[Ecm-discuss] gmp-ecm 6.4.3: Unexpected test failure
Paul.Zimmermann at loria.fr
Lun 31 Déc 16:57:40 CET 2012
> Date: Mon, 31 Dec 2012 12:41:02 +0100
> From: =?UTF-8?B?UGF0cmljayBMYXVlcg==?= <patrick at gentoo.org>
> since gforge is rather user-hostile please excuse my direct message to you -
> the test.ecm script has a reproducible failure in only one test:
> echo 5394204444759808120647321820789847518754252780933425517607611172590240019087317088600360602042567541009369753816111824690753627535877960715703346991252857 | $ECM $BATCH0 -sigma 805816989
> -go 345551 149827; checkcode $? 6
> -> return code is 0 instead of 6
> I'd like to figure out why this testcase fails. Testing with yafu gave me a proper factorization, so it's either gmp-ecm misfiring or the test case being incorrect in some way.
> Thanks for any pointers to get this resolved,
> Patrick Lauer
I guess this was with GMP 5.1.0. We use some internal (and undocumented)
GMP functions whose functionality changed from GMP 5.0.x to GMP 5.1.0.
This is fixed with the development version. Thus either use the development
version, or use GMP 5.0.5.
In any case don't use GMP-ECM if "make check" fails.
More information about the Ecm-discuss