Gilad Ben-Yossef <[EMAIL PROTECTED]> writes:
> Oleg Goldshmidt wrote:
>
> > Purely from memory, it is POSIX, though I would use STDOUT_FILENO
> > or fileno(stdout).
>
> OK. I see I wasn't clear about what i was asking.
Not to me, but that's my problematic internal parser. Obviously I read
the question as "is fd 1 guaranteed to be stdout?". I should have
given you the credit of knowing that even when you are really tired.
I apologize. I was tired... ;-)
Apparently, there is no guarantee that the operation in the test will
succeed (see below). I suspect that on most platforms std{in,out,err}
(or whatever the macros expand to) are static, and maybe there are
platforms that return those pointers from fdopen().
At least it is not the case with glibc on Linux:
#include <stdio.h>
#include <string.h>
#include <errno.h>
#include <limits.h>
int main(void) {
int rc = 0;
int newfd;
FILE* new = NULL;
char file[PATH_MAX] = "/tmp/fd";
fprintf(stderr,"stdout: %p (initial)\n", (void*)stdout);
if ((newfd = dup(fileno(stdout))) == -1) {
perror(strerror(errno));
rc = 1;
goto CLEAN_UP;
}
if (freopen(file,"w",stdout) == NULL) {
perror(strerror(errno));
rc = 2;
goto CLEAN_UP;
}
fclose(stdout);
fprintf(stderr,"stdout: %p (closed)\n", (void*)stdout);
if (dup2(newfd,1) != 1) {
perror(strerror(errno));
rc = 3;
goto CLEAN_UP;
}
if ((new = fdopen(1,"w")) == NULL) {
perror(strerror(errno));
rc = 4;
goto CLEAN_UP;
}
fprintf(stderr,"new: %p\n", (void*)new);
fprintf(stderr,"stdout: %p (final)\n", (void*)stdout);
CLEAN_UP:
return rc;
}
produces
$ ./a.out
stdout: 0x4212db00 (initial)
stdout: 0x4212db00 (closed)
new: 0x8049b80
stdout: 0x4212db00 (final)
It does look like a bug.
--
Oleg Goldshmidt | [EMAIL PROTECTED]
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
