Oops

fivemack · 2017-11-30, 15:56

If a relations file is corrupt in such a way that it contains '1' as a putative factor, current-SVN msieve goes into an infinite loop in mp_is_prime_1.

Changing line 133 to reject p==1 fixes this

fivemack · 2017-11-30, 17:30

Sorry, misdiagnosis: the file has to contain a line starting '1,0:' to trigger the problem (infinite loop in mp_is_prime_1 called from gnfs/relation.c:133)

As far as I can see, with lpb=33, composite factors larger than 2^32 will not get rejected; I don't know if this has been an issue in actually existing jobs.

2017-11-30, 15:56	#1
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 7²×131 Posts	Oops If a relations file is corrupt in such a way that it contains '1' as a putative factor, current-SVN msieve goes into an infinite loop in mp_is_prime_1. Changing line 133 to reject p==1 fixes this

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Oops - New Prime! (M49 related)	Prime95	News	631	2017-11-10 20:35
Merging Computers OOPS	petrw1	PrimeNet	2	2017-04-15 03:24
Oops (need help recovering factorization)	Dubslow	YAFU	6	2015-02-26 01:32
Oops i did it again. (Prime found)	ltd	Prime Sierpinski Project	21	2006-01-04 14:50
Oops!	Xyzzy	Lounge	4	2003-01-06 22:18

2017-11-30, 17:30	#2
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 7²×131 Posts	Sorry, misdiagnosis: the file has to contain a line starting '1,0:' to trigger the problem (infinite loop in mp_is_prime_1 called from gnfs/relation.c:133) As far as I can see, with lpb=33, composite factors larger than 2^32 will not get rejected; I don't know if this has been an issue in actually existing jobs.