C: Difference between revisions
Line 487: | Line 487: | ||
=== Standard libraries === |
=== Standard libraries === |
||
;ctype.h |
|||
* Provides character classification functions like <code>isdigit</code>, <code>islower</code>, <code>isupper</code>, <code>isalpha</code>... |
|||
;err.h |
;err.h |
||
*Provides standard error reporting functions like <code>err</code>, <code>errx</code>... These replaces the need for custom <code>die(int errcode, const char * fmt, ...)</code> functions. |
|||
<source lang="c"> |
<source lang="c"> |
||
if ( bind(sockfd, (struct sockaddr *)&server, sizeof(server)) < 0 ) { |
if ( bind(sockfd, (struct sockaddr *)&server, sizeof(server)) < 0 ) { |
Revision as of 06:31, 5 October 2018
References
- The C Library Reference Guide
- C Preprocessor reference
- manpage manpages-dev
- manpage manpages-posix-dev
sudo apt-get install manpages-dev manpages-posix-dev
- On this wiki
Examples
See C examples.
Tips
Designated Struct Initializer
struct {
int a;
int b;
} s = { .a = 1, .b = 2 };
Same applies for union.
See GCC or StackOverflow.
Variadic Macros
See reference here.
#define eprintf(...) fprintf (stderr, __VA_ARGS__)
/* eprintf ("%s:%d: ", input_file, lineno)
==> fprintf (stderr, "%s:%d: ", input_file, lineno) */
#define eprintf(args...) fprintf (stderr, args)
#define eprintf(format, ...) fprintf (stderr, format, __VA_ARGS__)
An handy macro hack that counts the number of parameters passed before expansion (See [1] and [2]):
/* The PP_NARG macro returns the number of arguments that have been
* passed to it. This compensates for lack of __VA_NARGS__.
* Macros written by Laurent Deniau See http://en.wikipedia.org/wiki/Variadic_macro.
*/
#define PP_NARG(...) \
PP_NARG_(__VA_ARGS__,PP_RSEQ_N())
#define PP_NARG_(...) \
PP_ARG_N(__VA_ARGS__)
#define PP_ARG_N( \
_1, _2, _3, _4, _5, _6, _7, _8, _9,_10, \
_11,_12,_13,_14,_15,_16,_17,_18,_19,_20, \
_21,_22,_23,_24,_25,_26,_27,_28,_29,_30, \
_31,_32,_33,_34,_35,_36,_37,_38,_39,_40, \
_41,_42,_43,_44,_45,_46,_47,_48,_49,_50, \
_51,_52,_53,_54,_55,_56,_57,_58,_59,_60, \
_61,_62,_63,N,...) N
#define PP_RSEQ_N() \
63,62,61,60, \
59,58,57,56,55,54,53,52,51,50, \
49,48,47,46,45,44,43,42,41,40, \
39,38,37,36,35,34,33,32,31,30, \
29,28,27,26,25,24,23,22,21,20, \
19,18,17,16,15,14,13,12,11,10, \
9,8,7,6,5,4,3,2,1,0
/* Some test cases */
PP_NARG(A) -> 1
PP_NARG(A,B) -> 2
PP_NARG(A,B,C) -> 3
PP_NARG(A,B,C,D) -> 4
PP_NARG(A,B,C,D,E) -> 5
PP_NARG(1,2,3,4,5,6,7,8,9,0,
1,2,3,4,5,6,7,8,9,0,
1,2,3,4,5,6,7,8,9,0,
1,2,3,4,5,6,7,8,9,0,
1,2,3,4,5,6,7,8,9,0,
1,2,3,4,5,6,7,8,9,0,
1,2,3) -> 63
Variadic I/O Functions
See [3] for some reference information on variable functions.
The standard variadic I/O functions are:
#include <stdio.h>
#include <stdarg.h>
int vprintf(const char *format, va_list ap);
int vfprintf(FILE *stream, const char *format, va_list ap);
int vsprintf(char *str, const char *format, va_list ap);
int vsnprintf(char *str, size_t size, const char *format, va_list ap);
Example of use:
//This works both on win32 / linux
#include <stdio.h>
#include <stdarg.h>
void vario(const char* formatstr,...)
{
va_list args;
va_start(args, formatstr);
vprintf(formatstr, args);
va_end(args); /* cleanup - DON'T FORGET */
}
Note on win32:
- There exists alternative functions ([4]) prefixed with a '_', such as
_vprintf_p
,_vfprintf_l
, ...
Temporary variable names for Macro
/* UNIQ(x) creates a unique variable name that depends on the current source line as returned by __LINE__. We need
Several intermediate macros because identifier are not expanded in macro if they are used along with # or ## in
macro definition.
Example: #define SCAN_MY(var,n) {int UNIQ(x); for(UNIQ(x)=0; UNIQ(x)<n; ++UNIQ(x)) printf(var[UNIQ(x)]);}
*/
#define UNIQ__(x,y) x ## y
#define UNIQ_(x,y) UNIQ__(x,y)
#define UNIQ(x) UNIQ_(x,__LINE__)
Static assertions
Static assertions are assertions that can be verified at compile time.
From Static_assertions (WP):
#define SASSERT(pred) switch(0){case 0:case pred:;}
SASSERT( BOOLEAN CONDITION );
static char const static_assertion[ (BOOLEAN CONDITION)
? 1 : -1
] = {'!'};
Variadic macros
From GNU CC Manual - Variadic Macros.
#define eprintf(...) fprintf (stderr, __VA_ARGS__) // C99 Standard
#define eprintf(args...) fprintf (stderr, args) // GNU extension
#define eprintf(args...) fprintf (stderr, ## args) // GNU extension
Some caveats:
#define eprintf(format,...) fprintf (stderr, format, __VA_ARGS__)
Looks more descriptive, but now at least one argument must be provided, except if compiler supports the construction (gcc, vc).
Static assert
Static assert that supports sizeof()
operator: http://www.pixelbeat.org/programming/gcc/static_assert.html.
#define ASSERT_CONCAT_(a, b) a##b
#define ASSERT_CONCAT(a, b) ASSERT_CONCAT_(a, b)
/* These can't be used after statements in c89. */
#ifdef __COUNTER__
#define STATIC_ASSERT(e,m) \
;enum { ASSERT_CONCAT(static_assert_, __COUNTER__) = 1/(int)(!!(e)) }
#else
/* This can't be used twice on the same line so ensure if using in headers
* that the headers are not included twice (by wrapping in #ifndef...#endif)
* Note it doesn't cause an issue when used on same line of separate modules
* compiled with gcc -combine -fwhole-program. */
#define STATIC_ASSERT(e,m) \
;enum { ASSERT_CONCAT(assert_line_, __LINE__) = 1/(int)(!!(e)) }
#endif
Call a function at a given absolute address
Using a nice typedef
(SO):
typedef int func(void);
func* f = (func*)0xdeadbeef;
int i = f();
Avoiding the assignment:
typedef int func(void);
int i = ((func*)0xdeadbeef)();
All in one line:
int i = ((int (*)(void))0xdeadbeef)();
Complex or unusual type-casting
// Multi-dimensional arrays
uint8_t array[20][8]; // array is uint8 (*)[8]
uint8_t (*p)[8];
p = array; // CORRECT!
Deal with misaligned data in a portable way
The recommended solution is to use memcpy
[5]:
_Bool check_ip_header_sum (const char * p, size_t size)
{
uint32_t temp;
uint64_t sum = 0;
memcpy (&temp, p, 4); sum += temp;
memcpy (&temp, p + 4, 4); sum += temp;
memcpy (&temp, p + 8, 4); sum += temp;
memcpy (&temp, p + 12, 4); sum += temp;
memcpy (&temp, p + 16, 4); sum += temp;
for (size_t i = 20; i < size; i+= 4) {
memcpy (&temp, p + i, 4);
sum += temp;
}
do {
sum = (sum & 0xFFFF) + (sum >> 16);
} while (sum & ~0xFFFFL);
return sum == 0xFFFF;
}
Using C99 compound literals
C99 constant literal are useful to build local constants without the need to create a variable for them. For instance, from SO:
if (setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR, &(int){ 1 }, sizeof(int)) < 0)
error("setsockopt(SO_REUSEADDR) failed");
We use the compound literal (int){ 1 }
to create a constant variable and then return its address. The alternative would be to define a variable:
int enable = 1;
if (setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR, &enable, sizeof(enable)) < 0)
error("setsockopt(SO_REUSEADDR) failed");
Parse integer
Use strtol
:
char * endptr;
port = strtol(optarg, &endptr, 10); // *endptr == '\x00' if all bytes parsed
if (errno || *endptr || port > 65535) {
fprintf("Invalid port number %s.\n", optarg);
}
Pits
Multiple definition of uninitialized variables
By default, gcc allow multiple definition of uninitialized variables! This is really EVIL. Solutions:
References:
|
Let's consider foo.c int foo;
bar.c int bar=0;
and main.c int foo;
int bar=0;
int main(int argc, char** argv)
{
return foo+bar;
}
We have the following: $ gcc main.c bar.c
# /tmp/ccjIc7Is.o:(.bss+0x0): multiple definition of `bar'
# /tmp/ccFys0wm.o:(.bss+0x0): first defined here
# collect2: ld returned 1 exit status
gcc main.c foo.c
# No error !!!
gcc does not report any error for duplicate definitions of uninitialized variables. We can force it as follows: gcc -fno-common main.c foo.c
# /tmp/ccnMOLlf.o:(.bss+0x0): multiple definition of `foo'
# /tmp/cc5YeNxE.o:(.bss+0x0): first defined here
# collect2: ld returned 1 exit status
gcc -Wl,--warn-common main.c foo.c
# /tmp/ccm9Z06a.o: warning: multiple common of `foo'
# /tmp/ccJIfrYq.o: warning: previous common is here
|
Undefined reference
Undefined reference with gcc although library is given gcc -lm mymath.c -o mymath /tmp/ccI4NbJJ.o: In function `V': mymath.c:(.text+0x235): undefined reference to `pow' |
The solution is to have the -lm at the end. Order does make a difference (at least since gcc 4.6). gcc mymath.c -o mymath -lm |
No sequence point in assignment operator
The following code is wrong [6]:
uint32_t
swaphalves(uint32_t a)
{
a = (a >>= 16) | (a <<= 16);
return a;
}
There is no sequence point here, and so we don't know anything about the order of operations here. Of course the code can be easily fixed:
uint32_t
swaphalves(uint32_t a)
{
return (a >> 16) | (a << 16);
}
More on sequence points [7], [8], [9] (! links for C++, there might be difference with C).
Strict Aliasing
See [10]
Addr-of array, &a
int a[4];
int *p;
p=&a+2; // WRONG - ACCESS (void *)a + 2*sizeof(a)
p=a+2; // CORRECT - ACCESS (void *)a + 2*sizeof(int)
p=&a[2]; // CORRECT
*p=1;
Handy macros
#define NUM(A) ( sizeof(A) / sizeof((A)[0]) ) // Number of elements in an array
How-To
First we must compile with Position-Independent Code (PIC):
gcc -c -Wall -Werror -fpic foo.c
We use -shared
to generate a dynamically loaded library:
gcc -shared -o libfoo.so foo.o
To link with the library, we must usually gives the location of the library:
$ gcc -Wall -o test main.c -lfoo
# /usr/bin/ld: cannot find -lfoo
# collect2: ld returned 1 exit status
gcc -L/home/username/foo -Wall -o test main.c -lfoo
Likewise, we must tell the location of the library at runtime using LD_LIBRARY_PATH
:
./test
# ./test: error while loading shared libraries: libfoo.so: cannot open shared object file: No such file or directory
LD_LIBRARY_PATH=/home/username/foo:$LD_LIBRARY_PATH ./test
We use rpath
at compilation-time to hardcode the library path and avoids the use of LD_LIBRARY_PATH
(but also the cost of system configuration flexibility):
gcc -L/home/username/foo -Wl,-rpath=/home/username/foo -Wall -o test main.c -lfoo
./test
The best option is to make the library available system-wide using ldconfig
:
cp /home/username/foo/libfoo.so /usr/lib
chmod 0755 /usr/lib/libfoo.so
ldconfig
ldconfig -p | grep foo
gcc -Wall -o test main.c -lfoo # We don't need -L ... anymore
ldd test | grep foo
# libfoo.so => /usr/lib/libfoo.so (0x00a42000)
./test # We don't need LD_LIBRARY_PATH
# This is a shared library test...
# Hello, I'm a shared library
We can build Windows DLL with:
- MinGW gcc on Windows (MSYS/MinGW).
- MinGW gcc from package mingw-w64 on Debian / Ubuntu.
The recipe is similar to building a library on Linux. We only need to export or import explicitly each DLL entries using __declspec
statements [11].
Content of example_dll.cpp (see http://www.mingw.org/wiki/sampledll for more export examples):
<
#define EXPORT_DLL
#include "example_dll.h"
int Double(int x)
{
return 2 * x;
}
Content of example_dll.h:
#ifndef EXAMPLE_DLL_H
#define EXAMPLE_DLL_H
#if defined(EXPORT_DLL)
#define DLL_DECL __declspec(dllexport)
#else
#define DLL_DECL __declspec(dllimport)
#endif
extern "C" int DLL_DECL Double(int x);
#endif // EXAMPLE_DLL_H
To build the DLL:
g++ -c example_dll.cpp
g++ -shared -o example_dll.dll example_dll.o # We build for the MINGW toolchain only
Note: If need to use the DLL with another toolchain than MinGW, we must add -Wl,--out-implib,libexample_dll.a
to the linker call [12].
To use our DLL, content of example_exe.cpp:
#include <stdio.h>
#include "example_dll.h"
int main(void)
{
printf("%d\n", Double(333));
return 0;
}
We build our example:
g++ -c example_exe.cpp
g++ -o example_exe.exe example_exe.o -L. -lexample_dll
- Building / linking without the __declspec declarations
- Use
-no-undefined
at build time to export all functions into the DLL. - Use
--enable-runtime-pseudo-reloc
to import all functions automaticaly.
Libraries
Standard libraries
- ctype.h
- Provides character classification functions like
isdigit
,islower
,isupper
,isalpha
...
- err.h
- Provides standard error reporting functions like
err
,errx
... These replaces the need for customdie(int errcode, const char * fmt, ...)
functions.
if ( bind(sockfd, (struct sockaddr *)&server, sizeof(server)) < 0 ) {
err(1, "bind failed");
}
- unistd.h
- Provides
getopt
for parsing of command-line parameters (see C examples).
P99
P99 is not a regular library but instead a set of include files that offer powerful preprocessor macros (like P99_DUPL(...)
, P99_NARG(...)). From the website:
P99 is not a C library in the classical sense but merely a collection of include files:
- There is no binary library to be linked to your executable. The few functions that are provided are small wrappers that are compiled directly into your code.
- There is nothing to configure, P99 include files should work out of the box with any conforming C99 compiler.
Debugging
See general page on Debugging.
Embedded C
Some tips to make embedded development with C.
Enum are not always int
Some compilers do not necessarily map enum types to int. The standard says it is platform-dependent. Some compiler may use (or be configured to use) the smallest date type that can hold the value of all enumerators.
To prevent strange behaviour, always enforce the enum size as follows:
typedef enum MyEnum_t {
first = 1,
second = 2,
_forceint = 0x7FFFFFFF // To force 32-bit (signed) enum
} MyEnum_t;
Using packed enum
Some compiler supports storing enum using the smallest data type (signed or unsigned):
- Command-line options
- On gcc, use option
--short-enums
or-fshort-enums
:
gcc --short-enums -c file.c
gcc -fshort-enums -c file.c
- On clang, use option
-fshort-enums
:
clang -fshort-enums -c file.c
- ARM compilers have similar options
- Source code
- gcc and alike: Use pragma
__attribute__((packed))
typedef enum {a,b} __atribute__((packed)) letters;
- VC: Support only packed structure
#pragma pack(push,1)
typedef struct {
letters x;
}
#pragma pack(pop)
Win32 vs Linux
Linux | Win32 | Comment |
---|---|---|
snprintf |
_snprintf |
deprecated, replaced by _snprintf_s but not compatible
|