[CRIU] BUG: CRIU corrupt floating point state after checkpoint

Diyu Zhou zhoudiyupku at gmail.com
Fri Sep 27 19:09:16 MSK 2019


Thanks for pointing that out. It is indeed an embarrassing bug.

On Fri, Sep 27, 2019 at 6:14 AM Cyrill Gorcunov <gorcunov at gmail.com> wrote:
>
> On Wed, Sep 25, 2019 at 09:23:23PM +0300, Cyrill Gorcunov wrote:
> >
> > Great, thanks! So you've narrowed down the bug. Will take a look, thanks!
>
> 1) You've a bug in your program
>
> for  (thrd = 0 ; thrd < nthreads ; thrd++)  {
>    if  ((wspace[thrd] = calloc(1,sizeof(double *) * nelem)) == NULL)
>      ERRperror("calloc workspaces")
> }
>
> you allocate _pointers_ to doubles not doubles themself and then
> threat this space as full of doubles, which is obviously wrong.
>
> Actually I fixed this problem, but issue remans, so I think there
> might be other bugs in the code. Continue investigating...


More information about the CRIU mailing list