1 Conformance of the GNU libc with various standards
2 ==================================================
4 The GNU libc is designed to be conformant with existing standard as
5 far as possible. To ensure this I've run various tests. The results
12 The hdrchk test suite is available from the Open Group at
14 ftp://ftp.rdg.opengroup.org/pub/unsupported/stdtools/hdrchk/
16 I've last run the suite on 2000-08-13 on a Linux/ix86 system with the
17 following results [*]:
19 FIPS No reported problems
21 POSIX90 No reported problems
23 XPG3 No reported problems
25 XPG4 No reported problems
27 POSIX96 Same as for UNIX98 (see below).
28 UNIX98 The message queue implementation is missing:
30 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
31 /****** <mqueue.h> - Missing include file ******/
32 /****** Start of Definitions for file mqueue.h ******/
33 extern int mq_close();
34 extern int mq_getattr();
35 extern int mq_notify();
36 extern mqd_t mq_open();
37 extern ssize_t mq_receive();
39 extern int mq_setattr();
40 extern int mq_unlink();
42 struct mq_attr { <members> };
43 struct sigevent { <members> };
44 /****** End of Definitions for file mqueue.h ******/
45 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
47 [*] Since the scripts are not clever enough for the way gcc handles
48 include files (namely, putting some of them in gcc-local directory) I
49 copied over the iso646.h, float.h, and stddef.h headers and ignored the
50 problems resulting from the splitted limits.h file).
53 Technical C standards conformance issues in glibc
54 =================================================
56 If you compile programs against glibc with __STRICT_ANSI__ defined
57 (as, for example, by gcc -ansi, gcc -std=c89, gcc -std=iso1990:199409
58 or gcc -std=c99), and use only the headers specified by the version of
59 the C standard chosen, glibc will attempt to conform to that version
60 of the C standard (as indicated by __STDC_VERSION__):
62 GCC options Standard version
63 -ansi ISO/IEC 9899:1990
64 -std=c89 ISO/IEC 9899:1990
65 -std=iso9899:199409 ISO/IEC 9899:1990 as amended by Amd.1:1995 *
66 -std=c99 ISO/IEC 9899:1999
68 * glibc does not support this standard version.
70 (Note that -std=c99 is not available in GCC 2.95.2, and that no
71 version of GCC presently existing implements the full C99 standard.)
73 You may then define additional feature test macros to enable the
74 features from other standards, and use the headers defined in those
75 standards (for example, defining _POSIX_C_SOURCE to be 199506L to
76 enable features from ISO/IEC 9945-1:1996).
78 There are some technical ways in which glibc is known not to conform
79 to the supported versions of the C standard, as detailed below. Some
80 of these relate to defects in the standard that are expected to be
81 fixed, or to compiler limitations.
84 Defects in the C99 standard
85 ===========================
87 The definition of macros such as INT8_C in <stdint.h> and <inttypes.h>
88 is not implementable (Defect Report #209); this is expected to be
89 fixed in a Technical Corrigendum to make the macros yield a constant
90 expression of the promoted type (for example, int rather than char)
91 rather than needing to be able to represent constants of type char.
92 glibc follows this corrected version.
94 Several of the <fenv.h> functions are specified to return void, but
95 Defect Report #202 points out that under some circumstances they may
96 need to return an error status. They are expected to be corrected to
97 return int; glibc follows this corrected specification.
100 Implementation of library functions
101 ===================================
103 The implementation of some library functions does not fully follow the
104 standard specification:
106 C99 added additional forms of floating point constants (hexadecimal
107 constants, NaNs and infinities) to be recognised by strtod() and
108 scanf(). The effect is to change the behavior of some strictly
109 conforming C90 programs; glibc implements the C99 versions only
110 irrespective of the standard version selected.
112 C99 added %a as another scanf format specifier for floating point
113 values. This conflicts with the glibc extension where %as, %a[ and
114 %aS mean to allocate the string for the data read. A strictly
115 conforming C99 program using %as, %a[ or %aS in a scanf format string
116 will misbehave under glibc.
122 The macros __STDC_IEC_559__, __STDC_IEC_559_COMPLEX__ and
123 __STDC_ISO_10646__ are properly supposed to be defined by the
124 compiler, and to be constant throughout the translation unit (before
125 and after any library headers are included). However, they mainly
126 relate to library features, and the necessary magic has yet to be
127 implemented for GCC to predefine them to the correct values for the
128 library in use, so glibc defines them in <features.h>. Programs that
129 test them before including any standard headers may misbehave.
131 GCC doesn't support the optional imaginary types. Nor does it
132 understand the keyword _Complex before GCC 3.0. This has the
133 corresponding impact on the relevant headers.
135 glibc's use of extern inline conflicts with C99: in C99, extern inline
136 means that an external definition is generated as well as possibly an
137 inline definition, but in GCC it means that no external definition is
138 generated. When GCC's C99 mode implements C99 inline semantics, this
139 will break the uses of extern inline in glibc's headers. (Actually,
140 glibc uses `extern __inline', which is beyond the scope of the
141 standard, but it would clearly be very confusing for `__inline' and
142 plain `inline' to have different meanings in C99 mode.)
144 glibc's <tgmath.h> implementation is arcane but thought to work
145 correctly; a clean and comprehensible version requires compiler
148 For most of the headers required of freestanding implementations,
149 glibc relies on GCC to provide correct versions. (At present, glibc
150 provides <stdint.h>, and GCC doesn't.) GCC's <float.h> is missing
151 FLT_EVAL_METHOD and DECIMAL_DIG.
153 Implementing MATH_ERRNO, MATH_ERREXCEPT and math_errhandling in
154 <math.h> needs compiler support: see
156 http://sources.redhat.com/ml/libc-hacker/2000-06/msg00008.html
157 http://sources.redhat.com/ml/libc-hacker/2000-06/msg00014.html
158 http://sources.redhat.com/ml/libc-hacker/2000-06/msg00015.html
164 There are various technical issues with the definitions contained in
165 glibc's headers, listed below. The list below assumes current CVS GCC
166 as of 2001-01-10, and relates to i686-linux; older GCC may lead to
167 more problems in the headers.
169 Note that the _t suffix is reserved by POSIX, but not by pure ISO C.
170 Also, the Single Unix Specification generally requires more types to
171 be included in headers (if _XOPEN_SOURCE is defined appropriately)
174 <ctype.h> should not declare size_t.
176 <inttypes.h> should not declare wchar_t.
178 <signal.h> should not declare size_t.
180 <stdint.h> should not declare wchar_t.
182 <stdio.h> should not declare or use wchar_t or wint_t.
184 <wchar.h> does not support AMD1; to support it, the functions
185 fwprintf, fwscanf, wprintf, wscanf, swprintf, swscanf, vfwprintf,
186 vwprintf, vswprintf and fwide would need to be declared when
187 __STDC_VERSION__ >= 199409L and not just for C99.
189 <wctype.h> should not declare size_t.