Akos requests to stop using Patch files - READ ONLY

ForumsCruncher's Corner

Gary Roberts
Gary Roberts
Joined: 9 Feb 05
Posts: 3781
Credit: 3540273903
RAC: 3922553
Topic 191433

Only Akos to post here, please, if he has any further information to announce regarding the removal of patched apps and the restoration of the standard app.

PLEASE NOTE:

Akos has requested that we stop using patched applications.

If you have been using applications patched using any of the patch files released by Akos over the last few days, you should, as quickly as possible, disable the patched app and restore the standard app. How you do this will vary depending on your circumstances and whether or not you were using the anonymous platform mechanism (app_info.xml).

I don't believe that it's really necessary to trash your entire cache, particularly if you were just using the patched version 4.02. However if you were using 4.10 and app_info.xml, things are a bit more tricky. If you are prepared to assume full responsibility yourself, you could suspend EAH, abort the current result and edit client_state.xml to allocate your remaining 4.10 results back to 4.02. These are not the only things you would need to do but I'm assuming people should be able to figure out the basics like setting no new work and stopping and starting BOINC, etc, at the appropriate times.

In both cases, you would need to put back the stock 4.02 app in place before restarting BOINC. I suspect that if you don't have the stock app already there, BOINC will notice and will trash the current work, then download a fresh copy of the app and then download new work. So I suspect you will waste your current work anyway if you don't have the app there, in place, first. You did keep a copy of the pristine stock app, right?? :).

As a personal comment, people need to understand the rigorous nature of the scientific methodology. If you are going to present ground breaking discoveries supported by mathematical analysis of carefully gathered data, you really need to be 200% confident in exactly what mathematical analysis was done, exactly how it was done, and precisely what checks and balances were in place to guarantee against any possibility whatsoever of spurious results. If you aren't able to fully account for all the details of how you arrived at your answers, peer group review will simply throw your results out the window. And then you could really complain about energy wastage!!

Some really silly things are being said. We should all take a huge chill pill and wait for those in charge of the project and its science outcomes to advise us of how best to proceed from here. In the meantime, fix your computers and keep crunching. Just keep in mind that you are still crunching approx 3.5 times faster than you were 6 months ago. Project staff have always said that they expect further speedups to come after further work to refine the app. We just have to learn to be patient.

PS: I'm leaving some of this thread intact as a record of a few things that happened along the way. People should ignore any links to patches that might still remain.

Cheers,
Gary.

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 560
Credit: 4410312
RAC: 0

Last stable

Message 40165 in response to (parent removed)

Last stable version:

S5S0003.DAT

Patcher program:

patcher

Gary Roberts
Gary Roberts
Joined: 9 Feb 05
Posts: 3781
Credit: 3540273903
RAC: 3922553

An important point has been

An important point has been raised by Ananas over in "Problems and Bug Reports". We have just seen the situation where the 0004 patch was found to produce invalid results. This could be picked up early by noticing that the message "Checked, but no consensus yet" was given immediately the second result arrived. In your results list this would show as "pending" until such time as a third result was sent out and returned to give the final verdict. Obviously this could take a while.

It is most likely that such a third result would be sent to a machine running the stock app and the correct decision would be made. But what happens, if by chance, the third result is also sent to a machine running the 0004 patched app. Then, because the two "wrong" results would most likely agree with each other, the wrong result would be validated and the science would be adversely affected. Fortunately we seem to have caught the problem with 0004 pretty quickly and hopefully all people running it are paying attention and have disabled it quickly.

To protect against this happening again, I think that new releases should be tested first by a very small group of willing testers and only released to the general public for broader testing after validated results have been seen.

If you have opinions about this please don't post here. Use the thread called "S5 patches - Proposal for a very small alpha test group" to make your comments about this. After opinions are aired, we might call for volunteers and Akos can choose who he wants.

EDIT: Added links.

EDIT2: Getting people to slow down and take a bit of care is harder than trying to stop the universe expanding. I should have known better when the feeding frenzy started :). I just hope that people pay close attention and that the integrity of the project isn't compromised.

Cheers,
Gary.

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 560
Credit: 4410312
RAC: 0

Last stable

Message 40171 in response to message 40165

Last stable version:

S5S0007.DAT

Patcher program:

patcher

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 560
Credit: 4410312
RAC: 0

Hi people! Please, don't

Hi people!

Please, don't use any patched versions of Einstein@Home app.
The bad results caused some problems in the database.

Thanks!

Gary Roberts
Gary Roberts
Joined: 9 Feb 05
Posts: 3781
Credit: 3540273903
RAC: 3922553

^Bump^

^Bump^

Cheers,
Gary.

roadrunner_gs
roadrunner_gs
Joined: 7 Mar 06
Posts: 94
Credit: 3362498
RAC: 0

RE: ^Bump^ Errr.... For

Message 40180 in response to message 40179

Quote:
^Bump^

Errr.... For what do you need a bumper in a sticky thread?
Im puzzled.

Gary Roberts
Gary Roberts
Joined: 9 Feb 05
Posts: 3781
Credit: 3540273903
RAC: 3922553

RE: RE: ^Bump^ Errr....

Message 40182 in response to message 40180

Quote:
Quote:
^Bump^

Errr.... For what do you need a bumper in a sticky thread?
Im puzzled.

Don't be puzzled - there's method in the madness :).

Many people set up their forum preferences to flag a thread when there are new messages. There wasn't a new message but I had just edited the first post to bring to people's attention, Akos' request to stop using patched apps. I wanted people to be aware of the edited information so I added a dummy new post.

Cheers,
Gary.

Pepo_2
Pepo
Joined: 17 Aug 05
Posts: 15
Credit: 458309
RAC: 0

RE: RE: RE: ^Bump^ Errr

Message 40183 in response to message 40182

Quote:
Quote:
Quote:
^Bump^

Errr.... For what do you need a bumper in a sticky thread?

Don't be puzzled - there's method in the madness :).

Many people set up their forum preferences to flag a thread when there are new messages. There wasn't a new message but I had just edited the first post to bring to people's attention, Akos' request to stop using patched apps. I wanted people to be aware of the edited information so I added a dummy new post.


Garry, my browser jumped to "first unread message in the thread", exactly to your ^BUMP^, and your first edited post was left over the screen, If roadrunner_gs would not ask, I'd never imagine that there is some very new and important update to any older post (breaking the usual 1 hour barrier :)

Maybe you might note such updates in your bumps :-)

Peter

Gary Roberts
Gary Roberts
Joined: 9 Feb 05
Posts: 3781
Credit: 3540273903
RAC: 3922553

RE: Maybe you might note

Message 40184 in response to message 40183

Quote:


Maybe you might note such updates in your bumps :-)

Peter

Yes, you are quite right. It would have been a lot smarter to have stated "Please go read the first post for some important new information".

However, I was in a hurry at the time and don't claim to be particularly smart. It was the first thing that popped into my head to quickly bring the thread as a whole to people's attention.

I'm sorry for stuffing it up.

Cheers,
Gary.

_AF_HFR_ ToOm
[AF>HFR] ToOm
Joined: 11 Feb 05
Posts: 3
Credit: 753076
RAC: 0

Lot of users are still using

Lot of users are still using Akosf opti...

What about them ??
Is it really bad for the project ?
Can you send them personal message or something..?

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.