.h>, except that a macro named static_assert is not defined in header <cassert> or in header <assert.h>.
Note: This kind of problem is handled in a slightly different way in 18.10p7-8.
20.7.13 C library
Change table 45:
Functions: calloc malloc aligned_alloc free realloc
Change paragraph 3:
The functions aligned_alloc(), calloc(), malloc(), and realloc() do not attempt to allocate storage by calling ::operator new() (18.6).
Change the reference to the C standard for paragraph 4:
See also: ISO C Clause 7.11.2 7.22.3.
Change paragraph 5:
Storage allocated directly with aligned_alloc(), malloc(), calloc(), or realloc() is implicitly declared reachable (see 3.7.4.3) on allocation, ....
Change the reference to the C standard for paragraph 7:
See also: ISO C Clause 7.11.2 7.24.
20.12.8 Date and time functions
Change table 60:
Macros: NULL CLOCKS_PER_SEC TIME_UTC Types: size_t clock_t time_t Struct: tm timespec Functions: asctime clock difftime localtime strftime ctime gmtime mktime time timespec_get
Change the references to the C standard for paragraph 2:
See also: ISO C Clause 7.12 7.27. , Amendment 1 Clause 4.6.4.
21.8 Null-terminated sequence utilities
Change the references to the C standard for paragraph 15:
See also: ISO C 7.3, 7.10.7, 7.10.8, and 7.11. Amendment 1 4.4, 4.5, and 4.6 7.4, 7.22.7, 7.22.8, 7.24, 7.28, 7.29 and 7.30.
22.6 C library locales
Change the reference to the C standard for paragraph 3:
See also: ISO C Clause 7.4 7.11.
25.5 C library algorithms
Change the reference to the C standard for paragraph 4:
See also: ISO C Clause 7.10.5 7.22.5.
26.3 The floating-point environment
The changes to be made to this section will depend on the direction to be taken on P0175.
For editorial consistency, replace everything from section header 26.3.1 through paragraph 2 with the following:
Table N describes header <cfenv>.
Table N — Header <cfenv> synopsis
Macros: FE_DIVBYZERO FE_OVERFLOW FE_DOWNWARD FE_UPWARD FE_INEXACT FE_UNDERFLOW FE_TONEAREST FE_DFL_ENV FE_INVALID FE_ALL_EXCEPT FE_TOWARDZERO Types: fenv_t fexcept_t Functions: feclearexcept fesetexceptflag fesetround fesetenv fegetexceptflag fetestexcept fegetenv feupdateenv feraiseexcept fegetround feholdexcept The contents of this header are the same as the Standard C library header <fenv.h>.
26.8 C library
Change paragraph 1:
The header <ctgmath> simply includes the headers <ccomplex> <complex> and <cmath>.
Change paragraph 2:
[ Note: The overloads provided in C by type-generic macros are already provided in <ccomplex> <complex> and <cmath> by “sufficient” additional overloads. —end note ]
Change the references to the C standard for paragraph 11:
See also: ISO C 7.5, 7.10.2, 7.10.6 7.12, 7.22.2, 7.22.6.
27.4.1 Overview
Change paragraph 3:
Mixing operations on corresponding wide- and narrow-character streams follows the same semantics as mixing such operations on FILEs, as specified in Amendment 1 of the ISO C standard.
27.9.2 C library files
Change paragraph 1:
Table 133 describes header <cstdio>. [ Note: C++ does not define the function gets. —end note ]
Add a new paragraph following paragraph 1:
The contents of this header are the same as the Standard C library header <stdio.h>.
Change table 133:
clearerr fopen fsetpos putchar snprintf vscanf fclose fprintf ftell puts sprintf vsnprintf feof fputc fwrite remove sscanf vsprintf ferror fputs getc rename tmpfile vsscanf fflush fread getchar rewind tmpnam vfscanf fgetc freopen perror scanf ungetc fgetpos fscanf printf setbuf vfprintf fgets fseek putc setvbuf vprintf
Change the references to the C standard for paragraph 2:
See also: ISO C 7.9, Amendment 1 4.6.2 7.21.
Change paragraph 3:
Table 134 describes header <cinttypes>. [ Note: The macros defined by <cinttypes> are provided unconditionally. In particular, the symbol __STDC_FORMAT_MACROS, mentioned in footnote 182 of the C standard, plays no role in C++. —end note ]
C.5 C standard library
TBD, with help from LWG.
D.3 C standard library headers
Change paragraph 1:
For compatibility with the C standard library and the C Unicode TR, the C++ standard library provides the 26 22 C headers, as shown in Table 154.
Change table 154:
<assert.h> <inttypes.h> <signal.h> <stdio.h> <wchar.h> <complex.h> <iso646.h> <stdalign.h> <stdlib.h> <wctype.h> <ctype.h> <limits.h> <stdarg.h> <string.h> <errno.h> <locale.h> <stdbool.h> <tgmath.h> <fenv.h> <math.h> <stddef.h> <time.h> <float.h> <setjmp.h> <stdint.h> <uchar.h>
Since all C headers have been deprecated for some time already, it seems reasonable to delete the ones for which the corresponding C++ header is going to be deprecated.
Add a new paragraph:
The use of any of the C++ headers <ccomplex>, <cstdalign>, <cstdbool>, or <ctgmath>, is deprecated.
Related proposals
P0035R1 “Dynamic memory allocation for over-aligned data”, Clark Nelson P0175R0 “Synopses for the C library”, Thomas Köppe, Richard Smith P0270R0 “Removing C dependencies from signal handler wording”, Hans-J. BoehmTừ khóa » C99 Vs C11 Vs C17
-
What Is C17 And What Changes Have Been Made To The Language?
-
C17 (C Standard Revision) - Wikipedia
-
Is There Any Major Difference Between C18, C11 & C99? Is C ...
-
C17
-
C99, C11, C17, With C23 On The Horizon. The Only Thing Stagnant On ...
-
What's The Difference Between C – Now And Then | Electronic Design
-
[PDF] Brief Overview Of C89 Vs C99 Vs C11
-
C18 (C Standard Revision) - HandWiki
-
What Is The Difference Between C11, C17 And C18? | Sololearn
-
Standards (Using The GNU Compiler Collection (GCC))
-
P0063R0: C++17 Should Refer To C11 Instead Of C99 - Open-std
-
C11 And C17 Standard Support Arriving In MSVC - C++ Team Blog
-
About: C11 (C Standard Revision) - DBpedia