Re: open() creates a new file even if it fails with EMFILE

Hi Christian,

On 5/4/2026 11:39 AM, Christian Franke via Cygwin wrote:
> Could be reproduced with 3.6.9 and current HEAD.
> 
> Testcase:
> 
> $ uname -r # or 3.6.9-1.x86_64
> 3.7.0-0.483.ga1f347c0d5b8.x86_64
> 
> $ cat openmany.c
> #include <errno.h>
> #include <fcntl.h>
> #include <stdio.h>
> #include <sys/stat.h>
> 
> int main()
> {
>    for (int i = 0; i < 10000; i++) {
>      char name[32];
>      snprintf(name, sizeof(name), "file-%04d.tmp", i);
>      errno = 0;
>      int fd = open(name, O_WRONLY | O_CREAT | O_EXCL, 0666);
>      if (fd >= 0)
>        continue;
>      printf("open(%s, ...)=%d (errno=%d)\n", name, fd, errno);
>      struct stat st;
>      printf("stat(%s, ...)=%d\n", name, stat(name, &st));
>      break;
>    }
>    return 0;
> }
> 
> $ gcc -o openmany openmany.c
> 
> $ ls file-*.tmp
> ls: cannot access 'file-*.tmp': No such file or directory
> 
> $ ulimit -n
> 3200
> 
> $ ./openmany
> open(file-3197.tmp, ...)=-1 (errno=24)
> stat(file-3197.tmp, ...)=0
> 
> $ ls file-*.tmp
> file-0000.tmp
> file-0001.tmp
> ...
> file-3195.tmp
> file-3196.tmp
> file-3197.tmp
> 
> $ grep -w 24 /usr/include/sys/errno.h
> #define EMFILE 24       /* File descriptor value too large */
> 
> 
> Looks like the fd is allocated too late.

Thanks for the report and especially for the STC.  It is during fdtable 
extension that the limit is checked; by then NtCreateFile() has already 
been called successfully.  We need to consider a fix that avoids 
needlessly creating a file that can't be operated on anyway.

..mark


-- 
Problem reports:      https://cygwin.com/problems.html
FAQ:                  https://cygwin.com/faq/
Documentation:        https://cygwin.com/docs.html
Unsubscribe info:     https://cygwin.com/ml/#unsubscribe-simple