Re: [Valgrind-users] helgrind and rwlock

[Julian S. <js...@ac...>]

On Sunday 30 March 2008 16:35, Ivan Skytte Jørgensen wrote:
> On Friday 28 March 2008 02:01:31 Julian Seward wrote:
> > > Yes, it is possible. But in order to avoid false positives they cannot
> > > be online algorithms - they have to be run "after the future is known",
> > > ie. based on a trace file.
> >
> > Or at the end of execution, analysing data collected during the run.
> >
> > Do you have pointers to paper(s) describing this?
>
> Strange, I thought this was a well-researched subject, but I can only find
> three relevant papers at citeseer: [...]

On the basis that many ideas in CS have been reinvented several
times by different generations of researchers, I would not be
surprised to find that the database people had this all figured
out by about 1970 :-)  After all, concurrency and locking is a
big deal in database-world.

J