]> git.ipfire.org Git - thirdparty/gcc.git/blob - gcc/README.Portability
[Ada] Get rid of more references to Universal_Integer in expanded code
[thirdparty/gcc.git] / gcc / README.Portability
1 Copyright (C) 2000-2020 Free Software Foundation, Inc.
2
3 This file is intended to contain a few notes about writing C code
4 within GCC so that it compiles without error on the full range of
5 compilers GCC needs to be able to compile on.
6
7 The problem is that many ISO-standard constructs are not accepted by
8 either old or buggy compilers, and we keep getting bitten by them.
9 This knowledge until now has been sparsely spread around, so I
10 thought I'd collect it in one useful place. Please add and correct
11 any problems as you come across them.
12
13 I'm going to start from a base of the ISO C90 standard, since that is
14 probably what most people code to naturally. Obviously using
15 constructs introduced after that is not a good idea.
16
17 For the complete coding style conventions used in GCC, please read
18 http://gcc.gnu.org/codingconventions.html
19
20
21 String literals
22 ---------------
23
24 Some compilers like MSVC++ have fairly low limits on the maximum
25 length of a string literal; 509 is the lowest we've come across. You
26 may need to break up a long printf statement into many smaller ones.
27
28
29 Empty macro arguments
30 ---------------------
31
32 ISO C (6.8.3 in the 1990 standard) specifies the following:
33
34 If (before argument substitution) any argument consists of no
35 preprocessing tokens, the behavior is undefined.
36
37 This was relaxed by ISO C99, but some older compilers emit an error,
38 so code like
39
40 #define foo(x, y) x y
41 foo (bar, )
42
43 needs to be coded in some other way.
44
45
46 Avoid unnecessary test before free
47 ----------------------------------
48
49 Since SunOS 4 stopped being a reasonable portability target,
50 (which happened around 2007) there has been no need to guard
51 against "free (NULL)". Thus, any guard like the following
52 constitutes a redundant test:
53
54 if (P)
55 free (P);
56
57 It is better to avoid the test.[*]
58 Instead, simply free P, regardless of whether it is NULL.
59
60 [*] However, if your profiling exposes a test like this in a
61 performance-critical loop, say where P is nearly always NULL, and
62 the cost of calling free on a NULL pointer would be prohibitively
63 high, consider using __builtin_expect, e.g., like this:
64
65 if (__builtin_expect (ptr != NULL, 0))
66 free (ptr);
67
68
69
70 Trigraphs
71 ---------
72
73 You weren't going to use them anyway, but some otherwise ISO C
74 compliant compilers do not accept trigraphs.
75
76
77 Suffixes on Integer Constants
78 -----------------------------
79
80 You should never use a 'l' suffix on integer constants ('L' is fine),
81 since it can easily be confused with the number '1'.
82
83
84 Common Coding Pitfalls
85 ======================
86
87 errno
88 -----
89
90 errno might be declared as a macro.
91
92
93 Implicit int
94 ------------
95
96 In C, the 'int' keyword can often be omitted from type declarations.
97 For instance, you can write
98
99 unsigned variable;
100
101 as shorthand for
102
103 unsigned int variable;
104
105 There are several places where this can cause trouble. First, suppose
106 'variable' is a long; then you might think
107
108 (unsigned) variable
109
110 would convert it to unsigned long. It does not. It converts to
111 unsigned int. This mostly causes problems on 64-bit platforms, where
112 long and int are not the same size.
113
114 Second, if you write a function definition with no return type at
115 all:
116
117 operate (int a, int b)
118 {
119 ...
120 }
121
122 that function is expected to return int, *not* void. GCC will warn
123 about this.
124
125 Implicit function declarations always have return type int. So if you
126 correct the above definition to
127
128 void
129 operate (int a, int b)
130 ...
131
132 but operate() is called above its definition, you will get an error
133 about a "type mismatch with previous implicit declaration". The cure
134 is to prototype all functions at the top of the file, or in an
135 appropriate header.
136
137 Char vs unsigned char vs int
138 ----------------------------
139
140 In C, unqualified 'char' may be either signed or unsigned; it is the
141 implementation's choice. When you are processing 7-bit ASCII, it does
142 not matter. But when your program must handle arbitrary binary data,
143 or fully 8-bit character sets, you have a problem. The most obvious
144 issue is if you have a look-up table indexed by characters.
145
146 For instance, the character '\341' in ISO Latin 1 is SMALL LETTER A
147 WITH ACUTE ACCENT. In the proper locale, isalpha('\341') will be
148 true. But if you read '\341' from a file and store it in a plain
149 char, isalpha(c) may look up character 225, or it may look up
150 character -31. And the ctype table has no entry at offset -31, so
151 your program will crash. (If you're lucky.)
152
153 It is wise to use unsigned char everywhere you possibly can. This
154 avoids all these problems. Unfortunately, the routines in <string.h>
155 take plain char arguments, so you have to remember to cast them back
156 and forth - or avoid the use of strxxx() functions, which is probably
157 a good idea anyway.
158
159 Another common mistake is to use either char or unsigned char to
160 receive the result of getc() or related stdio functions. They may
161 return EOF, which is outside the range of values representable by
162 char. If you use char, some legal character value may be confused
163 with EOF, such as '\377' (SMALL LETTER Y WITH UMLAUT, in Latin-1).
164 The correct choice is int.
165
166 A more subtle version of the same mistake might look like this:
167
168 unsigned char pushback[NPUSHBACK];
169 int pbidx;
170 #define unget(c) (assert(pbidx < NPUSHBACK), pushback[pbidx++] = (c))
171 #define get(c) (pbidx ? pushback[--pbidx] : getchar())
172 ...
173 unget(EOF);
174
175 which will mysteriously turn a pushed-back EOF into a SMALL LETTER Y
176 WITH UMLAUT.
177
178
179 Other common pitfalls
180 ---------------------
181
182 o Expecting 'plain' char to be either sign or unsigned extending.
183
184 o Shifting an item by a negative amount or by greater than or equal to
185 the number of bits in a type (expecting shifts by 32 to be sensible
186 has caused quite a number of bugs at least in the early days).
187
188 o Expecting ints shifted right to be sign extended.
189
190 o Modifying the same value twice within one sequence point.
191
192 o Host vs. target floating point representation, including emitting NaNs
193 and Infinities in a form that the assembler handles.
194
195 o qsort being an unstable sort function (unstable in the sense that
196 multiple items that sort the same may be sorted in different orders
197 by different qsort functions).
198
199 o Passing incorrect types to fprintf and friends.
200
201 o Adding a function declaration for a module declared in another file to
202 a .c file instead of to a .h file.