On Oct 2, 1:58=A0pm, Alexander Chemeris <alexander.cheme...@[EMAIL PROTECTED]
>
wrote:
> On Oct 1, 9:53=A0pm, "Dmitriy V'jukov" <dvyu...@[EMAIL PROTECTED]
> wrote:
>
> > I am always ready to consider all suggestions. Can you provide example
> > how execution history must look according to your proposal?
> > Btw, I have some items related to execution history in my todo list:
> > =A0- for load operation output which store writes value returned by
the
> > load operation.
> > =A0- output happens-before matrix for all operations
> > =A0- output what operations cause data race
> > What do you think? Do you find those improvements useful?
>
> As an other (not so active) user of Relacy I would like to say
> that all this three bullets are very im****tant. Having they
> implemented
> reading of Relacy logs will become much easier. Now you
> have to manually trace back the cause of data races, etc,
> while after these improvements will be done, the reasons of
> some errors should be clear right after the first look.
What about points 1 and 3 - I agree. But what about point 2? Is it
will be really useful?
It will require substantial amount of code, and will consume
substantial amount of output.
Provided that we have particular execution history which leads to
error, do we need happens-before matrix?
Dmitriy V'jukov
|
