Is this a bug? (something about immortal processes and ps going crazy...)

[Joaquin Ferruz Melero]

Hello all,

I attach a small program which does something simple: First it
programs a posix timer; the timer will send the process a SIGKILL signal
when it expires, so that the process should die. Then it loops without
end. It has also a handler for SIGUSR1 which will print the remaining
count for the timer if such signal is sent to the process.

Well, as a matter of fact, the process won't die! If you kill it
BEFORE the timer has expired, it disappears, as expected. However, if
you wait for the timer to send the process SIGKILL, you find that a)
This does not kill the process b) You CANNOT kill the process, even with
SIGKILL. c) If you type ps -l -p <process id>, "ps" will loop forever
while it prints the stuff for the non-dead process over and over...

It happens under QNX 6.2, and QNX 6.1; it does not happen under
Solaris 2.6. The process dies, however, if the timer sends SIGTERM, and
its handler calls exit(0).

Is this a system bug?

Best regards,

Joaquín Ferruz

[Robert Rutherford]

We have reported a similar problem and so have a few others. Do you also
find that "psin" fails in this scenario?

So far the silence from QSSL has been deafening.

Rob Rutherford

"Joaquin Ferruz Melero" <fer...@cartuja.us.es> wrote in message
news:3DF5E94B...@cartuja.us.es...

[Chris McKillop]

If you have a test case and an issue, best to post it on the QDN groups.

chris

Joaquin Ferruz Melero <fer...@cartuja.us.es> wrote:
> [-- text/plain, encoding 8bit, 29 lines --]

> Joaqu?n Ferruz
>
>
>
> [-- application/x-unknown-content-type-cfile, encoding base64, 27 lines, name: TEMP.C --]
>
> [-- text/x-vcard, encoding 7bit, 11 lines, name: ferruz.vcf --]
>

--
Chris McKillop <c...@qnx.com> "The faster I go, the behinder I get."
Software Engineer, QSSL -- Lewis Carroll --
http://qnx.wox.org/

[Joaquin Ferruz Melero]

To R. Rutherford-> I have not tested psin; I will when I have some time.
To chris-> Which of them? There are quite a few.

In any case the would-be bug isn't a much of a problem for me currently; I have found some

workarounds.

Regards,

Joaquín Ferruz

[Geoff]

I have found that psin fails in lots of scenarios. Mainly when I want to
scroll down the viewable list of processes, and when I least expect it.
Pity really as it's quite an informative utility.

I haven't been inclined to mention it here before as I'm being quite
brutal with my machines in the area of interrupts and the odd timer or
three, and it could be argued that I am provoking it.

I'm learning to live without it (psin) but I find pidin and sin lacking
in some of the information I am after about what's going on in the
system.

Geoff.

--

Realtime Technology Systems Pty Ltd
2 Hadleigh Circuit
Isabella Plains
ACT 2905
AUSTRALIA

Phone: 61-2-6291 3833
Fax: 61-2-6291 3838
Mobile: 0413 634 667
Email: ge...@rtts.com.au

[David Gibbs]

Robert Rutherford <ru...@nospamplease.ruzz.com> wrote:
> We have reported a similar problem and so have a few others. Do you also
> find that "psin" fails in this scenario?

> So far the silence from QSSL has been deafening.

There was a PRed signal/termination issue against 6.2.0 that was, I think,
fixed for 6.2.1.

I haven't further commented on this particular issue, because the source
was mime-encrypted. If it had been posted in plain, I probably would have
grabbed it and tested it.

Mime-encoding is ok for posting binary data (though I prefere uuencoded),
but C source is plain text data, and doesn't need to be mime-encoded.

-David
--
QNX Training Services
http://www.qnx.com/support/training/
Please followup in this newsgroup if you have further questions.

[Joaquin Ferruz Melero]

I correct myself... "psin" works fine until requested to slay "temp"; in this
case, "psin" will hang, probably while waiting for "temp" to exit, something
that won't happen. I cannot blame "psin" for that; the same would happen to me
If I were patient enough... :-)

Regards,

Joaquin Ferruz

Joaquin Ferruz Melero wrote:

> I have tested my program with "psin". I have found no malfunction, even
> when "temp" was hanged and it could not be killed with signals. "psin"
> reported the state "NANOSLEEP". As I said before, ps will go crazy, but only
> if requested to display the full set of data ("-l" option).
>
> Regards,
>
> Joaquín Ferruz

[Joaquin Ferruz Melero]

I have tested my program with "psin". I have found no malfunction, even
when "temp" was hanged and it could not be killed with signals. "psin"
reported the state "NANOSLEEP". As I said before, ps will go crazy, but only
if requested to display the full set of data ("-l" option).

Regards,

Joaquín Ferruz