Re: [Valgrind-developers] helgrind

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

On Monday 07 October 2002 1:14 am, Jeremy Fitzhardinge wrote:
> Can go into a few more details about what's wrong with helgrind?  Does
> it get the Eraser algorithm wrong, or does it just need some more
> refinement?
>
> I read the Eraser paper and the extension suggested in "Visual Threads",
> and it all looks pretty straightforward.  Once I get some profiling
> machinery going, I'm definitely interested in working on helgrind.

It generates bazillions of errors.  We don't know why.

The only _known_ cause for tyhis is that glibc has a counter for
the number of relocations done by the dynamic linker, and this is
incremented without locking.  That's allegedly harmless because
it's only for stats purposes, but it's still something that needs
to be suippressed or worked around.  Problem with suppression is
that there is no fixed call stack at the error point; it happens
at the first use of every dynamic symbol (or something) so the
normal stack-based-identification of the suppression mechanism
won't work.  We haven't thought of a way round this yet.

If you want to investigate that would be great.  Having a usable
race detector would be really excellent.

J