[arch-dev-public] [signoff] gmp-5.0.2-2
Upstream patch for the following bug (until 5.0.3 is release with this fix): There is a GMP bug in mpn/generic/dcpi1_bdiv_q.c that causes incorrect computation with extremely low probability, for unnormalised dividends. The bug might also result in a decrement of memory above dividends, with some potential security impact. The bug might affect several user-level routines, e.g., mpz_invert, but only for operands of at least 3000 bits (for smaller operands, a simpler algorithm is used). Testsuite fully passes. Signoff both, Allan
On 18 July 2011 14:51, Allan McRae <allan@archlinux.org> wrote:
Upstream patch for the following bug (until 5.0.3 is release with this fix):
There is a GMP bug in mpn/generic/dcpi1_bdiv_q.c that causes incorrect computation with extremely low probability, for unnormalised dividends. The bug might also result in a decrement of memory above dividends, with some potential security impact. The bug might affect several user-level routines, e.g., mpz_invert, but only for operands of at least 3000 bits (for smaller operands, a simpler algorithm is used).
Testsuite fully passes.
Signoff both,
Since the test suite passes, signoff both. :)
On Mon, Jul 18, 2011 at 09:51:36PM +1000, Allan McRae wrote:
Upstream patch for the following bug (until 5.0.3 is release with this fix):
There is a GMP bug in mpn/generic/dcpi1_bdiv_q.c that causes incorrect computation with extremely low probability, for unnormalised dividends. The bug might also result in a decrement of memory above dividends, with some potential security impact. The bug might affect several user-level routines, e.g., mpz_invert, but only for operands of at least 3000 bits (for smaller operands, a simpler algorithm is used).
Testsuite fully passes.
Signoff both, Allan
Signoff x86_64. dave
participants (3)
-
Allan McRae
-
Dave Reisner
-
Evangelos Foutras