Mersenne.ca bug...

[Stargate38]

Look at the factor-bounds graphs at the bottom:

http://www.mersenne.ca/exponent/74207321

If you point to the red circle (which represents the 1 known factor), it shows a negative number, which, if you add to the known prime factor, you get 2⁶⁴. Is this a problem with 64-bit signed integers?

[LaurV]

Sorry for waking this up, I didn't know where to post, not that I would care much about the well ordered topics on the forum but I know James is a busy guy and he only read very few topics here around.

So, there is a small cosmetic bug on mersenne.ca site:

The http://www.mersenne.ca/exponent/57885161 would need a small manual tweak, it does not show the triple arrow going right. I think the pointers were added by hand (blearh! ) and they are not "computed" by reading the exponents DB. So, the news of the fact that we discovered a new prime escaped to the parsing guy there, hehe.

And because we are here, the dots designating "uncertain" mersenne primes in the table at http://www.mersenne.org/primes/ won't stay purple, but appear blue and change back to blue when you navigate in the table. I would like them to remain purple (or violet, whatever that color is).

[LaurV]

Nice title for a book. For a new subforum too.. But let's stay inside of this topic for now. Anyhow, this worth a new post. Related to the former one, but more interesting. I continued to play with those arrows and guess what I found:

I know how to find new mersenne primes. Proven formula!

All I have to do is to take a table of prime numbers, then James' site, start from a prime, any prime, and click the arrows in the table. Any arrow, left or right, in any numbers. And follow close with the numbers in the table, watch them tight. If any prime is skipped, i.e. I have it in the table, but the arrow jumps over it, then it is a high probability I found a new prime:

Example: M74207281. You fill any exponent in James' site, a bit lower than this (and click any right arrows) or a bit higher than this (and click left arrows), then the number is skipped, the form can't find any prime between 74207279 and 74207297, no matter what arrows you click.

But if you fill 74207281 exactly, it shows you that is prime.

For sure I have to do this procedure with exponents over 332M, watch which one is skipped, and I will find a new mersenne prime. It takes much shorter than a LL test...

[vsuite]

Originally Posted by LaurV

...And because we are here, the dots designating "uncertain" mersenne primes in the table at http://www.mersenne.org/primes/ won't stay purple, but appear blue and change back to blue when you navigate in the table. I would like them to remain purple (or violet, whatever that color is).

What dots?

{What arrows?}

[LaurV]

Originally Posted by vsuite

What dots?

Click the second link in my post, scroll down. Hover the mouse over the big blue dots in the graph. Then hover the mouse over the table, while the graph is still visible. Repeat...

{What arrows?}

Yes, that was my question too, "which arrows?"
Click the first link in my post. Please navigate to former mersenne prime, by clicking the tripple-arrow pointing left. Repeat, to see other mersenne primes. Try in the opposite direction. When you reach back the starting point, click the triple arrow pointing right, to go to the next mersenne prime... "Which arrow??".

Related my second post, click on http://www.mersenne.ca/exponent/74207281.
Yeah, voila! is prime. Click any arrows you like there on the right of the exponent.
Try to find it back, using the arrows.
Told you so!

License agreement (EULA): by reading this post, you agree to the fact that you are not allowed to use my algorithm to find new mersenne primes, as I described it in my second post.

Edit: I just sent a PM to James, I am sure he didn't see this topic (including the first message from Stargate) and he will take care of the issues as fast as his time allows, as usual.

[James Heinrich]

I don't generally browse the forum, I usually only see stuff in a handful of subscribed threads (until LaurV bothers me with PMs ). Please post any mersenne.ca related bugs in the mersenne.ca thread.

Originally Posted by Stargate38

Look at the factor-bounds graphs at the bottom:
Is this a problem with 64-bit signed integers?

Yes, unfortunately.
PHP is very good at automagically typecasting variables as needed. Unfortunately it sometimes gets it wrong and it's hard to override. In this case, the factor is stores as a string, and used as a string, but in one place it's used as a string array key. But... for small factors it looks like an integer so PHP uses it as an integer array key instead of a string key, and in that context it gets cast as a negative number when in the bitrange just below PHP_INT_MAX (either 2³¹-2³² or 2⁶³-2⁶⁴).
I gave the code a severe talking-to and told it not to do that again. I also had to prepend a non-numeric character to the array key to force it to always handle it as a string.

Originally Posted by LaurV

http://www.mersenne.ca/exponent/57885161 would need a small manual tweak, it does not show the triple arrow going right. I think the pointers were added by hand (blearh! ) and they are not "computed" by reading the exponents DB.

Not added by hand (that would be silly ) but a 1-char typo had it reading the `prime` table instead of the `primes` table it was supposed to. Fixed now.

Originally Posted by LaurV

the dots designating "uncertain" mersenne primes in the table at http://www.mersenne.org/primes/ won't stay purple, but appear blue and change back to blue when you navigate in the table. I would like them to remain purple (or violet, whatever that color is).

#7F2FFF if you must know
I thought we were talking about mersenne.ca bugs here, but I have fixed that for you anyways

Originally Posted by LaurV

Any arrow, left or right, in any numbers. And follow close with the numbers in the table, watch them tight. If any prime is skipped, i.e. I have it in the table, but the arrow jumps over it, then it is a high probability I found a new prime

Well of course. The arrow buttons are generated for next exponent with/without factors that are not prime. So I have to query the list of Mersenne primes (I have the full list up to M(2³²) of course) and skip those ones.

[LaurV]

Stargate's bug does not seem to be fixed, still shows the negative number. Additionally, now it shows a "NaN" instead of the number of bits for the factor. This is in tooltip when you hover over the big dot in the factors table.

The violet dots in the primes graph of PrimeNet is fixed - it looks very good now!

The arrow links are not fixed. They work well up to M48, but the tripple-right of M48 is still not showing, and also for the next prime exponent after the M48. It starts appearing again after the second next prime exponent after M48, when you can click it and go to M49. Around M49, forward navigation is ok, backward navigation still skips M49.

Thanks a million for your efforts to sort out these small cosmetic things. We really appreciate!

[James Heinrich]

Originally Posted by LaurV

Stargate's bug does not seem to be fixed
The arrow links are not fixed

I think they are, but you have not sufficiently refreshed your browser. The mersenne.ca exponent pages are quite forcibly cached, so you'll need to Ctrl-F5 forcibly-refresh your browser to see the new, fixed pages.

[vsuite]

Oh no. The Rapid Prime Finding Feature (RPFF) has been disabled. So we have to use the Leisurely {prime} Locater (LL) instead.

[LaurV]

Yeah, you killed my algorithm... Bad James!

{ do I have to hit Ctrl-F5 to every exponent? it seems so...}

{yes, it does that, if I visited the page in the past... hmmm... I just killed all the history for all the pages ending in .ca, and it seems to work very well now}

[petrw1]

Above the GREEN line is what I entered just before I clicked on "Calculate" on the Left.

Below the GREEN line is the output I got … NOTE:
"Use B1/B2 above" was changed to "Save 2-LL Tests"

It didn't do that a few weeks ago.