Re: [PATCH] eal: replace rte_panic instances to return an error value

[Thomas Monjalon <thomas@monjalon.net>]

20/03/2018 23:42, Arnon Warshavsky:
> > Thanks for working on this important topic.
> 
> With pleasure :)
> 
> > My feeling is that we could replace most of them by a log + return.
> > I did not think you would add a new macro. Why you chose this way?
> 
> This was meant to keep the code shorter, and imply to the reader that this
> return is actually meant to be fatal
> >
> > >   I would like to define a device health state that can be monitored from
> > >   the side,and this will be an independant patch.
> >
> > You mean when a device become unusable?
> 
> Yes. Obviously not a simple issue, but essential for refraining from panic
> in the interrupt/data-path context,
> while allowing to detect and execute on the slow/management path.
> 
> > > - Some previously panicing void functions where changed to return a
> > > value, with callers modified accordingly.
> >
> > If the function is exposed to the application, I think it is an ABI change
> > and should follow the deprecation process.
> 
> In this case I thought there would be no actual change for the user as the
> transition is from returning void to int,
> and existing calls should continue to behave as before (except for not
> crashing)

You are talking about API, and I agree the old applications can keep
considering the functions as void.
But I was talking about ABI, meaning: can we use an old application
without recompiling and update only the DPDK (in .so file)?