Automatic Submission Spider for Workers...

[chalsall]

Originally Posted by flashjh

You're on the right track, as this was discussed before. I just haven't found a way to get the spider to do it correctly.

Is that all it would take to help PrimeNet understand the method used to find the factor? Easily done.

It is time I climbed back into that Spider's code, and bring it up to 0.3. Several people have pointed out bugs, and I've noticed a few myself (for very special cases).

In order to test the "multi-line" submission requirement, I would need some sample data to work with. So could I ask one or more people to please PM me an example of a non-factor found with mfakt* (one line), and then another with a factor found (multi-line).

Please only provide me with examples which have already been submitted to PrimeNet. I'll then have one of my machines find a cheap factor, and will simulate the submission with the factor I find using a CPU. (Amusingly, I don't myself have a GPU. Or, at least, not any which are not busy doing other work.)

[kladner]

Originally Posted by chalsall

In order to test the "multi-line" submission requirement, I would need some sample data to work with. So could I ask one or more people to please PM me an example of a non-factor found with mfakt* (one line), and then another with a factor found (multi-line).

Please only provide me with examples which have already been submitted to PrimeNet. I'll then have one of my machines find a cheap factor, and will simulate the submission with the factor I find using a CPU. (Amusingly, I don't myself have a GPU. Or, at least, not any which are not busy doing other work.)

Do I understand correctly that you want results.txt lines with examples of "no factor for Mxxxxxxx", and the pair starting with "Mxxxxx has a factor:..."?

Would an example of a split submission of a "...has a factor" pair be helpful? My memory is that this resulted in a "partial credit given" message when the first line was submitted without the second. I wanted to see what PrimeNet says about the exponent, but then remembered it's down for maintenance.

[chalsall]

Originally Posted by kladner

Do I understand correctly that you want results.txt lines with examples of "no factor for Mxxxxxxx", and the pair starting with "Mxxxxx has a factor:..."?

Yes. I've already received an example, but additional ones would be useful.

Originally Posted by kladner

Would an example of a split submission of a "...has a factor" pair be helpful? My memory is that this resulted in a "partial credit given" message when the first line was submitted without the second. I wanted to see what PrimeNet says about the exponent, but then remembered it's down for maintenance.

As many examples as the Spider will ever encounter would be useful.

Thanks.

[flashjh]

Originally Posted by chalsall

Is that all it would take to help PrimeNet understand the method used to find the factor? Easily done.

It is time I climbed back into that Spider's code, and bring it up to 0.3. Several people have pointed out bugs, and I've noticed a few myself (for very special cases).

In order to test the "multi-line" submission requirement, I would need some sample data to work with. So could I ask one or more people to please PM me an example of a non-factor found with mfakt* (one line), and then another with a factor found (multi-line).

Please only provide me with examples which have already been submitted to PrimeNet. I'll then have one of my machines find a cheap factor, and will simulate the submission with the factor I find using a CPU. (Amusingly, I don't myself have a GPU. Or, at least, not any which are not busy doing other work.)

I'm no expert on this, but the discussion started here. Once you get to #1235, it shows that it has something to do with the 'no factor' line.

Good discussion, nothing has come out of it yet. James was working on it; I just asked him about it the other day here. It would be nice to get the spider submission correct, but I've even tried manual submission and I get the same results.

Thanks for working it!

[Dubslow]

I must point out that currently, the spider submits exponents in exactly the same way as the manual results page. If you don't believe me, when PrimeNet is up, submit results without the spider there and look at the URL that is produced when you hit "submit". Any testing to try and fix the PrimeNet detection system can be done just the same from that webpage as from modifying the spider. As I recall, we've already tried messing with the mfakt* result line format, and nothing has gotten PrimeNet to report correctly; therefore we decided we'd have to wait for James to fix PrimeNet's result parser.

[bcp19]

Originally Posted by Dubslow

I must point out that currently, the spider submits exponents in exactly the same way as the manual results page. If you don't believe me, when PrimeNet is up, submit results without the spider there and look at the URL that is produced when you hit "submit". Any testing to try and fix the PrimeNet detection system can be done just the same from that webpage as from modifying the spider. As I recall, we've already tried messing with the mfakt* result line format, and nothing has gotten PrimeNet to report correctly; therefore we decided we'd have to wait for James to fix PrimeNet's result parser.

From your description I am guessing you are cutting and pasting? I know when I upload the fle I get correct results, broken down by program, with factors at the bottom of the page, and the URL is unchanged (meaning if I hit F5 it reloads the manual assignment page). I don't know the difficulty, but maybe have the spider upload your results file rather than send it line by line?

[Dubslow]

Hmm... well then what I said is definitely true for C/V, but I'll go check out results file upload after PN is online again...

The Spider acts as if you copy and paste one line at a time into the box and hit submit (or in the case of the factor, both lines at the same time actually, let me verify this. I think it might just report the "M%d has a factor" line. Brb. Edit: Yes, it only reports the first line of results, however even with the second line PrimeNet still fails in general at parsing it correctly, especially with the new output format in mfaktc 0.18.).

[chalsall]

Originally Posted by Dubslow

The Spider acts as if you copy and paste one line at a time into the box and hit submit (or in the case of the factor, both lines at the same time).

You are only half correct.

The spider submits one line at a time to the manual results page. Trust me on this; I wrote it.

What has been pointed out is perhaps it would be better if all the result lines related to a particular candidate were submitted at the same time.

[Dubslow]

Originally Posted by Dubslow

The Spider acts as if you copy and paste one line at a time into the box and hit submit ...
Edit: Yes, it only reports the first line of results, however even with the second line PrimeNet still fails in general at parsing it correctly, especially with the new output format in mfaktc 0.18.

Reposted so that thread turns yellow and you see my later edit :P

[flashjh]

I've disabled my automatic spider submissions until after the new exponent is verified (Monday-ish). 'not_completely_submitted' files were building up. It a good and bad day all at the same time.

[bcp19]

Originally Posted by flashjh

I've disabled my automatic spider submissions until after the new exponent is verified (Monday-ish). 'not_completely_submitted' files were building up. It a good and bad day all at the same time.

Just when I got mine set up too :( At least GPU72 is up and we can get assignments if needed.