[QUOTE=ATH;422818]What was the "version 17 shift bug disaster"? I think I heard there was once a prime made public that turned out to be a bug?[/QUOTE]

It was a bug that caused faulty residues when dealing with exponents of certain sizes (I forget the details). v17 results had to be tossed out, and even when you upgraded to the fixed version I think you had to start over.

I remember searching in the data for any potential exponents where both verifications were done by the buggy version and fortunately didn't find any, so it was handled pretty well at the time.

[QUOTE=ATH;422817]Here is the list of old Mersenne Prime threads if anyone is feeling nostalgic:
...
2003-06-01: M16811549 False alarm M40: [url]http://www.mersenneforum.org/showthread.php?t=672[/url]
...[/QUOTE]

I just looked through that... I never saw them mention what the actual false positive exponent was. I looked in the database and didn't see it in there either. Was it ever actually announced?

What's interesting is the fact that I still see however many false positives per year, whereas in that thread it sounds like these events were considered unexpected, weird, and should be really rare if they occur at all. Hmm...

I have a guess on the exponent based on the available info: exponent range, and the fact that one of the verifying results in June 2003 was from George and the other is from Guillermo. But the false positive result itself is not in the data. It may have been scrubbed.
[URL="http://www.mersenne.org/M16811549"]M16811549[/URL]

If I understand correctly prime has been found, and it is verified. If so, why hurry? Day or 10 days will not change anything. :smile:
Wait with patience :)

[QUOTE=pepi37;422833]Wait with patience :)[/QUOTE]

Indeed. And "Manage the Message".

After all, if done correctly, this could attract new and serious LL'ing workers which will help find the /next/ MP faster than we might otherwise.

Our last MP announcement was a great success; perfectly timed. :smile:

Why is this info like this appearing at Chris Caldwell page?
 

[URL="http://primes.utm.edu/primes/status.php"]http://primes.utm.edu/primes/status.php[/URL]
[QUOTE]
Before primes are added to the List of Largest Known Primes, they must be first be verified, comments must be checked and they must meet the size requirements. Below we show the status of these primes (if any) that are awaiting verification (of any age)[COLOR="Red"][U][COLOR="DarkRed"] as well as those modified (for any reason) in the last 72 hours.[/COLOR][/U][/COLOR] ([I]highlighted by me[/I]) Click on the prime's id for more detailed information. The color code is at the bottom of the page.
id prime digits who when comment
111120 257885161 - 1 17425170 G13 Feb 2013 Mersenne 48? (**)
85527 243112609 - 1 12978189 G10 Aug 2008 Mersenne 47? (**)
88847 242643801 - 1 12837064 G12 Jun 2009 Mersenne 46? (**)
120897 1365 · 21931200 + 1 581353 L1134 Jan 2016
[/QUOTE]

In "Status" these numbers are presented as if they are about to come (or have come in recently) into the list.

Mersenne 48, 47 and 46 are, as we all know, not recent at all.



Can anyone give a reasonable explanation?
I suspect it has to do with the new prime, but unless the corresponding exponent is smaller than 42,643,801, I can't image why would that be.

[QUOTE=Madpoo;422831]I just looked through that... I never saw them mention what the actual false positive exponent was. I looked in the database and didn't see it in there either. Was it ever actually announced?[/QUOTE]

Yes, it was 16811549 as I wrote. I was only said to be +/- 100k of 2^24=16777216 in that false M40 thread, but it was revealed later in the M41 thread in post #166: [url]http://www.mersenneforum.org/showthread.php?t=2475&page=10[/url]

[QUOTE=rudy235;422837][URL="http://primes.utm.edu/primes/status.php"]http://primes.utm.edu/primes/status.php[/URL]


In "Status" these numbers are presented as if they are about to come (or have come in recently) into the list.

Mersenne 48, 47 and 46 are, as we all know, not recent at all.



Can anyone give a reasonable explanation?
I suspect it has to do with the new prime, but unless the corresponding exponent is smaller than 42,643,801, I can't image why would that be.[/QUOTE]

They're present there because there was some sort of update in the last 72 hours. From [URL="http://primes.utm.edu/primes/page.php?id=111120"]M57885161's[/URL] page, for example, you can see the update is this comment:

[QUOTE] Chris Caldwell writes (16 Jan 2016):
Every Mersenne number less than this one has been checked once for primality by GIMPS (by October 4, 2015), but not twice. Since single checks have a relatively high error rate, this may not be the 48th Mersenne prime by size. We will remove the question mark when GIMPS completes checking all lower exponents twice and comparing the residues.[/QUOTE]

This is old news, but apparently Chris was getting the page up to date. (he probably knows it'll have a new Mersenne on it soon)

[QUOTE=Mini-Geek;422841]


This is old news, but apparently Chris was getting the page up to date. (he probably knows it'll have a new Mersenne on it soon)[/QUOTE]

Old news indeed.:rolleyes:

[QUOTE=ATH;422818]We will never know what caused prime95 to generate a false positive[/QUOTE]

This was the first time prime95 gave a false positive. After a dozen or so more such cases and several years, I did find the bug that can cause prime95 to generate a false positive.


[QUOTE=ATH;422818]What was the "version 17 shift bug disaster"? I think I heard there was once a prime made public that turned out to be a bug?[/QUOTE]

V17 was the first version that supported shift counts. Due to a 32-bit overflow bug, all results for exponents above 2^22 had to be tossed. Several weeks of LL test were discarded.

GIMPS has never been made public a prime that later turned out to be composite. That's what the independent verifications are for!

[QUOTE=Prime95;422853]This was the first time prime95 gave a false positive. After a dozen or so more such cases and several years, I did find the bug that can cause prime95 to generate a false positive.[/quote]

I don't seem to recall a thread about that - could you point me to one, or give a brief summary? (I just spent some time re-reading the original 'M40' thread ... lots of interesting history in there, including my first failed attempts at multithreaded code.)

[quote]GIMPS has never been made public a prime that later turned out to be composite. That's what the independent verifications are for![/QUOTE]

Since AFAIK numbers can't change their character I think the real worry is making public a composite with the mistaken claim of it being prime, but I know that's what you meant to say. :P

Remind me - what time Tuesday does the PR embargo officially get lifted?

[QUOTE=ewmayer;422859]I don't seem to recall a thread about that - could you point me to one, or give a brief summary? (I just spent some time re-reading the original 'M40' thread ... lots of interesting history in there, including my first failed attempts at multithreaded code.)



Since AFAIK numbers can't change their character I think the real worry is making public a composite with the mistaken claim of it being prime, but I know that's what you meant to say. :P

Remind me - what time Tuesday does the PR embargo officially get lifted?[/QUOTE]

IIRC, 0900 US Eastern time (UTC-5), so 1400 UTC. That translates to 0600 in your neck of the woods, Ernst.