Re: [Valgrind-users] Suppression bug with Memcheck:Value16

[Nicholas N. <nj...@ca...>]

On Wed, 11 Jun 2003, Nicholas Nethercote wrote:

> > Figuring I had the supression rule wrong, I reran with
> > --gen-suppressions=yes.  However, when valgrind hit this event in
> > the code, it crashed and said:
> >
> > ==2336== Thread 2:
> > ==2336== Use of uninitialised value of size 16
> > ==2336==    at 0x40D64260: __nvsym04804 (in
> > /usr/lib/libGLcore.so.1.0.3123)
> > ==2336==
> > ==2336== ---- Print suppression ? --- [Return/N/n/Y/y/C/c] ----
> > Memcheck: the `impossible' happened:
> >    unexpected size for Value
> > Basic block ctr is approximately 104700000
>
> It's a bug/oversight -- Valgrind doesn't support suppressions for 16 byte
> value errors.  I will investigate, hopefully get back to you soon.

I just added support for "Value16" and "Addr16" suppressions to CVS.  I
would appreciate it if you could test the changes.

Note that three other value error sizes -- 10, 28, 108 -- are possible,
but not suppressible.  These strange sizes are for various floating point
ops, and we'll add support for suppressing them when someone needs it.

N