invalid read in mpn_copyi?

[Emmanuel Thomé]

Suppressions make it possible to silence some errors by matching on
some fragment of the stack trace.

Example (from alex's example program in the thread pointed to by marc):

quiche /tmp $ cat gmp-copyid.supp
{
        partialloads_copyd
        Memcheck:Addr8
        fun:__gmpn_copyd
}
{
        partialloads_copyi
        Memcheck:Addr8
        fun:__gmpn_copyi
}
quiche /tmp $ valgrind --suppressions=gmp-copyid.supp ./test_mpmod
(no longer complains).

See also http://valgrind.org/docs/manual/mc-manual.html#mc-manual.suppfiles

I don't know if we are to regard this as too coarse or not. Certainly
if some non-sse implementation of copy[id] gets passed a pointer in
the middle of nowhere, preventing it from complaining is a bad idea.
So in such cases, the suppression file above would be coarser than
partial-loads.

E.

On Tue, Jan 21, 2014 at 10:48 PM, Niels Möller <nisse at lysator.liu.se> wrote:
> David Harvey <d.harvey at unsw.edu.au> writes:
>
>> I can see why this might be considered a non-bug, and perhaps it's even intentional for performance reasons, but it seemed dodgy enough to me to report it.
>
> If the warnings disappear with --partial-loads-ok=yes to valgrind, the
> reads are most likely intentional.
>
> It would be nice to be able to tell valgrind on a per object file (or
> per function or even per instruction) if the code is expected to do such
> "partial loads". Is that possible?
>
> Regards,
> /Niels
>
> --
> Niels Möller. PGP-encrypted email is preferred. Keyid C0B98E26.
> Internet email is subject to wholesale government surveillance.
> _______________________________________________
> gmp-bugs mailing list
> gmp-bugs at gmplib.org
> https://gmplib.org/mailman/listinfo/gmp-bugs