Possible flaw in MPI code w.r.t. sign flag

Heiko Stamer HeikoStamer at gmx.net
Tue May 14 20:58:52 CEST 2019


Hi Werner,

On 13 May 2019 16:31, Werner Koch wrote:

> We need to handle the +-0 case first.  What about this:

I am not sure whether the fix in do_mpi_cmp() is enough. Perhaps
other comparison operators or output functions of the MPI code
have to be changed too. However, I have not looked at this yet.

Another solution to the underlying issue: avoid setting the sign
flag for that special case in all arithmetic operators (add, ...).

--
Heiko




More information about the Gcrypt-devel mailing list