A question about my completed results.

[endless mike]

Recently, I have been doing some unneeded triple checks to help confirm or deny the stability of an overclock that I've been working on. I have manually submitted results for a few exponents to see about them matching previous verified results, and have had to adjust my ratios and voltages in my quest for overclocked stability. My questions is about the two exponents listed here:
42195079 and 42196381
If you look at these results, you can see that I submitted matching residues confirming their previous first time checks towards the beginning of January (8th and 9th). These were assigned to me and both done on EM_Core_i5_2320. I then manually added them to the machine I was trying to stress test and submitted mismatching residues through the manual results page January 31st. These two results were marked bad, and show up as such on my results page. I then adjusted my overclock and re-ran the same exponents, getting results that matched the previously verified tests. These were submitted on February 9th. They show up as verified on each exponents' results page, but on my overall results page, they show as bad.

Manual testing	42195079	C - Bad	        2016-02-09 15:03	0.0	6DEEC52B179568__	73.2553
Manual testing	42195079	C - Bad	        2016-01-31 22:25	0.0	CAC668B940DC3D__	73.2553
EM_Core_i5_2320	42195079	C - Verified	2016-01-09 03:30	14.5	6DEEC52B179568__	62.1938

Manual testing	42196381	C - Bad	        2016-02-09 15:03	0.0	F56E2EAFA97D05__	73.2576
Manual testing	42196381	C - Bad	        2016-01-31 22:25	0.0	728C125569B308__	73.2576
EM_Core_i5_2320	42196381	C - Verified	2016-01-08 21:50	14.2	F56E2EAFA97D05__	62.1957

[Dubslow]

Tis naught but a known bug in how PrimeNet computes if a result is bad or not. Madpoo...?

[fivemack]

I thought there was a policy that multiple checks by the same person didn't count.

[Dubslow]

Originally Posted by fivemack

I thought there was a policy that multiple checks by the same person didn't count.

They don't count towards marking the exponent as verified, but that's not the policy in question here.

What's in question is that on the personal LL results page, it erroneously shows that the second run by the same machine is incorrect, even though said second run was in fact correct (even though the first one wasn't). Essentially, if a machine turns in a bad result on an exponent, all subsequent re-tests by that particular machine are marked as "bad" in this particular report, even if the secondary tests were in fact good (and the exponent status page correctly shows them as good).

Such a situation may arise as in this and the other linked thread, where the person in question reruns the same exponent to verify that the machine has been fixed since the first faulty test.

[Madpoo]

Originally Posted by Dubslow

Tis naught but a known bug in how PrimeNet computes if a result is bad or not. Madpoo...?

I'll take a look. It's probably something to do with how the data is pulled out of SQL into a PHP array... I'll see what I can make of it.

[Madpoo]

Originally Posted by Madpoo

I'll take a look. It's probably something to do with how the data is pulled out of SQL into a PHP array... I'll see what I can make of it.

It's actually a problem in the SQL query itself. An implicit join between tables that *should* be matching on the residue as well, apparently.

I've worked up a kind of fix for it, just need to test it.

EDIT: Try this page and see if it looks better:
http://www.mersenne.org/results/default.mock.php

I worked in matching up the residue itself between different tables... it should be better.

[endless mike]

Originally Posted by Madpoo

I've worked up a kind of fix for it, just need to test it.

EDIT: Try this page and see if it looks better:
http://www.mersenne.org/results/default.mock.php

I worked in matching up the residue itself between different tables... it should be better.

That test page looks good, results like I was expecting to see. As an aside, I like the mismatched category, it's quite useful for the work we've done on strategic double checking.

[Madpoo]

Originally Posted by endless mike

That test page looks good, results like I was expecting to see. As an aside, I like the mismatched category, it's quite useful for the work we've done on strategic double checking.

Yeah... the "C - Mismatch" used to just show "C - Unverified" but highlighted in yellow to indicate there was a mismatch involved, but it wasn't clear what was going on... when I made the other change I decided to try and remedy that.

I'll push that new version out and call it good.