Re: bug in glibc?

Ya I think I understand now... Thanks for your inputs... The buffers
are copied "as is" and the fork() call is flushing the buffers of the
child. The parent outputs its buffered data after many loops.

@wayne: You are right, the only difference being flush of buffers is
not triggered by printf but by fork(), because, as you can see, even
after you remove printf in child the output doesn't change, except
that there are no new-lines.

What I wish to say now is that: We can see that fork() call is
flushing the buffers in the child, but if there is a fork library call
which can flush the buffers before calling fork() system call, then
the output will be close to intuition. I would like to know your
opinions regarding this.

Thank you all,
KK.


On Jan 18, 2008 10:11 PM, Wayne Pinette <Wpinette@xxxxxx> wrote:
Im sorry to say that the bug is not in glibc, the bug is in your code/logic.

fork() is doing exactly what it is suppose to do in this case, this is why:


from man fork : The fork() function shall create a new proces. The new process shall be an exact copy of the calling process...and so on.

In your code :

main()
{
int a = 0;

while (1) {
printf("XXX(%d)",++a); <--- AT this point the buffer is not flushed, it is still full
(void) sleep(1);
if (fork() == 0) { <0---- at the point fork is called, the buffer is still not flushed, it is full
printf("\n"); <-- at thsi point you flush the buffer, but you have flushed it off the first process, the copy has already been made, and was
made with a full buffer.
exit(0);
}
}
}

If you put the "\n" in the original printf statement :

printf("XXX(%d)\n",++a);

You will have flushed the buffer BEFORE the fork and consequently before the exact copy of the process, and it will work the way you want it to.


Wayner


Dr Alan J Bartlett <stxsl_ajb@xxxxxxxxxxx> 18/01/2008 8:21 am >>>







KK,

I am running CentOS 5, update 1 ( == RHEL 5.1) and thought I would mimic your test.

$ uname -a
Linux stxsl 2.6.18-53.1.4.el5.stxsl #1 SMP Tue Jan 8 21:29:57 GMT 2008 i686 i686 i386 GNU/Linux
$ rpm -q glibc
glibc-2.5-18.el5_1.1
$ ldd zxc
linux-gate.so.1 => (0x00343000)
libc.so.6 => /lib/libc.so.6 (0x00b3d000)
/lib/ld-linux.so.2 (0x0016e000)
$ cat zxc.c
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

main()
{
int a = 0;

while (1) {
printf("XXX(%d)",++a);
(void) sleep(1);
if (fork() == 0) {
printf("\n");
exit(0);
}
}
}
$ ./zxc
XXX(1)
XXX(1)XXX(2)
XXX(1)XXX(2)XXX(3)
XXX(1)XXX(2)XXX(3)XXX(4)
XXX(1)XXX(2)XXX(3)XXX(4)XXX(5)
XXX(1)XXX(2)XXX(3)XXX(4)XXX(5)XXX(6)

$

You are not alone!

Regards,
Alan.


_________________________________________________________________
Free games, great prizes - get gaming at Gamesbox.
http://www.searchgamesbox.com--
redhat-list mailing list
unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/redhat-list


--
redhat-list mailing list
unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subject=unsubscribe

https://www.redhat.com/mailman/listinfo/redhat-list


--
redhat-list mailing list
unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/redhat-list

Relevant Pages

  • Re: [PATCH/RFC] Shared page tables
    ... > sharing, yeah. ... > up front, then never fork during the test, so fork performance doesn't ... > mmap() randomization doesn't affect fork at all, ... It has to be the case that they all agree on what addresses the buffers are ...
    (Linux-Kernel)
  • Re: Process.fork weirdness
    ... other threads can keep filling buffers before the fork happens. ... seems like there should be an option to flush everything. ... Don't mix threads and fork, ... often a good idea to use exit! ...
    (comp.lang.ruby)
  • Re: bug in glibc?
    ... //you can remove the printf here, ... Hence the data in the buffers is getting accumulated. ... You fork the whole process, so the buffers are copied "as is", so the first process ...
    (RedHat)
  • Re: timing a fork
    ... if ($kid) { ... What it does is say that the parent isn't interested in ... Once a parent process has forked it can carry on on its own while the child ... In fact what a call to 'system' does is to fork a child ...
    (perl.beginners)
  • Re: Ruby lacks atfork : The evil that lives in fork...
    ... so that there's no way to ensure mutex to share ... Give the resource to the child. ... fork() in a multi-threaded program. ... multi-threaded I/O libraries, which are almost sure to be invoked ...
    (comp.lang.ruby)