[Bug stdio/21544] New: stdio fprintf, printf returning >0 after underlying fd closed

["dslowik500 at gmail dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=21544

            Bug ID: 21544
           Summary: stdio fprintf, printf returning >0 after underlying fd
                    closed
           Product: glibc
           Version: 2.23
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: stdio
          Assignee: unassigned at sourceware dot org
          Reporter: dslowik500 at gmail dot com
  Target Milestone: ---

As in the code below,
After opening a regular file FILE *fp to write to, then closing the underlying
file descriptor, a call to fprintf on fp returns number of chars given to
fprintf. I think it should return -1 as an strace shows that first fprintf
seems to trigger a call to fstat on the underlying fd which fails with EBADF
errno. fprintf will return -1 only after enough chars have been added into the
stdio buffer to trigger a write syscall, and that write error is propagated
back through fprintf, but not the initial fstat error. The question is: Should
fprintf report the error when it first detects it through fstat or wait until
write later reports the same error? 
[Note that if the fd is closed 'after' first fprintf call, it will have to wait
for a full stdio buffer to trigger write to detect error; but I'm thinking
first error detected should be reported by stdio in either case.]

A similar thing happens with printf writing to stdout after closing
STDOUT_FILENO. That code and discussion is at:
https://stackoverflow.com/questions/44034025/stdio-to-terminal-after-closestdout-fileno-behavior

OS:
Linux HAL 4.4.0-79-generic #100-Ubuntu SMP Wed May 17 19:58:14 UTC 2017 x86_64
x86_64 x86_64 GNU/Linux

compiler:
gcc (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609

code for fprintf to regular file:
#include <unistd.h>
#include <stdio.h>
#include <string.h>
#include <errno.h>

int main() {
  FILE *fp = fopen("out.txt", "w");

  // Close underlying file descriptor:
  int fd = fileno(fp);
  close(fd);

  // write fails:
  if (write(fd, "Direct write\n", 13) != 13)
    fprintf(stderr, "Error on write after close(fd): %s\n", strerror(errno));

  // but fprintf() calls continue fine, ferror(fp) = 0, until first write()
call on filled stdio buffer:
  int rtn = 1, cum = 0;
  errno = 0;
  //  while (rtn>0) {  // uncomment while loop to fill stdio buffer.
    if ((rtn = fprintf(fp, "fprintf after close(fd)\n")) < 0 || ferror(stdout)) 
      fprintf(stderr, "Error on fprintf after close(fp): %s\n",
strerror(errno));
    cum += rtn;
    fprintf(stderr, "fprintf returned %d, cum: %d\n", rtn, cum);
  // }

  // Only on fflush is error detected (if write not yet triggered above):
  if (fflush(fp) || ferror(fp))
    fprintf(stderr, "Error on fflush(fp): %s\n", strerror(errno));
}

-- 
You are receiving this mail because:
You are on the CC list for the bug.