Re: This may be a bug

[Bob Proulx]

>   I am a graduate at UBC. These days, I am implementing a software. I
> tested my program on different versions of Linux. My program can run
> well on Linux 2.14, but failed on Linux 2.16. Since my program used
> named pipe to communicate data between two processes, I checked mkfifo
> command and mkfifo() function on the two Linux versions, and I find that
> Linux 2.16 failed to create named pipe. The following is my result:
[...]
>   The result is same if using mkfifo() in a C program. 
[...]
>  >ls -l feng
>  -rw-------    1 qfzhang  grads           0 Mar 23 15:45 feng
>   The result is same if using mkfifo() in a C program.
[...]
>  It seems that Linux 2.16 failed to create name pipe.

I agree completely with your assessment and thank you for your report.
However this mailing list is for the mkfifo command and not the
mkfifo() library call.  The mkfifo command uses the system call just
like your C program.  If it did not work for your C program then it
won't work for the mkfifo C program either.  I will note that I cannot
reproduce your problem here, however.  It might be a local problem.

We can't help you with the library routines from glibc.  And perhaps
this might even be a deeper problem with the mknod() kernel system
call on your machine.

I suggest you make your report to the glibc mailing list and see what
information they can provide.

  address@hidden
  http://mail.gnu.org/mailman/listinfo/bug-glibc

Bob