Hi Calvin,
Thanks for reporting this bug. To be honest, when DMTCP was growing
fast, we weren't paying enough attention to issues such as ctrl-C.
Now we expect that the DMTCP core will change only slowly, while
each plugin (including internal DMTCP plugins) can continue to be
developed separately. So, we're very interested in testing more of
these cases now.
If 2.4.0 rc2 sometimes segfaults on restart, we'd definitely like to
reproduce that locally here, and fix it.
Here are some questions that will help us to analyze this:
1. Which programming language interpreter were you testing on?
2. On the netbook, was the Atom a 32-bit CPU or 64-bit?
3. You said that on the Fedora 20 machine with 2.4.0 rc2, the
restart would _sometimes_ segfault. Given a single checkpoint image,
would that same image sometimes checkpoint on restart? Or would
the same file always segfault on restart, and a new checkpoint image
might always restart? (In other words, is there sometimes a
"bad ckpt image" that will always segfault on restart, or is it
that a single ckpt image will sometimes segfault on restart and
sometimes not?)
We greatly appreciate you reporting this, and we'd definitely llke
to have it fixed for the final 2.4.0 release. Thank you for writing.
Best,
- Gene
On Fri, Apr 24, 2015 at 12:56:05AM -0400, Calvin Ostrum wrote:
> On Thu, Apr 23, 2015 at 8:52 PM I wrote:
>
> > I want to checkpoint a programming language interpreter's environment
> > which I am using interactively. When using this environment, if an
> > expression I am evaluating turns out not to help, normally I can hit
> > control-c to interrupt the computation and return me to the
> > enviroment's user prompt.
>
> Oops, that should be "turns out not to *halt*".
>
> Some more details. I was using the dmtcp that comes with Fedora 20,
> which was version 2.2. So I tried the newest version 2.4.0 rc2, and
> it does pass the control-c as expected to the process that was
> checkpointed. However, I find that 2.4.0 rc2 occasionally segfaults
> when starting up a restart.
>
> Next I tried on my netbook (which is an Atom processor, the other
> machine is an i3 540). I first tried the 2.4.0 rc2 version with a
> default ./configure. It checkpoints the process without complaint,
> but without fail segfaults on the restart. So I tried 2.3.1, and it
> has the same problem as the 2.2 on the other machine does: it quits
> completely when it gets a control-c.
>
> Is it correct to assume that control-c never did work correctly in
> earlier versions and that is one thing that is being fixed in 2.4? I
> might assume this since I have two separate examples where the
> control-c had this behavior in pre 2.4 versions. But this seems
> unlikely since the control-c behavior is so standard and so needed by
> so many.
>
> ------------------------------------------------------------------------------
> One dashboard for servers and applications across Physical-Virtual-Cloud
> Widest out-of-the-box monitoring support with 50+ applications
> Performance metrics, stats and reports that give you Actionable Insights
> Deep dive visibility with transaction tracing using APM Insight.
> http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
> _______________________________________________
> Dmtcp-forum mailing list
> Dmtcp-forum@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
Dmtcp-forum mailing list
Dmtcp-forum@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
