I have come across many questions related to this but my requirement is quite different. I developed a C application(mib-test.c) earlier which uses fprintf at lot of places. fprintf logs the messages in a file.
But now I need to print these messages to console also. I am looking for a solution in which I do not need to change fprintf in all the places.
Is there a way to do it?
No, there is not, you will have to edit each instance. A better approach to begin with would have been to create your own logging function or macro. I would suggest you to do that now, and save yourself some time if you ever change mind again in the future.
Example using a function:
void log(const char *fmt, ...)
{
va_list args;
va_start(args, fmt);
vfprintf(stderr, fmt, args);
va_end(args);
va_start(args, fmt);
vfprintf(logfile, fmt, args);
va_end(args);
}
Example using a macro:
#define log(fmt, ...) do { \
fprintf(stderr, fmt, __VA_ARGS__); \
fprintf(logfile, fmt, __VA_ARGS__); \
} while (0)
See also: do { ... } while (0) — what is it good for?
Note that using a macro will evaluate expressions twice. For example log("%d\n", get_int()); using the macro form will call get_int() twice. You should only use the macro if you plan to pass simple values.
Both solutions assume that logfile is global and already opened for writing. You could actually define it static inside the function and handle the initial opening checking if (logfile == NULL) before doing anything.
@marcobonelli wrote a great answer. So for the replacing part, if you're not using an IDE that can do all the replacements for you, you could probably use this simple sed command to do all the replacement. Assume you have a log function like in this answer, then just do this:
sed 's/fprintf(logfile/log(/g' -i <file>
This is incomplete (error handling is omitted), but it demonstrates the principle of replacing filedescriptors via dup() and dup2():
#include <stdlib.h>
#include <unistd.h>
#include <string.h>
#include <stdio.h>
#include <fcntl.h>
int errtee(char * path);
/*
** Function to write to *both* stderr and a named file
** Return the new filedescriptor, which should be
** used instead of STDERR_FILENO.
** (or -1 on error)
*/
int errtee(char * path){
int out[2];
pipe(out);
int pid=fork();
if(pid<0){ // error
perror( "fork() failed\n" );
return -1;
}
if(pid>0){ // parent
close(out[0]);
return out[1];
}
if(pid==0){ // child
int fd_extra;
char buff[512];
fd_extra = open(path, O_CREAT | O_WRONLY, 0666);
if (fd_extra < 0) return -1;
close(out[1]);
dup2(out[0], STDIN_FILENO);
close(out[0]);
while(1) {
unsigned done, todo;
int cnt, this;
cnt = read(STDIN_FILENO, buff, sizeof buff) ;
if (cnt <=0) break;
for(todo = cnt, done = 0; done < todo; done += this) {
this = write(STDERR_FILENO, buff+done, todo-done);
if (cnt <=0) break; // this is not perfect ...
}
for(todo = cnt, done = 0; done < todo; done += this) {
this = write(fd_extra, buff+done, todo-done);
if (cnt <=0) break; // this is not perfect ...
}
}
close( fd_extra);
}
exit(0);
}
int main(void){
int new_fd;
int old_fd;
char buff[222];
new_fd = errtee("Mylogfile.log" );
if (new_fd < 0) {
perror( "errtee() failed\n" );
exit(1);
}
// Save the stdin filedescriptor
old_fd = dup( STDERR_FILENO);
// ... and replace it by the one from the subprocess.
dup2(new_fd, STDERR_FILENO);
// Send some output to stderr ...
while (fgets(buff, sizeof buff, stdin) ) {
fprintf(stderr, "%s", buff);
}
// Restore the old stdin filedescriptor
dup2(old_fd, STDERR_FILENO);
// ... and check if stderr still works ...
fprintf(stderr,"Completed\n" );
return 0;
}
Note: diagnostic output should go to stderr. If you keep it that way, it can always be replaced by a method like above.
[It is also possible to fetch the fileno() from the fopen'd log file, and do it the other way around. YMMV]
