> This looks good to me. And just to make sure I'm understanding properly,
> OPEN_REL only valid if it is within
> or contained in a sub-directory tree of the ORIG_CWD.
That is correct.
> One other question, do we get into any trouble with regards to the CWD during
> restore?
>
> Example in program:
> chdir("./subdir");
> system("some_other_program"); // Checkpoint occurs while some_other_program
> is running
>
> So the CWD of some_other_program during run is ./subdir. When we restore
> (assuming same location),
> does that mean that some_other_program thinks that its CWD is ./subdir or .?
> I believe if it's not
> in ./subdir, then we could run into some strange behavior if it thereafter
> tries to open files with relative paths.
This is an excellent point. DMTCP currently doesn't restore the CWD.
We should come up with similar heuristics for restoring CWD as well. I
think the ones for files can act as a decent guide for now.
Kapil
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_mar
_______________________________________________
Dmtcp-forum mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
