1 uClibc and Glibc are not the same -- there are a number of differences which
2 may or may not cause you problems. This document attempts to list these
3 differences and, when completed, will contain a full list of all relevant
7 1) uClibc is smaller than glibc. We attempt to maintain a glibc compatible
8 interface, allowing applications that compile with glibc to easily compile with
9 uClibc. However, we do not include _everything_ that glibc includes, and
10 therefore some applications may not compile. If this happens to you, please
11 report the failure to the uclibc mailing list, with detailed error messages.
13 2) uClibc is much more configurable then glibc. This means that a developer
14 may have compiled uClibc in such a way that significant amounts of
15 functionality have been omitted.
17 3) uClibc does not even attempt to ensure binary compatibility across releases.
18 When a new version of uClibc is released, you may or may not need to recompile
21 4) malloc(0) in glibc returns a valid pointer to something(!?!?) while in
22 uClibc calling malloc(0) returns a NULL. The behavior of malloc(0) is listed
23 as implementation-defined by SuSv3, so both libraries are equally correct.
24 This difference also applies to realloc(NULL, 0). I personally feel glibc's
25 behavior is not particularly safe. To enable glibc behavior, one has to
26 explicitly enable the MALLOC_GLIBC_COMPAT option.
28 4.1) glibc's malloc() implementation has behavior that is tunable via the
29 MALLOC_CHECK_ environment variable. This is primarily used to provide extra
30 malloc debugging features. These extended malloc debugging features are not
31 available within uClibc. There are many good malloc debugging libraries
32 available for Linux (dmalloc, electric fence, valgrind, etc) that work much
33 better than the glibc extended malloc debugging. So our omitting this
34 functionality from uClibc is not a great loss.
36 5) uClibc does not provide a database library (libdb).
38 6) uClibc does not support NSS (/lib/libnss_*), which allows glibc to easily
39 support various methods of authentication and DNS resolution. uClibc only
40 supports flat password files and shadow password files for storing
41 authentication information. If you need something more complex than this,
42 you can compile and install pam.
44 7) uClibc's libresolv is only a stub. Some, but not all of the functionality
45 provided by glibc's libresolv is provided internal to uClibc. Other functions
46 are not at all implemented.
48 8) libnsl provides support for Network Information Service (NIS) which was
49 originally called "Yellow Pages" or "YP", which is an extension of RPC invented
50 by Sun to share Unix password files over the network. I personally think NIS
51 is an evil abomination and should be avoided. These days, using ldap is much
52 more effective mechanism for doing the same thing. uClibc provides a stub
53 libnsl, but has no actual support for Network Information Service (NIS).
54 We therefore, also do not provide any of the headers files provided by glibc
55 under /usr/include/rpcsvc. I am open to implementing ldap based password
56 authentication, but I do not personally intend to implement it (since I have no
59 9) uClibc's locale support is not 100% complete yet. We are working on it.
61 10) uClibc's math library only supports long double as inlines, and even
62 then the long double support is quite limited.
64 11) uClibc's libcrypt does not support the reentrant crypt_r, setkey_r and
65 encrypt_r, since these are not required by SuSv3.
67 12) uClibc directly uses the kernel types to define most opaque data types.
69 13) uClibc directly uses the linux kernel's arch specific 'stuct stat'.
71 <other things as we notice them>
75 ****************************** Manuel's Notes ******************************
77 Some general comments...
79 The intended target for all my uClibc code is ANSI/ISO C99 and SUSv3
80 compliance. While some glibc extensions are present, many will eventually
81 be configurable. Also, even when present, the glibc-like extensions may
82 differ slightly or be more restrictive than the native glibc counterparts.
83 They are primarily meant to be porting _aides_ and not necessarily
86 Now for some details...
90 1) Leap seconds are not supported.
91 2) /etc/timezone and the whole zoneinfo directory tree are not supported.
92 To set the timezone, set the TZ environment variable as specified in
93 http://www.opengroup.org/onlinepubs/007904975/basedefs/xbd_chap08.html
94 or you may also create an /etc/TZ file of a single line, ending with a
95 newline, containing the TZ setting. For example
96 echo CST6CDT > /etc/TZ
97 3) Currently, locale specific eras and alternate digits are not supported.
98 They are on my TODO list.
102 1) The only multibyte encoding currently supported is UTF-8. The various
103 ISO-8859-* encodings are (optionally) supported. The internal
104 representation of wchar's is assumed to be 31 bit unicode values in
105 native endian representation. Also, the underlying char encoding is
106 assumed to match ASCII in the range 0-0x7f.
107 2) In the next iteration of locale support, I plan to add support for
108 (at least some) other multibyte encodings.
112 1) The target for support is SUSv3 locale functionality. While nl_langinfo
113 has been extended, similar to glibc, it only returns values for related
115 2) Currently, all SUSv3 libc locale functionality should be implemented
116 except for wcsftime and collating item support in regex.
120 1) Conversion of large magnitude floating-point values by printf suffers a loss
121 of precision due to the algorithm used.
122 2) uClibc's printf is much stricter than glibcs, especially regarding positional
123 args. The entire format string is parsed first and an error is returned if
124 a problem is detected. In locales other than C, the format string is checked
125 to be a valid multibyte sequence as well. Also, currently at most 10 positional
126 args are allowed (although this is configurable).
127 3) BUFSIZ is configurable, but no attempt is made at automatic tuning of internal
128 buffer sizes for stdio streams. In fact, the stdio code in general sacrifices
129 sophistication/performace for minimal size.
130 4) uClibc allows glibc-like custom printf functions. However, while not
131 currently checked, the specifier must be <= 0x7f.
132 5) uClibc allows glibc-like custom streams. However, no in-buffer seeking is
134 6) The functions fcloseall() and __fpending() can behave differently than their
136 7) uClibc's setvbuf is more restrictive about when it can be called than glibc's
137 is. The standards specify that setvbuf must occur before any other operations
138 take place on the stream.
139 8) Right now, %m is not handled properly by printf when the format uses positional
141 9) The FILEs created by glibc's fmemopen(), open_memstream(), and fopencookie()
142 are not capable of wide orientation. The corresponding uClibc routines do
143 not have this limitation.
144 10) For scanf, the C99 standard states "The fscanf function returns the value of
145 the macro EOF if an input failure occurs before any conversion." But glibc's
146 scanf does not respect conversions for which assignment was surpressed, even
147 though the standard states that the value is converted but not stored.
149 glibc bugs that Ulrich Drepper has refused to acknowledge or comment on
150 ( http://sources.redhat.com/ml/libc-alpha/2003-09/ )
151 -----------------------------------------------------------------------
152 1) The C99 standard says that for printf, a %s conversion makes no special
153 provisions for multibyte characters. SUSv3 is even more clear, stating
154 that bytes are written and a specified precision is in bytes. Yet glibc
155 treats the arg as a multibyte string when a precision is specified and
157 2) Both C99 and C89 state that the %c conversion for scanf reads the exact
158 number of bytes specified by the optional field width (or 1 if not specified).
159 uClibc complies with the standard. There is an argument that perhaps the
160 specified width should be treated as an upper bound, based on some historical
161 use. However, such behavior should be mentioned in the Conformance document.
162 3) glibc's scanf is broken regarding some numeric patterns. Some invalid
163 strings are accepted as valid ("0x.p", "1e", digit grouped strings).
164 In spite of my posting examples clearly illustrating the bugs, they remain
165 unacknowledged by the glibc developers.
166 4) glibc's scanf seems to require a 'p' exponent for hexadecimal float strings.
167 According to the standard, this is optional.
168 5) C99 requires that once an EOF is encountered, the stream should be treated
169 as if at end-of-file even if more data becomes available. Further reading
170 can be attempted by clearing the EOF flag though, via clearerr() or a file
171 positioning function. For details concerning the original change, see
172 Defect Report #141. glibc is currently non-compliant, and the developers
173 did not comment when I asked for their official position on this issue.
174 6) glibc's collation routines and/or localedef are broken regarding implicit
175 and explicit UNDEFINED rules.
177 More to follow as I think of it...