[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[nptl] Re: OSDL Bug 3770
> That would be not conform to SUS. printf() is line buffered, the "\n" at the
> end ensures that the line is flushed. I believe, if you do the test unter a
> normal text console (doing under X might be more tricky), you should see at
> least that the alarm thread is started. IMHO.
Well, I'll have to check then ;) I did the testing through a serial
line...
> > To know where the code really hangs, I would need to add some "fflush
> > (stdout);". But I would prefer to do this against a testcase which does
> > not require I power-cycle the box each time... Can you provide me with
> > this new version?
>
> >From my opinion, the fact that you need to hard-reboot the machine is
> already a sign that something is wrong with the patch... Or perhaps with my
> test. Note however that this test has been now tested on a wide range of
> platforms/OS (OK, that's by no means a proof for correctness).
Well, the ctrl-C is ignored. I thought this come from that the keyboard
handling mechanism may have a lesser priority than the test program? (I
think this is impossible, but...)
> By the way, you didn't answered my question: Do you have the same effect,
> without patch?
Oops sorry. No, without the patch the behavior has been the same since
we raised this problem: some threads do not execute.
> I believe, doing the way you're proposing would result another test case. In
> each thread you would have to periodically trap to the kernel to get the
> current clock. You wouldn't have a pure compute bound thread anymore...
Sure -- or maybe have a sample computation, and then scale the
computation to last about 60 seconds...
> I have right now a hand full of work, and I am not exactly sure when I can
> design that new version. But perhaps you have time?
If you give me a design, I may try to code it, but I don't really know
what is the good way to do this ?
Thanks,
Seb.
--
-------------------------------
Sebastien DECUGIS
NPTL Test & Trace Project
http://nptl.bullopensource.org/
"You may fail if you try.
You -will- fail if you don't."
Liste de diffusion nptl
Pour se désinscrire : mailto:nptl_request@bullopensource.org?subject=unsubscribe