Factor submission problems [FIXED]

[kugano]

All,

I've looked long and hard at the broken factor submission script. We've ruled out code / scripting problems, but still don't know for sure what's going on. The submission script's Perl environment is behaving completely unpredictably -- for some factors, I can make it work simply by inserting a print statement in a certain spot. I've seen this behavior caused by memory corruption in low-level languages, but that shouldn't be possible in Perl. And I even rewrote a different factor submission script, from scratch, independent of Louie's code, and it still exhibited the same unpredictable behavior.

We also reverted to the old kernel, but this did not fix the problem.

We do have more things to try:

1 - Upgrade Perl. We're still running the 5.6 series, so it's severely due for an upgrade anyway. This isn't a trivial upgrade, though, with the modules we're using.

2 - Rewrite the factor submission script without the use of an external verification program. The "random" behavior only comes out when Perl tries to invoke an external process. There are large-integer modules for Perl we can use to avoid having to start a new process. So far I haven't seen any problems or strange behavior with internal Perl code.

Unfortunately I can't do either of these tonight, but we should be able to do #1 by the weekend. If we're still having problems then, we'll look at rewriting the script.

Until then, keep track of your unverified factors so they can be submitted later. You can keep trying the script if you want; the worst it can do is reject them.

Sorry!

[Death]

may be you should use something like php or C++?

[kugano]

may be you should use something like php or C++?

Maybe not.

[Death]

Originally posted by kugano
Maybe not.

I just asked. I wish that this can help.... =(((

[Frodo42]

I have at least one factor that may be passed during the weekend (I'm not quite sure at which speed the tests are dealt out at the moment).

Is it possible to add these few factors, mostly found using P-1, manually to the database. It is afterall quite a lot of work that is saved with every factor.

[Nuri]

Originally posted by Frodo42
I have at least one factor that may be passed during the weekend (I'm not quite sure at which speed the tests are dealt out at the moment).

Did you try submitting that factor seperately? (that one only). May be it will work.

[hc_grove]

Originally posted by Nuri
Did you try submitting that factor seperately? (that one only). May be it will work.

None of the factors I've tried submitting today worked, whether I tried submitting them sperately or more at a time. Among those were:

181022579901641 | 19249*2^5530478+1

which is already in the active window (next n -> next n+200k), but probably a week from being tested.

[kugano]

I just asked. I wish that this can help.... =(((

Sorry! You caught me in a grumpy mood this morning. I should probably avoid reading messages until I'm fully awake and able to exercise social restraint.

Anyway, for now, we're sticking with Perl. I strongly dislike PHP for many reasons that aren't really applicable to the discussion, and C++ is widely regarded as an inappropriate language for interactive web development. The bottom line is that Perl should work, and the fact that it isn't working means there's a problem with our installation. The language itself isn't at fault. We just have to find out what part of Perl we broke and fix it. Or upgrade to 5.8, which is a really good idea anyway!

[hc_grove]

Originally posted by hc_grove
None of the factors I've tried submitting today worked, whether I tried submitting them sperately or more at a time.

I have also tried submitting a few factors today, none of those I tried submitting in the usual submission form worked, but the one found using p-1-factoring luckily was big enough to use the largesieve submission form which worked.

[Mystwalker]

Originally posted by hc_grove
the one found using p-1-factoring luckily was big enough to use the largesieve submission form which worked.

That's a great idea for a short-term solution:
David, could you (or Louie) change the largesieve script so that it accepts lower factors as well?
Maybe we can bypass the strange behavior for the time being and allow everyone (including me ) to submit urgent findings.

[kugano]

Quick update on the factor submission problem. We still need to try upgrading to Perl 5.8. I had wanted to have that done by now, but I've had to spend a lot of time trying to get my car fixed -- coolant leaking into the engine block is a really nasty thing to have happen to a car. To buy a new car or not to buy a new car, that is the question... Oh well.

Anyway, I'm still hoping to get to 5.8 sometime soon. But I know people are anxious to get their factors in, so if anybody wants to email me their unsubmitted factors, I'll get them in from the console. My email address is danorris@seventeenorbust.com. I'm sure I don't need to remind everyone to be honest and not email me someone else's factors claiming to be their own ;-)

[kugano]

I *think* the problems are fixed. I have no idea what was wrong, but the upgrade to Perl 5.8.2 (along with the requisite upgrade to Apache 1.3.29 and mod_perl 1.29 and all new Perl modules) seems to have just made it go away. Please post here or email me if anyone still has problems submitting certain factors.

If you emailed me with your factors, you should have gotten a response that I added them. If you didn't, you can email me again or, preferably, try submitting them from the web script.

[MikeH]

Great news, just tried it, works fine.

13 of 13 verified in 1.25 secs.
13 of the results were new results and saved to the database.

[Nuri]

Works here too.

105 of 105 verified in 9.34 secs.
101 of the results were new results and saved to the database.

[Death]

7 of 7 verified and submitted.
Seems the problem is fixed, but still curios what was this.