255

Trying to create a macro which can be used for print debug messages when DEBUG is defined, like the following pseudo code:

#define DEBUG 1
#define debug_print(args ...) if (DEBUG) fprintf(stderr, args)

How is this accomplished with a macro?

jww
  • 97,681
  • 90
  • 411
  • 885
jfarrell
  • 4,480
  • 3
  • 21
  • 14
  • 1
    Will the compiler (gcc) optimize statements like if(DEBUG) {...} out, if in production code the DEBUG macro is set to 0 ? I understand that there are good reasons to leave the debug statements visible to the compiler, but a bad feeling remains. -Pat – Pat Feb 02 '10 at 10:48
  • 1
    Also see [Error when defining a stringising macro with \_\_VA_ARGS__](https://stackoverflow.com/q/18968070/608639) – jww Feb 25 '19 at 05:40

14 Answers14

488

If you use a C99 or later compiler

#define debug_print(fmt, ...) \
            do { if (DEBUG) fprintf(stderr, fmt, __VA_ARGS__); } while (0)

It assumes you are using C99 (the variable argument list notation is not supported in earlier versions). The do { ... } while (0) idiom ensures that the code acts like a statement (function call). The unconditional use of the code ensures that the compiler always checks that your debug code is valid — but the optimizer will remove the code when DEBUG is 0.

If you want to work with #ifdef DEBUG, then change the test condition:

#ifdef DEBUG
#define DEBUG_TEST 1
#else
#define DEBUG_TEST 0
#endif

And then use DEBUG_TEST where I used DEBUG.

If you insist on a string literal for the format string (probably a good idea anyway), you can also introduce things like __FILE__, __LINE__ and __func__ into the output, which can improve the diagnostics:

#define debug_print(fmt, ...) \
        do { if (DEBUG) fprintf(stderr, "%s:%d:%s(): " fmt, __FILE__, \
                                __LINE__, __func__, __VA_ARGS__); } while (0)

This relies on string concatenation to create a bigger format string than the programmer writes.

If you use a C89 compiler

If you are stuck with C89 and no useful compiler extension, then there isn't a particularly clean way to handle it. The technique I used to use was:

#define TRACE(x) do { if (DEBUG) dbg_printf x; } while (0)

And then, in the code, write:

TRACE(("message %d\n", var));

The double-parentheses are crucial — and are why you have the funny notation in the macro expansion. As before, the compiler always checks the code for syntactic validity (which is good) but the optimizer only invokes the printing function if the DEBUG macro evaluates to non-zero.

This does require a support function — dbg_printf() in the example — to handle things like 'stderr'. It requires you to know how to write varargs functions, but that isn't hard:

#include <stdarg.h>
#include <stdio.h>

void dbg_printf(const char *fmt, ...)
{
    va_list args;
    va_start(args, fmt);
    vfprintf(stderr, fmt, args);
    va_end(args);
}

You can also use this technique in C99, of course, but the __VA_ARGS__ technique is neater because it uses regular function notation, not the double-parentheses hack.

Why is it crucial that the compiler always see the debug code?

[Rehashing comments made to another answer.]

One central idea behind both the C99 and C89 implementations above is that the compiler proper always sees the debugging printf-like statements. This is important for long-term code — code that will last a decade or two.

Suppose a piece of code has been mostly dormant (stable) for a number of years, but now needs to be changed. You re-enable debugging trace - but it is frustrating to have to debug the debugging (tracing) code because it refers to variables that have been renamed or retyped, during the years of stable maintenance. If the compiler (post pre-processor) always sees the print statement, it ensures that any surrounding changes have not invalidated the diagnostics. If the compiler does not see the print statement, it cannot protect you against your own carelessness (or the carelessness of your colleagues or collaborators). See 'The Practice of Programming' by Kernighan and Pike, especially Chapter 8 (see also Wikipedia on TPOP).

This is 'been there, done that' experience — I used essentially the technique described in other answers where the non-debug build does not see the printf-like statements for a number of years (more than a decade). But I came across the advice in TPOP (see my previous comment), and then did enable some debugging code after a number of years, and ran into problems of changed context breaking the debugging. Several times, having the printing always validated has saved me from later problems.

I use NDEBUG to control assertions only, and a separate macro (usually DEBUG) to control whether debug tracing is built into the program. Even when the debug tracing is built in, I frequently do not want debug output to appear unconditionally, so I have mechanism to control whether the output appears (debug levels, and instead of calling fprintf() directly, I call a debug print function that only conditionally prints so the same build of the code can print or not print based on program options). I also have a 'multiple-subsystem' version of the code for bigger programs, so that I can have different sections of the program producing different amounts of trace - under runtime control.

I am advocating that for all builds, the compiler should see the diagnostic statements; however, the compiler won't generate any code for the debugging trace statements unless debug is enabled. Basically, it means that all of your code is checked by the compiler every time you compile - whether for release or debugging. This is a good thing!

debug.h - version 1.2 (1990-05-01)

/*
@(#)File:            $RCSfile: debug.h,v $
@(#)Version:         $Revision: 1.2 $
@(#)Last changed:    $Date: 1990/05/01 12:55:39 $
@(#)Purpose:         Definitions for the debugging system
@(#)Author:          J Leffler
*/

#ifndef DEBUG_H
#define DEBUG_H

/* -- Macro Definitions */

#ifdef DEBUG
#define TRACE(x)    db_print x
#else
#define TRACE(x)
#endif /* DEBUG */

/* -- Declarations */

#ifdef DEBUG
extern  int     debug;
#endif

#endif  /* DEBUG_H */

debug.h - version 3.6 (2008-02-11)

/*
@(#)File:           $RCSfile: debug.h,v $
@(#)Version:        $Revision: 3.6 $
@(#)Last changed:   $Date: 2008/02/11 06:46:37 $
@(#)Purpose:        Definitions for the debugging system
@(#)Author:         J Leffler
@(#)Copyright:      (C) JLSS 1990-93,1997-99,2003,2005,2008
@(#)Product:        :PRODUCT:
*/

#ifndef DEBUG_H
#define DEBUG_H

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif /* HAVE_CONFIG_H */

/*
** Usage:  TRACE((level, fmt, ...))
** "level" is the debugging level which must be operational for the output
** to appear. "fmt" is a printf format string. "..." is whatever extra
** arguments fmt requires (possibly nothing).
** The non-debug macro means that the code is validated but never called.
** -- See chapter 8 of 'The Practice of Programming', by Kernighan and Pike.
*/
#ifdef DEBUG
#define TRACE(x)    db_print x
#else
#define TRACE(x)    do { if (0) db_print x; } while (0)
#endif /* DEBUG */

#ifndef lint
#ifdef DEBUG
/* This string can't be made extern - multiple definition in general */
static const char jlss_id_debug_enabled[] = "@(#)*** DEBUG ***";
#endif /* DEBUG */
#ifdef MAIN_PROGRAM
const char jlss_id_debug_h[] = "@(#)$Id: debug.h,v 3.6 2008/02/11 06:46:37 jleffler Exp $";
#endif /* MAIN_PROGRAM */
#endif /* lint */

#include <stdio.h>

extern int      db_getdebug(void);
extern int      db_newindent(void);
extern int      db_oldindent(void);
extern int      db_setdebug(int level);
extern int      db_setindent(int i);
extern void     db_print(int level, const char *fmt,...);
extern void     db_setfilename(const char *fn);
extern void     db_setfileptr(FILE *fp);
extern FILE    *db_getfileptr(void);

/* Semi-private function */
extern const char *db_indent(void);

/**************************************\
** MULTIPLE DEBUGGING SUBSYSTEMS CODE **
\**************************************/

/*
** Usage:  MDTRACE((subsys, level, fmt, ...))
** "subsys" is the debugging system to which this statement belongs.
** The significance of the subsystems is determined by the programmer,
** except that the functions such as db_print refer to subsystem 0.
** "level" is the debugging level which must be operational for the
** output to appear. "fmt" is a printf format string. "..." is
** whatever extra arguments fmt requires (possibly nothing).
** The non-debug macro means that the code is validated but never called.
*/
#ifdef DEBUG
#define MDTRACE(x)  db_mdprint x
#else
#define MDTRACE(x)  do { if (0) db_mdprint x; } while (0)
#endif /* DEBUG */

extern int      db_mdgetdebug(int subsys);
extern int      db_mdparsearg(char *arg);
extern int      db_mdsetdebug(int subsys, int level);
extern void     db_mdprint(int subsys, int level, const char *fmt,...);
extern void     db_mdsubsysnames(char const * const *names);

#endif /* DEBUG_H */

Single argument variant for C99 or later

Kyle Brandt asked:

Anyway to do this so debug_print still works even if there are no arguments? For example:

    debug_print("Foo");

There's one simple, old-fashioned hack:

debug_print("%s\n", "Foo");

The GCC-only solution shown below also provides support for that.

However, you can do it with the straight C99 system by using:

#define debug_print(...) \
            do { if (DEBUG) fprintf(stderr, __VA_ARGS__); } while (0)

Compared to the first version, you lose the limited checking that requires the 'fmt' argument, which means that someone could try to call 'debug_print()' with no arguments (but the trailing comma in the argument list to fprintf() would fail to compile). Whether the loss of checking is a problem at all is debatable.

GCC-specific technique for a single argument

Some compilers may offer extensions for other ways of handling variable-length argument lists in macros. Specifically, as first noted in the comments by Hugo Ideler, GCC allows you to omit the comma that would normally appear after the last 'fixed' argument to the macro. It also allows you to use ##__VA_ARGS__ in the macro replacement text, which deletes the comma preceding the notation if, but only if, the previous token is a comma:

#define debug_print(fmt, ...) \
            do { if (DEBUG) fprintf(stderr, fmt, ##__VA_ARGS__); } while (0)

This solution retains the benefit of requiring the format argument while accepting optional arguments after the format.

This technique is also supported by Clang for GCC compatibility.


Why the do-while loop?

What's the purpose of the do while here?

You want to be able to use the macro so it looks like a function call, which means it will be followed by a semi-colon. Therefore, you have to package the macro body to suit. If you use an if statement without the surrounding do { ... } while (0), you will have:

/* BAD - BAD - BAD */
#define debug_print(...) \
            if (DEBUG) fprintf(stderr, __VA_ARGS__)

Now, suppose you write:

if (x > y)
    debug_print("x (%d) > y (%d)\n", x, y);
else
    do_something_useful(x, y);

Unfortunately, that indentation doesn't reflect the actual control of flow, because the preprocessor produces code equivalent to this (indented and braces added to emphasize the actual meaning):

if (x > y)
{
    if (DEBUG)
        fprintf(stderr, "x (%d) > y (%d)\n", x, y);
    else
        do_something_useful(x, y);
}

The next attempt at the macro might be:

/* BAD - BAD - BAD */
#define debug_print(...) \
            if (DEBUG) { fprintf(stderr, __VA_ARGS__); }

And the same code fragment now produces:

if (x > y)
    if (DEBUG)
    {
        fprintf(stderr, "x (%d) > y (%d)\n", x, y);
    }
; // Null statement from semi-colon after macro
else
    do_something_useful(x, y);

And the else is now a syntax error. The do { ... } while(0) loop avoids both these problems.

There's one other way of writing the macro which might work:

/* BAD - BAD - BAD */
#define debug_print(...) \
            ((void)((DEBUG) ? fprintf(stderr, __VA_ARGS__) : 0))

This leaves the program fragment shown as valid. The (void) cast prevents it being used in contexts where a value is required — but it could be used as the left operand of a comma operator where the do { ... } while (0) version cannot. If you think you should be able to embed debug code into such expressions, you might prefer this. If you prefer to require the debug print to act as a full statement, then the do { ... } while (0) version is better. Note that if the body of the macro involved any semi-colons (roughly speaking), then you can only use the do { ... } while(0) notation. It always works; the expression statement mechanism can be more difficult to apply. You might also get warnings from the compiler with the expression form that you'd prefer to avoid; it will depend on the compiler and the flags you use.


TPOP was previously at http://plan9.bell-labs.com/cm/cs/tpop and http://cm.bell-labs.com/cm/cs/tpop but both are now (2015-08-10) broken.


Code in GitHub

If you're curious, you can look at this code in GitHub in my SOQ (Stack Overflow Questions) repository as files debug.c, debug.h and mddebug.c in the src/libsoq sub-directory.

Jonathan Leffler
  • 730,956
  • 141
  • 904
  • 1,278
  • Anyway to do this so debug_print still works even if there are not arguments? IE debug_print("Foo"); ? – Kyle Brandt Oct 30 '09 at 14:53
  • There's always the old-fashioned hack: debug_print("%s\n", "Foo"); :D I'll try to work out whether there's an alternative. The GCC-only solution provides support for that. – Jonathan Leffler Oct 30 '09 at 14:59
  • What's the purpose of the `do while` here? – cYrus Jan 14 '12 at 02:19
  • @cYrus As Jonathan said in his answer, the purpose of the do { } while(0) is to trick C99 compilers into checking the debug code. Without it, the compiler might see that the condition in the if(DEBUG) is false and skip the code. This way the probability of bugs in the debug code is reduced. – Thomas Dignan Feb 20 '12 at 00:31
  • 4
    Years later, and this answer is still the most useful out of all the internets, on how to alias printk! vfprintf does not work in kernel space since stdio is not available. Thank you! `#define debug(...) \ do { if (DEBUG) \ printk("DRIVER_NAME:"); \ printk(__VA_ARGS__); \ printk("\n"); \ } while (0)` – Kevin Jul 18 '13 at 21:07
  • 9
    In your example with the keywords `__FILE__, __LINE__, __func__, __VA_ARGS__`, it won't compile if you have no printf parameters, i.e. if you just call `debug_print("Some msg\n");` You can fix this by using `fprintf(stderr, "%s:%d:%s(): " fmt, __FILE__, __LINE__, __func__, ##__VA_ARGS__);` The ##__VA_ARGS__ allows passing no parameters to the function. – mc_electron Jun 11 '14 at 18:55
  • @mc_electron: I believe your issue is covered in the sections on 'Single argument C99 variant' and 'GCC-specific Technique' in the main answer. If you agree, I suggest deleting your comment and flagging this one as obsolete. If not, we'd better have a discussion, maybe by email (see my profile). – Jonathan Leffler Jun 11 '14 at 20:58
  • could you explain more about "but it could be used as the left operand of a comma operator where the do { ... } while (0) version cannot." at the last paragraph of your answer? what do you mean by left operand of a comma operator? – bysreg Jun 05 '16 at 20:28
  • @bysreg: The comma operator has two operands, such as: `var1 = expr1, function(arg1, arg2)`. The left operand here is `var1 = expr1`; the right operand is `function(arg1, arg2)`. When executed, the left operand is evaluated (for its side effects — an assignment in this example), and then the value is discarded and the right operand is evaluated and its value is the value of the comma operator expression. With `#define debug_print(...) ((void)((DEBUG) ? fprintf(stderr, __VA_ARGS__) : 0))` you could write: `if (var1 == var2) debug_printf("var1 == var2\n"), var2++;` if you wanted to (but why?). – Jonathan Leffler Jun 05 '16 at 21:29
  • "Compared to the first version, you lose the limited checking that requires the 'fmt' argument" Can you clarify what checking is lost? Thanks. – LogicTom Jun 21 '16 at 01:42
  • 2
    @LogicTom: the difference is between `#define debug_print(fmt, ...)` and `#define debug_print(...)`. The first of these requires at least one argument, the format string (`fmt`) and zero or more other arguments; the second requires zero or more arguments in total. If you use `debug_print()` with the first, you get an error from the preprocessor about misusing the macro, whereas the second does not. However, you still get compilation errors because the replacement text is not valid C. So, it really isn't much of an difference — hence the use of the term 'limited checking'. – Jonathan Leffler Jun 21 '16 at 03:23
  • A slight more elegant and shorter way is to use a trailing else instead of the do while loop: `#define debug_print(fmt, ...) if (DEBUG) fprintf(stderr, fmt, __VA_ARGS__); else` – Guillermo Oct 02 '16 at 01:35
  • @Guillermo: Beauty and elegance are in the eye of the beholder. I prefer what I use to what you propose as it is more elegant to my eyes. – Jonathan Leffler Oct 02 '16 at 01:57
  • @JonathanLeffler Won't `__LINE__` report a wrong line if embedded like that? Shouldn't it be passed as an argument to the macro itself? – Zingam Sep 19 '17 at 05:15
  • @Zingam: I'm not sure why you think `__LINE__` would be handled wrongly. It is the value of the line number where the macro is invoked, not where it is defined. Did you test it? In the absence of a concrete example of how it is wrong, the answer is "No: `__LINE__` won't report the wrong value" for most definitions of 'wrong'. – Jonathan Leffler Sep 19 '17 at 05:20
  • @JonathanLeffler I believe I had such an issue in the past. I think the repoerted line won't be the same as the line where the macro was called. I'll test your example when I get to a compiler. – Zingam Sep 19 '17 at 05:24
  • @Zingam: If you find a reproducible problem, please let me know. It's likely to big for a comment — send me email about it (see my profile). Please do include platform information — especially which compiler, etc. If I can't reproduce the problem but the compiler that produces the problem for you is archaic (say, from the previous millennium, not the current one), then it is likely of limited interest, but I'd add a note about it. If it is a current version of a reasonably mainstream compiler, I'm very interested. If there's no problem, please consider removing your comments and flagging mine. – Jonathan Leffler Sep 19 '17 at 05:30
  • @JonathanLeffler suppose you want to call `#define debug_print(fmt, ...) \ do { if (DEBUG) fprintf(stderr, "%s:%d:%s(): " fmt, __FILE__, \ __LINE__, __func__, __VA_ARGS__); } while (0)` like this : `char *fmt = "anything : %s\n"; debug_print(fmt, "blah");` gcc would compile with error: `error: expected ‘)’ before ‘fmt’`. If I call it like `debug_print("anything : %s\n", "blah");` than it compiles successfully. What is the explanation about this compile error, and any idea of how to achieve it with a variable as an argument for the `fmt`? – HeinrichStack Nov 14 '17 at 15:27
  • @HeinrichStack: the error is easy to explain. The macro relies on string literal concatenation to ‘insert’ some format material before what you provide as an argument to the macro. When you use a variable for the format, you break the prerequisites. How to fix it? Instead of calling `fprintf()` directly, call a function that handles the printing itself. It would not need the constant string added to your format. It would probably use `flockfile()` and `funlockfile()` to give you thread safety (on a POSIX system). – Jonathan Leffler Nov 14 '17 at 16:22
  • 1
    @JonathanLeffler Im afraid, I need more of your explanation, of what did you mean with `handles the printing itself`, and why do you mention `flockfile()` ? – HeinrichStack Nov 15 '17 at 10:08
  • As far as I understand you are setting a debug level per translation unit. Why would not just set the level per application on startup that is read from some configuration? – St.Antario Apr 10 '19 at 07:38
  • 1
    The variant shown abover, @St.Antario, uses a single active debugging level across the entire application, and I usually use command line options to allow the debugging level to be set when the program is run. I also have a variant that recognizes multiple different subsystems, each of which is given a name and its own debugging level, so that I can use `-D input=4,macros=9,rules=2` to set the debug level of the input system to 4, the macros system to 9 (undergoing intense scrutiny) and the rules system to 2. There are endless variations on the theme; use whatever suits you. – Jonathan Leffler Apr 10 '19 at 16:57
  • How can I do something similar for stream-style prints? I want to control the printing using logic like comparing the message's severity to a chosen verbosity level, and have it used in user code like this: my_cerr(50) << "Filename is: " << fname << '\n'; What would my macro my_cerr(x) look like? – All The Rage Nov 20 '19 at 07:03
  • @AlltheRage — to be brutally honest, I've no idea how to write the macros shown for use in C++. C++ tries to avoid the C preprocessor when it can. The variable length, variable type argument lists supported by `printf()` etc don't work well with user-defined types. You should consult other authorities about how to write such code for C++ I/O streams. One possibility is to require the user to invoke `TRACE("Filename is: " << fname << '\n')` (so that the user types the RHS, and the macro then prefixes that with something suitable, such as `cerr << `. It probably isn't quite that easy, though. – Jonathan Leffler Nov 20 '19 at 20:49
  • I really like your variadic macro containing `__FILE__`, `__LINE__`, & `__func__`! Super useful! – Gabriel Staples Dec 14 '19 at 17:17
  • Update: just make sure to use `##__VA_ARGS__`, as commented here by @mc_electron, and later explained in this answer, to allow debug prints with a format string which requires no arguments. If you use only `__VA_ARGS__` instead, it's a compiler error. See the gcc documentation here (https://gcc.gnu.org/onlinedocs/cpp/Variadic-Macros.html), where it states: "the ‘##’ token paste operator has a special meaning when placed between a comma and a variable argument", if "the variable argument is left out when the [macro] is used, then the comma before the ‘##’ will be deleted." – Gabriel Staples Dec 14 '19 at 17:34
  • ..." This does not happen if you pass an empty argument, nor does it happen if the token preceding ‘##’ is anything other than a comma." – Gabriel Staples Dec 14 '19 at 17:34
  • Don't forget there are other compilers than GCC — using a GCC-specific extension is no help to those not using GCC (or a compiler compatible with GCC's extension). – Jonathan Leffler Dec 14 '19 at 18:01
  • how about using ({}) to put that if(DEBUG){code} part into a block? then that if won't matter for else statement outside – Atreyagaurav Apr 23 '20 at 12:56
  • 1
    That’s not C. I don’t use that feature of GCC. – Jonathan Leffler Apr 23 '20 at 13:04
32

I use something like this:

#ifdef DEBUG
 #define D if(1) 
#else
 #define D if(0) 
#endif

Than I just use D as a prefix:

D printf("x=%0.3f\n",x);

Compiler sees the debug code, there is no comma problem and it works everywhere. Also it works when printf is not enough, say when you must dump an array or calculate some diagnosing value that is redundant to the program itself.

EDIT: Ok, it might generate a problem when there is else somewhere near that can be intercepted by this injected if. This is a version that goes over it:

#ifdef DEBUG
 #define D 
#else
 #define D for(;0;)
#endif
mbq
  • 18,510
  • 6
  • 49
  • 72
  • 3
    As for `for(;0;)`, it might generate a problem when you write something like `D continue;` or `D break;`. – ACcreator Sep 10 '14 at 09:00
  • 2
    Got me; it seems very unlikely that it may occur on accident, though. – mbq Apr 03 '15 at 12:24
  • A more readable version just for printing might be `#define PRINTF if (0) printf`, as then you don't need the prefix everywhere. – Ed Graham Feb 01 '23 at 11:34
11

For a portable (ISO C90) implementation, you could use double parentheses, like this;

#include <stdio.h>
#include <stdarg.h>

#ifndef NDEBUG
#  define debug_print(msg) stderr_printf msg
#else
#  define debug_print(msg) (void)0
#endif

void
stderr_printf(const char *fmt, ...)
{
  va_list ap;
  va_start(ap, fmt);
  vfprintf(stderr, fmt, ap);
  va_end(ap);
}

int
main(int argc, char *argv[])
{
  debug_print(("argv[0] is %s, argc is %d\n", argv[0], argc));
  return 0;
}

or (hackish, wouldn't recommend it)

#include <stdio.h>

#define _ ,
#ifndef NDEBUG
#  define debug_print(msg) fprintf(stderr, msg)
#else
#  define debug_print(msg) (void)0
#endif

int
main(int argc, char *argv[])
{
  debug_print("argv[0] is %s, argc is %d"_ argv[0] _ argc);
  return 0;
}
Marcin Koziuk
  • 1,006
  • 6
  • 5
10

Here's the version I use:

#ifdef NDEBUG
#define Dprintf(FORMAT, ...) ((void)0)
#define Dputs(MSG) ((void)0)
#else
#define Dprintf(FORMAT, ...) \
    fprintf(stderr, "%s() in %s, line %i: " FORMAT "\n", \
        __func__, __FILE__, __LINE__, __VA_ARGS__)
#define Dputs(MSG) Dprintf("%s", MSG)
#endif
Christoph
  • 164,997
  • 36
  • 182
  • 240
9

I would do something like

#ifdef DEBUG
#define debug_print(fmt, ...) fprintf(stderr, fmt, __VA_ARGS__)
#else
#define debug_print(fmt, ...) do {} while (0)
#endif

I think this is cleaner.

LB40
  • 12,041
  • 17
  • 72
  • 107
  • I don't really like the idea of using a macro inside a test as a flag. Could you explain why the debug printing should be always checked ? – LB40 Oct 29 '09 at 16:47
  • 1
    @Jonathan: If the code only ever gets executed in debug mode, why should you care if it compiles in non-debug mode? `assert()` from the stdlib works the same way and I normally just re-use the `NDEBUG` macro for my own debugging code... – Christoph Oct 29 '09 at 16:52
  • using DEBUG in the test, if someone does an uncontrolled undef DEBUG, your code no longer compiles. right ? – LB40 Oct 29 '09 at 16:58
  • 4
    It is frustrating to enable debugging and then have to debug the debugging code because it refers to variables that have been renamed or retyped, etc. If the compiler (post pre-processor) always sees the print statement, it ensures that any surrounding changes have not invalidated the diagnostics. If the compiler does not see the print statement, it cannot protect you against your own carelessness (or the carelessness of your colleagues or collaborators). See 'The Practice of Programming' by Kernighan and Pike - http://plan9.bell-labs.com/cm/cs/tpop/. – Jonathan Leffler Oct 29 '09 at 17:07
  • @Jonathan: that's why you add `-DNDEBUG` only for release builds (or testing thereof) – Christoph Oct 29 '09 at 17:12
  • This is 'been there, done that' experience - I used essentially the technique described in this answer for a number of years (more than a decade). But I came across the advice in TPOP (see my previous comment), and then did enable some debugging code after a number of years, and ran into problems of changed context breaking the debugging. Several times, having the printing always validated has saved me from later problems. NB: You can simplify the non-debug macro replacement text to '((void)0)'. – Jonathan Leffler Oct 29 '09 at 17:13
  • 1
    @Christoph: well, sort of...I use NDEBUG to control assertions only, and a separate macro (usually DEBUG) to control debug tracing. I frequently do not want debug output to appear unconditionally, so I have mechanism to control whether the output appears (debug levels, and instead of calling fprintf() directly, I call a debug print function that only conditionally prints so the same build of the code can print or not print based on program options). I am advocating that for all builds, the compiler should see the diagnostic statements; however, it won't generate code unless debug is enabled. – Jonathan Leffler Oct 29 '09 at 17:26
  • @Jon, good explanations and nice references. Just one question, on what do you base your conditional prints (what is the condition inside the test.) You map the program options to variabes ? and then use these variables or rather define DEBUG according to the program options ? – LB40 Oct 29 '09 at 17:44
  • @LB: I actually call a printing function unconditionally, and it decides whether to print (see the headers in my answer - which is currently the accepted answer). The global variable debug was set to a debugging level (zero or negative, no output; increasing positive number give more output, in general), and the TRACE() macro included the debugging level as the first argument: TRACE((3, "Hi")); was a level 3 debugging statement - calling db_print(3, "Hi"). I normally have command line options to turn on debugging: -d turns on level 3, and -Dn turns it on to level n. [...to be continued...] – Jonathan Leffler Oct 29 '09 at 19:34
  • @LB: ... You could write the non-debug version of the macro as 'do { if (0) print_function(...); } while (0)'. You could have the debug version as 'do { if (debug) printf_function(...); } while (0)', etc. There are lots of ways to do it - C is more Perl like than Perl, sometimes! (Perl's motto is TMTOWTDI - Tim-Toady - There's More Than One Way To Do It.) – Jonathan Leffler Oct 29 '09 at 19:38
  • A minor point: I would use `((void)0)` instead of an empty do-loop – David R Tribble Oct 29 '09 at 23:59
9

According to http://gcc.gnu.org/onlinedocs/cpp/Variadic-Macros.html, there should be a ## before __VA_ARGS__.

Otherwise, a macro #define dbg_print(format, ...) printf(format, __VA_ARGS__) will not compile the following example: dbg_print("hello world");.

Jonathan Leffler
  • 730,956
  • 141
  • 904
  • 1,278
Chobits Tai
  • 91
  • 1
  • 2
  • 1
    Welcome to Stack Overflow. You are correct that GCC has the non-standard extension that you reference. The currently accepted answer does in fact mention this, including exactly the reference URL you give. – Jonathan Leffler Oct 27 '12 at 23:10
8
#define debug_print(FMT, ARGS...) do { \
    if (DEBUG) \
        fprintf(stderr, "%s:%d " FMT "\n", __FUNCTION__, __LINE__, ## ARGS); \
    } while (0)
eyalm
  • 3,366
  • 19
  • 21
  • Which version of C supports that notation? And, if it worked, the token pasting all the arguments like that means that you have only a very limited set of options for the format string, doesn't it? – Jonathan Leffler Oct 29 '09 at 16:28
  • @Jonathan: gcc (Debian 4.3.3-13) 4.3.3 – eyalm Oct 29 '09 at 16:46
  • 1
    OK - agreed: it is documented as an old GNU extension (section 5.17 of the GCC 4.4.1 manual). But you should probably document that it will only work with GCC - or maybe we've done that between us in these comments. – Jonathan Leffler Oct 29 '09 at 17:01
  • 1
    My intention was to show another style of using args and mainly to demonstrate the usage of __FUNCTION__ and __LINE__ – eyalm Oct 29 '09 at 19:32
2

This is what I use:

#if DBG
#include <stdio.h>
#define DBGPRINT printf
#else
#define DBGPRINT(...) /**/  
#endif

It has the nice benefit to handle printf properly, even without additional arguments. In case DBG ==0, even the dumbest compiler gets nothing to chew upon, so no code is generated.

5tenzel
  • 21
  • 1
2

So, when using gcc, I like:

#define DBGI(expr) ({int g2rE3=expr; fprintf(stderr, "%s:%d:%s(): ""%s->%i\n", __FILE__,  __LINE__, __func__, #expr, g2rE3); g2rE3;})

Because it can be inserted into code.

Suppose you're trying to debug

printf("%i\n", (1*2*3*4*5*6));

720

Then you can change it to:

printf("%i\n", DBGI(1*2*3*4*5*6));

hello.c:86:main(): 1*2*3*4*5*6->720
720

And you can get an analysis of what expression was evaluated to what.

It's protected against the double-evaluation problem, but the absence of gensyms does leave it open to name-collisions.

However it does nest:

DBGI(printf("%i\n", DBGI(1*2*3*4*5*6)));

hello.c:86:main(): 1*2*3*4*5*6->720
720
hello.c:86:main(): printf("%i\n", DBGI(1*2*3*4*5*6))->4

So I think that as long as you avoid using g2rE3 as a variable name, you'll be OK.

Certainly I've found it (and allied versions for strings, and versions for debug levels etc) invaluable.

John Lawrence Aspden
  • 17,124
  • 11
  • 67
  • 110
1

I've been stewing on how to do this for years, and finally come up with a solution. However, I didn't know that there were other solutions here already. First, at difference with Leffler's answer, I don't see his argument that debug prints should always be compiled. I'd rather not have tons of unneeded code executing in my project, when not needed, in cases where I need to test and they might not be getting optimized out.

Not compiling every time might sound worse than it is in actual practice. You do wind up with debug prints that don't compile sometimes, but it's not so hard to compile and test them before finalizing a project. With this system, if you are using three levels of debugs, just put it on debug message level three, fix your compile errors and check for any others before you finalize yer code. (Since of course, debug statements compiling are no guarantee that they are still working as intended.)

My solution provides for levels of debug detail also; and if you set it to the highest level, they all compile. If you've been using a high debug detail level recently, they all were able to compile at that time. Final updates should be pretty easy. I've never needed more than three levels, but Jonathan says he's used nine. This method (like Leffler's) can be extended to any number of levels. The usage of my method may be simpler; requiring just two statements when used in your code. I am, however, coding the CLOSE macro too - although it doesn't do anything. It might if I were sending to a file.

Against the cost the extra step of testing them to see that they will compile before delivery, is that

  1. You must trust them to get optimized out, which admittedly SHOULD happen if you have a sufficient optimization level.
  2. Furthermore, they probably won't if you make a release compile with optimization turned off for testing purposes (which is admittedly rare); and they almost certainly won't at all during debug - thereby executing dozens or hundreds of "if (DEBUG)" statements at runtime; thus slowing execution (which is my principle objection) and less importantly, increasing your executable or dll size; and hence execution and compile times. Jonathan, however, informs me his method can be made to also not compile statements at all.

Branches are actually relatively pretty costly in modern pre-fetching processors. Maybe not a big deal if your app is not a time-critical one; but if performance is an issue, then, yes, a big enough deal that I'd prefer to opt for somewhat faster-executing debug code (and possibly faster release, in rare cases, as noted).

So, what I wanted is a debug print macro that does not compile if it is not to be printed, but does if it is. I also wanted levels of debugging, so that, e.g. if I wanted performance-crucial parts of the code not to print at some times, but to print at others, I could set a debug level, and have extra debug prints kick in. I came across a way to implement debug levels that determined if the print was even compiled or not. I achieved it this way:

DebugLog.h:

// FILE: DebugLog.h
// REMARKS: This is a generic pair of files useful for debugging.  It provides three levels of 
// debug logging, currently; in addition to disabling it.  Level 3 is the most information.
// Levels 2 and 1 have progressively more.  Thus, you can write: 
//     DEBUGLOG_LOG(1, "a number=%d", 7);
// and it will be seen if DEBUG is anything other than undefined or zero.  If you write
//     DEBUGLOG_LOG(3, "another number=%d", 15);
// it will only be seen if DEBUG is 3.  When not being displayed, these routines compile
// to NOTHING.  I reject the argument that debug code needs to always be compiled so as to 
// keep it current.  I would rather have a leaner and faster app, and just not be lazy, and 
// maintain debugs as needed.  I don't know if this works with the C preprocessor or not, 
// but the rest of the code is fully C compliant also if it is.

#define DEBUG 1

#ifdef DEBUG
#define DEBUGLOG_INIT(filename) debuglog_init(filename)
#else
#define debuglog_init(...)
#endif

#ifdef DEBUG
#define DEBUGLOG_CLOSE debuglog_close
#else
#define debuglog_close(...)
#endif

#define DEBUGLOG_LOG(level, fmt, ...) DEBUGLOG_LOG ## level (fmt, ##__VA_ARGS__)

#if DEBUG == 0
#define DEBUGLOG_LOG0(...)
#endif

#if DEBUG >= 1
#define DEBUGLOG_LOG1(fmt, ...) debuglog_log (fmt, ##__VA_ARGS__)
#else
#define DEBUGLOG_LOG1(...)
#endif

#if DEBUG >= 2
#define DEBUGLOG_LOG2(fmt, ...) debuglog_log (fmt, ##__VA_ARGS__)
#else
#define DEBUGLOG_LOG2(...)
#endif

#if DEBUG == 3
#define DEBUGLOG_LOG3(fmt, ...) debuglog_log (fmt, ##__VA_ARGS__)
#else
#define DEBUGLOG_LOG3(...)
#endif

void debuglog_init(char *filename);
void debuglog_close(void);
void debuglog_log(char* format, ...);

DebugLog.cpp:

// FILE: DebugLog.h
// REMARKS: This is a generic pair of files useful for debugging.  It provides three levels of 
// debug logging, currently; in addition to disabling it.  See DebugLog.h's remarks for more 
// info.

#include <stdio.h>
#include <stdarg.h>

#include "DebugLog.h"

FILE *hndl;
char *savedFilename;

void debuglog_init(char *filename)
{
    savedFilename = filename;
    hndl = fopen(savedFilename, "wt");
    fclose(hndl);
}

void debuglog_close(void)
{
    //fclose(hndl);
}

void debuglog_log(char* format, ...)
{
    hndl = fopen(savedFilename,"at");
    va_list argptr;
    va_start(argptr, format);
    vfprintf(hndl, format, argptr);
    va_end(argptr);
    fputc('\n',hndl);
    fclose(hndl);
}

Using the macros

To use it, just do:

DEBUGLOG_INIT("afile.log");

To write to the log file, just do:

DEBUGLOG_LOG(1, "the value is: %d", anint);

To close it, you do:

DEBUGLOG_CLOSE();

although currently this isn't even necessary, technically speaking, as it does nothing. I'm still using the CLOSE right now, however, in case I change my mind about how it works, and want to leave the file open between logging statements.

Then, when you want to turn on debug printing, just edit the first #define in the header file to say, e.g.

#define DEBUG 1

To have logging statements compile to nothing, do

#define DEBUG 0

If you need info from a frequently executed piece of code (i.e. a high level of detail), you may want to write:

 DEBUGLOG_LOG(3, "the value is: %d", anint);

If you define DEBUG to be 3, logging levels 1, 2 & 3 compile. If you set it to 2, you get logging levels 1 & 2. If you set it to 1, you only get logging level 1 statements.

As to the do-while loop, since this evaluates to either a single function or nothing, instead of an if statement, the loop is not needed. OK, castigate me for using C instead of C++ IO (and Qt's QString::arg() is a safer way of formatting variables when in Qt, too — it's pretty slick, but takes more code and the formatting documentation isn't as organized as it might be - but still I've found cases where its preferable), but you can put whatever code in the .cpp file you want. It also might be a class, but then you would need to instantiate it and keep up with it, or do a new() and store it. This way, you just drop the #include, init and optionally close statements into your source, and you are ready to begin using it. It would make a fine class, however, if you are so inclined.

I'd previously seen a lot of solutions, but none suited my criteria as well as this one.

  1. It can be extended to do as many levels as you like.
  2. It compiles to nothing if not printing.
  3. It centralizes IO in one easy-to-edit place.
  4. It's flexible, using printf formatting.
  5. Again, it does not slow down debug runs, whereas always-compiling debug prints are always executed in debug mode. If you are doing computer science, and not easier to write information processing, you may find yourself running a CPU-consuming simulator, to see e.g. where the debugger stops it with an index out of range for a vector. These run extra-slowly in debug mode already. The mandatory execution of hundreds of debug prints will necessarily slow such runs down even further. For me, such runs are not uncommon.

Not terribly significant, but in addition:

  1. It requires no hack to print without arguments (e.g. DEBUGLOG_LOG(3, "got here!");); thus allowing you to use, e.g. Qt's safer .arg() formatting. It works on MSVC, and thus, probably gcc. It uses ## in the #defines, which is non-standard, as Leffler points out, but is widely supported. (You can recode it not to use ## if necessary, but you will have to use a hack such as he provides.)

Warning: If you forget to provide the logging level argument, MSVC unhelpfully claims the identifier is not defined.

You might want to use a preprocessor symbol name other than DEBUG, as some source also defines that symbol (eg. progs using ./configure commands to prepare for building). It seemed natural to me when I developed it. I developed it in an application where the DLL is being used by something else, and it's more convent to send log prints to a file; but changing it to vprintf() would work fine, too.

I hope this saves many of you grief about figuring out the best way to do debug logging; or shows you one you might prefer. I've half-heartedly been trying to figure this one out for decades. Works in MSVC 2012 & 2015, and thus probably on gcc; as well as probably working on many others, but I haven't tested it on them.

I mean to make a streaming version of this one day, too.

Note: Thanks go to Leffler, who has cordially helped me format my message better for StackOverflow.

CodeLurker
  • 1,248
  • 13
  • 22
  • 2
    You say "executing dozens or hundreds of `if (DEBUG)` statements at runtime, which do not get optimized out" — which is [tilting at windmills](https://en.wikipedia.org/wiki/Tilting_at_windmills). The whole point of the system I described is that the code is checked by the compiler (important, and automatic — no special build required) but the debug code is not generated at all because it _is_ optimized out (so there is zero runtime impact on code size or performance because the code is not present at runtime). – Jonathan Leffler Apr 10 '16 at 16:17
  • Jonathan Leffler: Thx for pointing out my mis-wording. I let my thoughts race faster than my fingers, being so glad to have this finished. I've revised my objections with "... 1) you must trust them to get optimized out, which admittedly should happen if you have a sufficient optimization level. 2) Furthermore, they won't if you make a release compile with optimization turned off for testing purposes; and they probably won't at all during debug - thereby executing dozens or hundreds of 'if (DEBUG)' statements at runtime - thereby increasing your executable or dll size, and execution times." – CodeLurker Nov 08 '16 at 17:50
  • For yours to do the other important thing mine is doing, you would have to have debug levels. While I often don't need a lot of them turned on, a few applications really benefit from being able to get a great level of detail about a time-critical loop with a simple "#define DEBUG 3", and then go back to much less verbose information with "#define DEBUG 1". I've never needed more than three levels, and thus, at least roughly 1/3 of my debugs compile already at release. If I've used level 3 recently, they probably ALL do. – CodeLurker Nov 08 '16 at 18:09
  • YMMV. The modern system I showed supports dynamic (runtime) setting of debug levels, so you can programmatically decide how much of the debug is produced at runtime. I usually used levels 1-9, though there's no upper limit (or lower limit; the default level is 0 which is usually off, but can be explicitly requested during active development if appropriate — it is not appropriate for long-term work). I chose a default level of 3; things can be tuned. This gives me a lot of control. If you really don't want to test the debug code when inactive, change the alternative to `((void)0)` — it's easy. – Jonathan Leffler Nov 08 '16 at 18:11
  • I don't see the debug levels in your code, although you mention it. For that, you would need "debug_print(3, fmt, ...)". Perhaps you have used that in the past and not posted it. As to not having to execute the code during debug, do you mean, "#define DEBUG_TEST ((void)0)"? "if ((void)0)" is a compile error. Also, in the debug functions, the IO can be as modern as one likes; as I mentioned. The .arg() alternative is indeed more modern than printf varargs. – CodeLurker Nov 08 '16 at 18:42
  • This needs to go to chat, but I'm not sure how we set up a chat and leave a reference here so other people can find it later. I hoped that SO might suggest it automatically, but it didn't do so yet. The prototype `extern void db_print(int level, const char *fmt,...);` has the debug level as the first argument: the outline usage of the macro is: `Usage: TRACE((level, fmt, ...))` where `"level" is the debugging level which must be operational for the output to appear.` (I now have a version `DB_TRACE(level, fmt, ...)` which uses single instead of double parentheses, but is otherwise the same.) – Jonathan Leffler Nov 08 '16 at 18:42
  • What I meant was that instead of `#define TRACE(x) do { if (0) db_print x; } while (0)` you would use `#define TRACE(x) ((void)0)`. – Jonathan Leffler Nov 08 '16 at 18:45
  • 1
    Ahh. It would have helped to have read the entire thing. It is a rather long post. I think that's got the essential points thus far. It turns out yours, like mine, can be used to compile or not compile all debug prints, and can support levels; although admittedly, yours can compile levels you aren't using - at a cost during debug. – CodeLurker Nov 08 '16 at 18:57
  • Let us [continue this discussion in chat](http://chat.stackoverflow.com/rooms/127660/discussion-between-jonathan-leffler-and-codelurker). – Jonathan Leffler Nov 08 '16 at 18:58
1

My favourite of the below is var_dump, which when called as:

var_dump("%d", count);

produces output like:

patch.c:150:main(): count = 0

Credit to @"Jonathan Leffler". All are C89-happy:

Code

#define DEBUG 1
#include <stdarg.h>
#include <stdio.h>
void debug_vprintf(const char *fmt, ...)
{
    va_list args;
    va_start(args, fmt);
    vfprintf(stderr, fmt, args);
    va_end(args);
}

/* Call as: (DOUBLE PARENTHESES ARE MANDATORY) */
/* var_debug(("outfd = %d, somefailed = %d\n", outfd, somefailed)); */
#define var_debug(x) do { if (DEBUG) { debug_vprintf ("%s:%d:%s(): ", \
    __FILE__,  __LINE__, __func__); debug_vprintf x; }} while (0)

/* var_dump("%s" variable_name); */
#define var_dump(fmt, var) do { if (DEBUG) { debug_vprintf ("%s:%d:%s(): ", \
    __FILE__,  __LINE__, __func__); debug_vprintf ("%s = " fmt, #var, var); }} while (0)

#define DEBUG_HERE do { if (DEBUG) { debug_vprintf ("%s:%d:%s(): HERE\n", \
    __FILE__,  __LINE__, __func__); }} while (0)
Tom Hale
  • 40,825
  • 36
  • 187
  • 242
1
#define PRINT_LOG(str_format, ...) { \
    time_t curtime=time (NULL); \
    struct tm *ltm = localtime (&curtime); \
    printf("[%d-%02d-%02d %02d:%02d:%02d] " str_format, \
        ltm->tm_year + 1900, ltm->tm_mon + 1, ltm->tm_mday, \
        ltm->tm_hour, ltm->tm_min, ltm->tm_sec, ##__VA_ARGS__); \
}
    
PRINT_LOG("[%d] Serving client, str=%s, number=%d\n", getpid(), "my str", 10);
  • Thank you for this code snippet, which might provide some limited, immediate help. A [proper explanation](https://meta.stackexchange.com/q/114762/9193372) would greatly improve its long-term value by showing why this is a good solution to the problem and would make it more useful to future readers with other, similar questions. Please edit your answer to add some explanation, including the assumptions you’ve made. – Syscall Mar 20 '21 at 15:46
0

I believe this variation of the theme gives debug categories without the need to have a separate macro name per category.

I used this variation in an Arduino project where program space is limited to 32K and dynamic memory is limited to 2K. The addition of debug statements and trace debug strings quickly uses up space. So it is essential to be able to limit the debug trace that is included at compile time to the minimum necessary each time the code is built.

debug.h

#ifndef DEBUG_H
#define DEBUG_H

#define PRINT(DEBUG_CATEGORY, VALUE)  do { if (DEBUG_CATEGORY & DEBUG_MASK) Serial.print(VALUE);} while (0);

#endif

calling .cpp file

#define DEBUG_MASK 0x06
#include "Debug.h"

...
PRINT(4, "Time out error,\t");
...
user358795
  • 145
  • 1
  • 7
0

If you don't care that the output goes to stdout, you can use this:

int doDebug = DEBUG;  // Where DEBUG may be supplied in compiler command
#define trace if (doDebug) printf

trace("whatever %d, %i\n", arg1, arg2);
trindflo
  • 311
  • 3
  • 12